Vol 2 No 4

(IJCNS) International Journal of Computer and Network Security, 1
Vol. 2, No. 4, April 2010
RD-Optimisation analysis for H.264/AVC scalable

video coding
Sinzobakwira Issa1, Abdi Risaq M. Jama2 and Othman Omar Khalifa3
1
Olympia College, School of Engineering
Persiaran Raja Chulan, 50200 Kuala Lumpur, Malaysia
Issa10issa@gmail.comgmail.com
2, 3
International Islamic University Malaysia, Department of Electrical and Computer Engineering
Jalan Gombak, Box: 53100 Kuala Lumpur, Malaysia
Abstract: The development of multimedia propagations and The reality shows that, for each second of the movie, the
applications has led to a greater expansion in the field of video requirement be at least 27MB of memory; as a result, the
transmission over a heterogeneous media as well as iterative entire movie will need almost 162GB of memory. If this
delivery platforms with dedicated content requirements. It is movie were stored on DVD’s, then considered the current
known that conventional video coding systems encode video DVD capacity of 4.7 GB, would roughly require 35 DVD’s.
content with given bitrates adapted to a specific function or Therefore, video needs to be compressed considerably for
application. As a result, conventional video coding does not meet
efficient storage and sharing over the web [1]
the fundamental requirements of the state-of-the-art flexible
digital media application. The newly technology based on
However, there are a lot of redundancies within the video
scalable video coding appears as a new modus operandi that has data that can be eliminated yielding file size reduction or
the ability to satisfy the underlying requirements. In this work, compression.
a multi-users scenario was considered for an optimum
performance between multiple streams. A rate distortion 2. H.264/AVC Scalable Video Coding
optimized video frame dropping strategy which can be applied
on active network nodes during high traffic intensity was 2.1 Basic H.264/AVC structure
developed. The concept of scalability here, come to introduce the
operability of high level of suppleness coding and decoding The H.264/AVC standard has a range of coding tools
systems. A base layer which can display the suitable quality of contributing to its high compression performance, flexibility
the premium file was considered and take care of the and robustness. However, the performance improvements
improvement of video quality. come at a cost of significantly high computational
Keywords: Bitrates, PSNR, bandwidth, multi-users scenario complexity. Therefore, encoder implementations should
and RDO. make use of the available coding tools effectively to achieve
the desired compression performance with the available
1. Introduction processing resources.
H.264/AVC is an extremely scalable video codec,
The past few decades, starting in the early nineties, a delivering excellent quality across the entire bandwidth
remarkable development has been achieved in the field of spectrum, from high definition television to the video
video compression. A lot of efforts were and still are being conferencing and 3G mobile multimedia. The following can
exerted for compressing, storing data in digital medium and thusly be summarized as the important differences.
allocation over the web. • Enhanced motion prediction capability
It is very crucial to have the idea of monochrome digital
• Use of a small block-size exact match transform
video data sequence which is a set of individual pictures
• Adaptive in-loop deblocking filter
called frames happening at predetermined time increments.
• Enhanced entropy coding methods
This frame needs to be considered as a light intensity of two
dimensions in terms of function of variable x and y; f(x, y),
where x and y denote special coordinates and the value off at
any point (x, y) is proportional of the brightness of the frame
or the gray level at the point for monochrome. The normal
standard speed at which these frames are displayed is 30
frames per second.
This representation is called canonical representative.
However, this canonical representation has negative impact
because it needs very huge amounts of memory, resulting in
impracticality of being stored or shared on the web or to be
launched into digital channel. The fact may seem as an
amusing game when we try to illustrate how it could be
done. Figure 1. H.264/AVC structure
The clear picture is an example of a 100 minutes movie
2.2 Scalable Video Coding
displayed at 30 frames per second with width of frame
640x480 pixels with each pixel taking 3 bytes of memory.
2 (IJCNS) International Journal of Computer and Network Security,
Scalable video coding is desirable in heterogeneous and Rate-distortion Optimisation (RDO) methods used in video
error-prone environments for various reasons. For example, compression are discussed in [6] [2], which include dynamic
scalable coding helps streaming servers avoid congestions in programming and Lagrange optimisation methods.
network by allowing the server to reduce the bitrate of A Lagrange optimisation method, which is also known as
bitstreams whilst still transmitting a useable bitstream. Lagrange multiplier method, offer computationally less
One application for scalability is to improve error complex (although sometimes sub-optimal) solutions to the
resilience in transport systems that allow different qualities optimisation problem was proposed. Due to its less complex
of service. nature, a specific form of the Lagrange optimisation method
For example, the essential information could be delivered has been used in rate-distortion optimisation of H.264/AVC
through a channel with high error protection. Scalability can [10].
also be used to enable different quality representations
depending on playback devices processing power. 3.2 Constrained Optimisation Problem
Devices with better processing power can decode and
display the full quality version, whereas devices having The objective function within source constraints is
lower processing power decode the lower quality version. minimized or maximized by the constrained optimization.
In the case here of video coding standards, this issue of
2.3 Types of SVC
constrained optimization can be considered as reducing the
There are three conventional types of scalability: temporal, amount of distortion of a given video sequence meaning to
quality and spatial. Temporal scalability enables adjustment strive looking to increase the number of bits that can be
of picture rate. encoded in exactly that particular coding sequence[4].
a) This is commonly carried out with either disposable Below is the mathematical representation of the constrained
pictures or disposable sub-sequences, which are optimization unit;
explained later on. Picture rate adjustment is then Let S represent all the allowable vectors and let B an
simply done by removing these disposable parts element of S, (BЄS). The objective function is defined for all
from the coded sequence thus lowering the frame B in S as D(B) and the constraint function R(B) is defined
rate. for all B in S . The constrained problem can be presented as:
b) In conventional quality scalability, also known as Given a constraint Rc, find
SNR scalability, an enhancement layer is achieved
with pictures having finer quantizers than the
particular picture in the lower reference layer[3]. BЄS
In coarse-granularity quality scalability, pictures in Subject to
enhancement layers may be used as prediction
references and therefore all the enhancement layer
pictures in a group of pictures typically have to be
disposed as a unit. In fine granularity scalability, The solution (BЄS*) to the problem satisfies that R (B*) ≤
the use of enhancement layer pictures as prediction Rc and D (B*) ≤D (B) for all B
sources is limited and therefore finer steps of In S*, where
bitrate can be achieved compared to coarse-
granularity scalability.
c) Finally, spatial scalability is used for creation of
multi-resolution bitstreams to meet different That is, if the solution to the problem is B*, then there is
display requirements or constraints and is very no other B in S which satisfies the constraint Rc, that will
similar to SNR scalability [5]. result in a smaller value for the objective function than D
A spatial enhancement layer enables recovery of coding loss (B*). The Lagrange multiplier theory offers a way of
between an up-sampled version of the reconstructed layer solving the above constrained problem (i.e. finding B*) by
used as a reference by the enhancement layer and a higher representing the problem as an unconstrained problem [3].
resolution version of the original picture.
3.3 Major Theorem
3. Rate Distortion Optimization
The constrained optimisation problem was presented earlier
3.1 Lagrangian multiplier method in previous section, equation (2). The Lagrange theory
represents the constrained problem as an unconstrained
In H.264/AVC, it is the art of the encoder to have the ability problem as follows:
of having the effective way of encoding a given video Theorem: for any λ≥0, the solution B*(λ) to the
sequence by selecting among numerous ranges of modes and unconstrained problem
parameters.
The encoder targets to achieve optimum rate distortion
performance by choosing the best of modes and parameters
of a given video. Doing this, the encoder would be looking
to minimize distortion in a sequence of particular video. This is considered as solution of the constrained problem in
(1) presenting Rc = R (B* (λ)) as the constraint.
Proof of the theorem
If B* (λ) is the solution to the constrained problem (4) then:
Therefore,
Where the solution to (4-11), M*, would satisfy,
If this is true for all B in S, it is true for a subset of B in S

where,
The optimum coding mode M* (if one exists) can be found
by solving (14). That means, when the macroblock is coded
Now, for the above subset and for any λ≥0: in mode M* it would satisfy the target rate (R (M*) =
Rc). All the other modes (if they exist) that satisfy R (M)≤ Rc
will have a higher distortion than D(M*).
The term D (M) + λ .R (M) in equation (4-9) is called the
Lagrangian rate-distortion cost. The mode that minimises
Therefore with the constraint Rc = R (B* (λ)), the solution the Lagrangian rate-distortion cost for a particular λ≥0
B* for the unconstrained problem is also the solution for the (which satisfies the rate constraint in the constraint
constrained problem. problem) is selected as the solution mode for the constrained
It should be noted that the theory does not guarantee a problem.
solution for the constrained problem. It only states that for
any λ≥0 of the unconstrained problem, there is a 4. Methodology
corresponding constrained problem which has the same
solution as the unconstrained problem. 4.1 Objective video quality measurement
3.4 Optimisation problem Objective video quality measurements are used to measure
the video quality, typically in situations where fast
Consider a macroblock, for which the encoder can encode (sometimes online) and repeatable measurements of the
the macroblock using only one of the ‘K’ possible modes distortion or the difference between the video under test and
given by the set m = {m1,, m2, … , mK }. Let ‘M’ (M Єm) be a reference video are needed [7].
the mode selected to code the macroblock. In the context of
H.264/AVC, these mode allocations could be any allowable 4.2 PSNR
combination of macroblock partition modes, Quantisation
Parameters (QP), choice of reference frames etc… so that The Peak Signal to Noise Ratio (PSNR) is the most
the K possible modes will include all the possible admissible commonly used objective measure of video quality. PSNR is
parameter combinations for the measured as follows:
Macroblock
Define the objective function D(M) and constraint
function R(M) , where D(M) and R(M) are distortion and
rate of the macroblock as a result of selecting a particular
coding mode. If the rate constraint is Rc, the constraint Where n is the bit depth and MSE is the Mean Squared
problem is defined as: Error between corresponding pixel values of the original
Find the coding mode M*, image and the current image of the sequence under test. For
M × N array of pixels, MSE is given by:
Subject to
Where Po (i, j) denotes a pixel from the original image

and Pi (i, j) denotes the corresponding pixel from the test
image. The parameters ‘i’ and ‘j’ point to a position in the
pixel arrays.
The MSE in itself can be a measure of distortion.
This may be written as an unconstrained problem using a However PSNR is preferred because the log scale provides a
Lagrange multiplier: more realistic mapping to quality variations. Therefore,
PSNR continues to be the most commonly used objective
quality measure [5].
5. Implementation
For the objectives to be achieved, software video simulation Based on Lagrangian Cost Function, if a video frame is to
tool JSVM was used to implement and test the algorithms. be sent on the outgoing link, it is first placed in the output
There are plenty and different H.264/AVC reference buffer. Note that, for simplicity, we don’t consider the buffer
software. JSVM was chosen for this research due to its limitations for the simulations in here.
flexibility of varying parameters. If the outgoing link cannot accommodate all the video
JSVM codec is commonly used to test new algorithms in packets, it will first drop the additional enhancement PR
the video community. The use of this reference software slices one by one. If the link is still overloaded, the spatial
enables realistic comparison of the performance of different enhancement layers are dropped next in the same spirit, i.e.,
algorithms developed by different researchers. The source scale out the enhancement layers completely sticking only to
code is mainly the same as the one used in the C the base layer. The optimized SVC offers better quality than
programming language [8]. the unoptimized SVC one
6. Results analysis
In this part of the simulation, basic parameters such as

frame rate of 30 Hz, number of frames 300 and group of
pictures 16 were taken into consideration. Set of stiff video
were used to evaluate the performance such as foreman,
garden, football, flower, Claire and Carphone. The PSNR
versus bitrates graph for various group of pictures were
studied in difference circumstances. Below are different
cases that were taken into consideration:
Initially, spatial dimensioning is represented by

QCIF and CIF, but was taken without additional progressive Figure 4. Rate distortion optimization for scalable coding
refinement (PR) slices. With additional PR, the transform
coefficients are refined thus the improvement of the Comparing to single-layer at higher bitrates, also, when the
reconstructive pictures’ quality. The performance clearly outgoing capacity Rout is larger than the required incoming
proves that the PSNR varies with the quality. rate, at 1670 kB/s, the RD-optimized single-layer coding
and unoptimized coding perform the same.
This is obvious, as at higher bitrates, the network link
will rarely overflow and very few or no video packets are
lost. However, if the outgoing rate is very small, it can be
seen that SVC strategy leads to good improvements in terms
of reconstructed video quality. Table 1 shows the
improvements obtained for individual video streams for the
outgoing link Rout = 600 Kbit/s.
Figure 2. Sequential scalable coding (Foreman)
In this case, several spatial resolution or bitrates are taken

into consideration or provided by the encoded bitstream.
The result shows that the PSNR is directly proportional to
bitrates.
Figure 5. evaluation of SVC and SLC
Table 1: comparison of different video streams
Sequences Scalable Video Coding Single layer coding
Optimized Unoptimized Optimized Unoptimized

(dB) (dB) (dB) (dB)
Garden 45.0008 38.5645 42.5682 40.2658
Foreman 34.5545 35.2564 32.5654 30.1254

Figure 3. Single Layer coding
Football 37.2356 37.0052 36.2545 36.5485 "Complexity Reduction of H.264 using Lagrange
Optimization Methods," IEE VIE 2005, Glasgow, 4~6
Flower 40.3215 39.0235 37.5468 37.6256
April, 2005.
Claire 36.2597 36.4566 31.2564 32.2564 [2] H. Kim and Y. Altunbasak, "Low-complexity
Carphone 41.3255 38.4552 38.2545 39.2545 macroblock mode selection for H.264/AVC encoders,"
presented at International Conference on Image
Processing, Singapore, 2004.
[3] K. P. Lim, "JVT -I020, Fast INTER Mode Selection."
7. Recommendations San Diego: ISO/IEC MPEG and ITU-T VCEG Joint
Video Team, 2003.
Although the video coding standards exhibit acceptable [4] X. Li. Scalable video compression via over complete
quality-compression performance in many visual motion compensated wavelet coding. Signal Processing:
communication applications, further improvements are Image Communication, special issue on
desired and more features need to be added, especially for subband/wavelet interframe video coding, 19:637—651,
some specific applications. The important considerations August 2004.
for video coding schemes to be used within future networks [5] S.-R. Kang, Y. Zhang, M. Dai, and D. Loguinov, \Multi-
could be bases on Compression efficiency, robustness with layer active queue management and congestion control
respect to packet loss, adaptability to different available for scalable video streaming," in Proc. IEEE ICDCS,
bandwidths and adaptability to memory and computational Tokyo, Japan, Mar. 2004, pp. 768{777}.
power for different clients. [7] T. Oelbaum, V. Baroncini, T. K. Tan, and C. Fenimore,
Several other communication and networking issues are also “Subjective quality assessment of the emerging
relevant, such as scalability, robustness, and interactivity. AVC/H.264 video coding standard,” International
A network with a single active node was considered, in our Broadcasting Conference (IBC), Sept., 2004.
simulations. This could be further enhanced to more [7] R. Leung and D. Taubman. Impact of motion on the
practical situations with a hierarchy of many active network random access efficiency of scalable compressed video.
nodes and perform rate shaping at every node accordingly. Proc. IEEE Int. Conf. Image Processing, 3:169—172,
Different values for the Lagrangian multiplier λ could be September 2005.
modeled for more stringent buffer conditions. A reasonable [8] R. Leung and D. Taubman. Perceptual mappings for
value for λ can be determined in maximizing the visual quality enhancement in scalable video
Lagrangian cost function, since λ is determined as a compression. Proc. IEEE Int. Conf. Image Processing,
function of buffer fullness. 2:65—68, September 2005.
The scalable video coding approach could be further [9] R. Leung and D. Taubman. Transform and embedded
extended to MCTF based scalable video codec which coding techniques for maximum efficiency and random
employs an open-loop architecture. accessibility in 3-D scalable compression. IEEE Trans.
Image Processing, 14(10):1632—1646, October 2005.
8. Conclusion [10] R. Leung and D. Taubman. Minimizing the perceptual
The choice of a Scalable Video Coding framework in this impact of visual distortion in scalable wavelet
context brings technical and economical advantages. Under compressed video. Proc. IEEE Int. Conf. Image
this framework, network elements can adapt the video Processing, October 2006.
streams to the channel conditions and transport the adapted [11] R. Leung and D. Taubman. Perceptual optimization for
video streams to receivers with acceptable perceptual scalable video compression based on visual masking
quality. The advantages of deploying such an adaptive principles. IEEE Trans. Circuits Syst. Video Technol.,
framework are that it can achieve suitable QoS for video submitted in 2006.
over wired and wireless networks, bandwidth efficiency and [12] T. Wedi and Y. Kashiwagi, “Subjective quality
fairness in sharing resources [11]. evaluation of H.264/AVC FRExt for HD movie
The adaptive scalable video coding technology produces content,” Joint Video Team document JVT-L033, July,
bitstreams decodable at different bitrates, requiring 2004.
different computational power and channel bitrate. In [13] ISO/IEC JTC 1/SC 29/WG 11 (MPEG), “Report of the
addition, the bitstream is organized with a hierarchical formal verification tests on AVC/H.264,” MPEG
syntax that enables users to easily extract only a subpart of document N6231, Dec., 2003 (publicly available at
the data contained in the bitstream and still being able to http://www.chiariglione.org/mpeg/quality_tests.htm).
decode the original input video but at a reduced spatial [14] T. Schierl, T. Stockhammer and T. Wiegand, "Mobile
resolution or frame rate. This process can be applied Video Transmission using Scalable Video Coding
recursively, that is, once a new bit stream is extracted out of (SVC)," IEEE Trans. On Circuits and Systems for
the original, it can undergo successive extractions Video Technology, Special issue on Scalable Video
corresponding to always lower resolutions. Coding, scheduled June 2007.
[15] S. Wenger, Y.-K. Wang and T. Schierl, “Transport and
References Signaling of SVC in IP networks,” IEEE Transactions
on Circuits and Systems for Video Technology, Special
[1] C. S. Kannangara, I. E. G. Richardson, M Bystrom, J. issue on Scalable Video Coding, scheduled for: March
Solera, Y. Zhao, A. MacLennan & R. Cooney, 2007.
Markov Based Mathematical Model of Blood Flow

Pattern in Fetal Circulatory System
Sarwan Kumar 1 , Sneh Anand 2 , Amit Sengupta 3
1
Dr B R Ambedkar National Institute of Techmology,
Jalandhar -144011 ,Punjab,India
pahujas@gmail.com
2
Indian Institute of Techmology
CBME Delhi, India
3
Indian Institute of Techmology
CBME Delhi, India
During the time of pregnancy it can be found that the

Abstract: This paper presents a novel approach to estimate
blood flow characteristics in the fetal circulatory system during
umbilical cord is in the form of a knot or at times this cord
pregnancy. We have developed a mathematical model of the is wrapped around the body of the baby. This is a common
fetal circulatory system by taking two nodes concept based on phenomenon and there is no prevention for this. This does
Markov model. As the oxygenated blood flows from mother side not pose any risk or threat to the baby or the mother. There
through placenta to fetus and deoxygenated blood from fetus to may be some complications of the placenta due to pregnancy
mother via umbilical cord. When it is simulated, the model the most common of which is placenta previa [10]. In this
shows how the oxygenated blood flows from placenta (one node) the placenta is attached over or near the cervix. With the
to the umbilicus (second node) and deoxygenated blood to growth of the fetus there is pressure on the placenta and due
placenta from the fetus. Also the same model is simulated at to this reason there may be bleeding [11]. If this condition
different conductivity path of the umbilical cord and available occurs there is need for medical care so that one can be
blood supply at placenta. Also it shows the effect of uterine
ensured of a safe labor for the baby. Due to many
contractions on the blood supply to the fetus. All simulations
complications there may be decrease in blood supply to fetus
have been performed in the Lab VIEW environment at various
conditions of vein and arteries. which leads to asphyxia and increase in heart rate [12].
Hence compromise of the fetal blood flow through the
umbilical cord vessels can have serious deleterious effects on
Keywords: Markov model, placenta, umbilicus cord, the health of the fetus and newborn. There for, it is
mathematical model, uterine contraction. necessary to know the blood profile in the fetus. This paper
discuss about a novel mathematical model for the
1. Introduction circulation of blood in the fetal circulatory system by taking
The baby develops in the uterus with a life support system two node concept based on Markov model [4] to know how
for the fetus and is composed of umbilical cord, placenta the blood profile. The same model is simulated at various
and amniotic fluid. The placenta is a pancake shaped conductivities of the blood vessels and available blood at
temporary organ which is attached to the uterus and is placenta. We have also demonstrated the effect of uterine
connected to fetus through the umbilical cord. The umbilical contractions on the blood profile which would be useful to
cord is the lifeline between the fetus and placenta. As soon it assist in developing a new bioelectric sensor for the
is formed it functions throughout pregnancy to protect the evaluation actual blood flow time.
vessels that travel between the fetus and the placenta. The
responsibility of the placenta is to act as a point of trade 2. Markov Model
between the circulatory system of the mother and the baby. A Markov model is a stochastic process whose dynamic
It is very important to know the relationship between behavior is such that its future development depends only on
concentration (quantity) of blood available at placenta and the present state space. In other words, the description of the
how quickly it passes to fetus through vein, the only path to present state fully captures all the information that could
carry good blood to fetus from mother and the waste influence the future evolution of the process. Being a
products of the fetus are transferred to the mother’s blood stochastic process means that all state transitions are
through umbilical arteries. Therefore the umbilical cord is probabilistic. At each step the system may change its state
called the life line and it is through this cord that the from the current state to another state (or remain in the
placenta and the fetus are attached to each other. There are same state) according to a probability distribution. The
three blood vessels in the umbilical cord, two small arteries changes of state are called transitions, and the probabilities
and a vein [10]. This cord can grow to a length of 50-60 cm associated with various state-changes are called transition
which allows the baby to have enough space to move safely probabilities. In other words, the description of the present
without damaging the placenta or the umbilical cord. The state fully captures all the information that could influence
placental conductivity increases with the age of pregnancy the future evolution of the process. In order to formulate a
[8]. Complete circulatory system is shown in figure 1. Markov model we must first define all the mutually
exclusive states of the system. The state of the system at t =0

are called the initial states (P0), and those representing a (5)
final or equilibrium state are final stage (P1). The set of
Markov state equations describes the probabilistic transition Similarly
from the initial to final states.
The transition probabilities must obey the following two (6)
rules:
1 The probability of transition in time ∆t from one The final solution of equation’s 5 and 6
state to another is given by gain into ∆t.
2 The probabilities of more than one transition in P0(s) = p0(s+u1 + u2) - p1(u1 + u2)/ (s+ gv)(s+ u1 +
time are infinitesimals of higher order and can be u2)- g v (u1 + u2)
neglected. (7)
3. Proposed Model
Figure 1 presents the complete fetal circulatory system and
its equivalent Markov chain in figure 2. The mostly (8)
problems related to node Ia are Intrauterine Growth
Restriction (IUGR) and preeclampsia [5 ]. These are due to And
high blood pressure, diabetes, infection, kidney disease, P1(s) = p1(s+ gv) + gv p1/(s+ gv)(s+ u1 + u2)- gv (u1 +
heart or respiratory disease, alcohol, drugs and cigarette u2) (9)
smoking ( figure 3) which may lead to fetal hypoxia, fetal
death, low birth weight, placenta abruption (figure 4)[5]. P1(t) = (p1- p0)gv / (u1 + u2 )+ p1 (u1 + u2 - gv )/
The problems related to Umbilical Cord i.e. node II are two (u1 + u2 )e- (u1 + u2) t (10)
vessels, long cord, nuchal cord and short cord figure 5. The
node Ib and node III are less significant in fetal circulation
and are ignored. The modified nodes representation of the
fetal circulation system and is equivalent signal flow graph
is shown in figure 6. In term of mathematical model as
described by the Markov Model [7], node I represents full of
oxygen rich blood toward mother side and node II represents
the fetus side. Umbilicus cord connects the two nodes.
There are two stages:
Stage 1: Placenta attached to the mother side, say node I,

full of oxygen
Stage II: Umbilicus, the entering point to fetus, say node II
Let us P0(t) the quantity of the good blood at node 1

P1(t) the quantity of blood reaches at node II through vein
gv the conductivity gain of the vein
u1 the conductivity gain of the artery1 Figure 1. Fetal Circulatory System
u2 the conductivity gain of the artery2
After ∆t the blood at node I and node II is given by
(1)
P1(t+∆t) = P1(t) (1- u1 ∆t) + P1(t) (1- u2 ∆t) + P0(t) gv ∆t

(2)
From equation 1 and 2
Figure 2. Fetal Circulatory System in nodes representation

P0(t+∆t) - P0(t) = - P0(t) gv ∆t + P1(t)( u1 + u2) ∆t (Node Ia Uterine Artery, Node Ib Placenta, Node II
Umbilical Cord, Node III Fetal Heart)
(3)
P0(t+∆t) - P0(t)/ ∆t = - P0(t) gv + P1(t)( u1 + u2) (4)
[4]. Here tissue is represented by capacitor in parallel with

resistance as shown in figure 8 [8 ] This time increases as
the conductivity (gv ) decreases because of knot or some
other reasons as shown in figure 9 where the time is
approximately 10 seconds. This may leads to the child a
number of dangerous effects (depression of the central
nervous system, breathing paralysis, etc.).
Figure 3. Problems related to node Ia and their causes
Figure 4. Problems to fetus due to Intrauterine Growth

Restriction (IUGR) and Preeclampsia Figure 6. Actual system and Equivalent Markov Signal
Flow Graph
This will increase further if the contractions are more [1].

The quantity of the blood supply from the mother is highly
affected the uterine contractions. Because the contractions
will increase the intramyometrial pressure (120mmHg)
compared to arterial pressure (85mmHg) [1]. The initial
blood supply is less in this case and less amount of blood
crossing the umbilical cord feeding fetus. This is the
situation of less oxygen to the fetus, may lead to hypoxia.
This is shown by our model as the output magnitude of P0 in
term of available blood quantity. The output is shown in
figure 10 for the 50% available blood. Maternal blood enters
the placenta through the spiral arteries (the terminal
branches of the uterine artery), which traverse the
myometrium-muscular contractile layer of the uterus and
flow into the intervelleous space. At this level the mother
exchanges substances with the fetus through the "placental
barrier." Anesthetics and analgesics are of low molecular
weight, and are easily exchanged by diffusion as a result of
Figure 5. Problems related to node II (Umbilical Cord) the concentration gradient between the maternal and fetal
compartments. Caldeyro-Barcia et.al have found that when
4. Simulation and analysis the uterus is at rest, without contraction, the mother's
arterial blood easily crosses the intervelleous space since the
The software is designed in LabVIEW and simulated at average arterial pressure is about 85 mmHg and the
different levels of P0, P1 ,gv ,u1 and u2 . When equation 8 is intramyometrial pressure external to the arteries is about 10
simulated for various values of conductivities of vein ( gv) mmHg. During uterine contractions, however, the
and two arteries (u1 , u2 ), we got the exponential curve. intramyometrial pressure rises to 120 mmHg, exceeding the
The response of the P0(t) is shown in figure 7 at gv unity or arterial pressure which, under such conditions, is about 90
100% conductivity. This indicates that the blood transfer mmHg. The arteries therefore become temporarily occluded
from mother side placenta to fetus through vein having because of the external pressure, and the placenta becomes
100% conductivity at the start of the process. This indicates disconnected from the maternal circulation. [1]
that the blood takes 4 second to reach to fetus. This If the conductivity is reduces to 10%, it take approximately
exponential curve due to simulation is same as the current 50 second to reach its final stage ie fetal heart. This very
discharge through capacitor and register which has already dangerous stage for the fetus as lesser oxygen is going to
been established in tissue impedance characterization [3],
fetus extremely asphyxia. The fetus may die because of less

blood or oxygen.
Figure 11. Deoxygenated blood flow curve, when both

arteries are good
Figure 7. Output at gv unity with blood flow time of 5
seconds
Figure 8. Tissue’s Cell Membrane & its Electrical

Equivalent Figure 12. Deoxygenated blood flow curve, when one artery
good
Figure 9. Output at gv 0.5 with blood flow time of 10

seconds
Figure 13. Blood flow time increases with decrease in
length of the umbilical vein
When equation 10 is simulated at various levels of arteries

path (u1 + u2 ). When both are working the response of this
is shown in figure 11. It takes approximately 2 second to
transfer the waste to placenta. This time increases as the
conductivity decreases. The setting time is doubled when
any one arteries is failed as shown in figure 12. This may
increase the acidic or PH composition level in the fetus
which can spoil umbilicus cord. The effect of length and
diameter of vein have also been simulated. The reaching
Figure 10. 50 % available blood during contractions and time also increases with the decrease in the length of the
same amount to fetus cord even the supply of the blood is 100% at placenta. The
result of which is shown in figure 13. This indicates the
blood is reached to fetus with longer time. When compared
the result with the blood flow from model with the actual
flow of Doppler FVW taken from the paper [5], it shows the impedance measurement derived using two tetrapolar
same blood flow pattern as actual. The figures 14(a) show probes of different sizes”, Biomedical Engineering ,pp
the simulated blood flow while 14(b) the actual flow. 1-7,2006.
[4] R.J Halter,., A. Hartov,., J.A. Heaney, K.D. Paulsen,.
A.R. Schned,., “Electrical Impedance Spectroscopy of
the Human Prostate”, IEEE Transactions on Biomedical
Engineering, pp 1321-1327 , 2007.
[5] A. Gaysen , S. K. Dua, A. Sengupta and Nagchoudhuri
, “ Effect of Non-Linearity Doppler Waveforms
Through Novel Model”, Biomedical Engineering
Online, pp1-13,2003.
[6] A S Gordon ,, J Strauss and G A Misrahy, “ Electrical
Impedance of Isolated Amnion”, Biophysical Journal,
,pp 855-865,2000.
(a)
[7] G. D. Clifford, F. Azuaje, P.E. McSharry , “Advanced
Methods and Tool for ECG Data Analysis”, Artech
House, pp 295-300 , 2006 .
[8] Guyton, Textbook of Medical Physiology, Eight
Editions, 1991.
[9] Ross and Wilson, Anatomy and Physiology in Health
and Illness, Tenth Edition , 2006.
[10] T. Erkinaro , “Fetal and Placental Haemodynamic
Responses to Hypoxaemia , Maternal and Vasopressor
Therapy in a Chronic Sheep Mode” l, Acta University ,
(b) pp-1-96, 2006.
Figure 14. Comparison of the flow of blood between result [11] J. C. Huhta , “ Fetal congestive heart failure” Seminars
from model and with actual flow, (a) Flow of blood response in Fetal & Neonatal Medicine 10, pp 542-552 , 2005.
of Markov model (b) Actual blood flow: A frame extracted [12] F. Kovacs, M. Torok, and I.Habermajer , “A Rule-
from Doppler FVW [5] Based Phonocardiographic Method for Long-Term
Fetal Heart Rate Monitoring” , IEEE Transactions on
5. Conclusion Biomedical Engineering , pp 124-130 , 2000.
The blood flow timing between the placenta-fetus and fetus - Authors Profile
placenta is given by the equation 8 and 10 respectively and
simulated using LabVIEW software. The flow is exponential Sarwan Kumar received the BTech and MTech degrees in
which shows that the umbilical cord structure (vein and Electrical Engineering from Regional Engineering College
arteries) acts as a capacitor in parallel to resistance. Time Kurukshetra in 1992 and 1997, respectively. He is associate
professor at National Institute of Technology Jalandhar. Now he is
taken by the blood to reach fetus is increases as conductivity
pursuing PhD from IIT Delhi, India under the guidance of
decreases. Also the time increases in case of lass quantity of
professors Sneh Anand IIT Delhi and Dr. Amit Sengupta, ,
blood is available due to uterine contractions, knot or any Consulting Obstetrician & Gynecologist (CHS), Mumbai.
other reason. The simulated results show the larger settling
time in case of short length. On the blood flow; it would be
useful to assist in developing a sensor for the evaluation of
conductivity of the umbilical cord and placenta during
pregnancy for the well-being of fetus. We are developing a
stand alone instrument for monitoring the various
parameters of the fetal model.
References
[1] C. Hernandez Sande, G. Rodriguez-Izquierdo, and M.

Iglesias,” Intermittent Drug Administration During
Labor and Protection of the Fetus, IEEE Transactions
on Biomedical Engineering, pp 615-619, 1983.
[2] S M Sims, E Daniel and R E Garfield, “Improved
Electrical Coupling in Uterine Smooth Muscle Is
Associated with Increased Numbers of Gap Junction”,
Journal of General of Physiology, pp-353-375, 1982.
[3] S Gandhi, D C Walker ,B.B. Brown and D. Anumba, “
Comparison of human uterine cervical electrical
Redevelopment of Graphical User Interface for

IPMS Web Portal
Hao Shi
Victoria University, School of Engineering and Science

Melbourne, Australia
hao.shi@vu.edu.au
course designer to create groups and manage group

Abstract: IPMS, short for Industry Project Management
System, is a web portal for industry project team management. activities besides course contents. Unfortunately there is no
IPMS is a very useful project managment tool to manage “one size fits all” solution in project management [8]. Many
students, allocate projects, coordinate supervisors and liaise with higher education institutions continue to build their own
industry sponsors. It has speeded up the process and allowed the project management tools as they provide significant
stakeholders to focus on their key tasks. However the originally benefits to teaching and learning. More and more project
developed IPMS is no longer working after migrating to a new teams in industry, academia, and the open source
server. As a result, the manual management was brought back
which was both times consuming and tedious for both project
community are increasingly reliant on web-based project
students and the course coordinator. This project aims to management portals [9].
upgrade IPMS to PHP5.0 and re-develop the new GUI The market for tools to improve software project
(Graphical User Interface) with enhanced system management and software quality management is fast
functionalities. In this paper, first the background information growing. It has been approved that vendors of software
about IPMS is described. Then the newly developed GUI is project and quality management tools can walk the talk by
presented and the usability test is conducted on the re-developed
using quantitative data to manage the development project
GUI. It is concluded that the newly developed GUI meets the
user requirements and is better than the existing GUI. and process [10]. Recently, agile software development
methods are popular because software should be developed
Keywords: IPMS, Project Management Systems, GUI in a short period. However, conventional project
(Graphical User Interface), Industry Project, Web Portal.
management techniques are often not adaptable to such new
development methodologies. A new tool based on the
1. Introduction communication model has been developed for agile software
Many final-year projects are offered at tertiary computing development [11], which allows to monitor product quality
degree programs to provide project students team work and control and progress control. Some of these tools focus
real world project experience under supervision of external mainly on project management for teaching and learning [8]
project sponsor and academic staff [1]. However, managing while others have full support for administrative tasks such
software project teams is a complex task [2]. It should have as student registration, team formation, project
occupied 20% time of the project coordinator but in the end confirmation, supervisor allocation and document
it took more than 80% [3]. In order to reduce administrative management [4, 12, 13].
load, many project management tools have been produced In this paper, it aims to upgrade the existing IPMS and
[4]. Some tools monitor full cycles of software engineering re-develop a new GUI for IPMS.
projects while others emphases more on aspects of the
management projects [2]. 2. Background
SourceForge is the best known web portal, currently hosts
IPMS (Industry Project Management System) is a “All-In-
over one hundred thousand projects and over a million users
One” web portal. It was primarily developed to automate
[5]. Open source tools such asDokuWiki, Trac, and
and streamline management of the final-year Industry
Subversion can be integrated to provide a low-cost platform
Projects at Faculty of Health, Engineering and Science at
for student collaboration on team projects [6]. By
Victoria University. IPMS prototype was initially developed
consolidating project artifacts in a central location, the wiki
based on the FDD methodology using Linux-Apache-
software serves as both a repository for project information
MySQL-PHP, (LAMP) three-tier client-server architecture
and a means of communication between team members and
[Martin] as shown in Figure 1. PHP (Hypertext
course instructors who may be working from different
Preprocessor) is the program language which generates
physical locations. Integrated version control helps students
dynamic web GUI (graphic user interface). Apache is
track changes in their documents and provides a safety net
employed as a web server running under Linux operating
for recovery of information that has been previously deleted
system while MySQL is database management system for
from project artifacts [6].
IPMS web pages and supports CMS (Contents Management
Even Moodle or Blackboard, dedicated eLearning tools
System). LDAP (Lightweight Directory Access Protocol) is
[7] have incorporated group management tools which allow
used for the user authentication [4, 11].
the
Figure 1. System topology [Martin]
Examples of the IPMS major menus are shown in Figure 2

and 3.
Figure 2. Student Menu [4]
Figure 3. Admin Menu [4]

3.1 Top-down Module Design
3. Redevelopment of IPMS GUI After the structured system analysis, the top-down design for
In the 1st Semester 2009, one of industry project teams were IPMS module using Gane and Sarson graphical
assigned the task to upgrade the exiting IPMS and re- specification technique is shown in Figure 4.
develop its GUI because it was no longer working after PHP 3.2 Admin Menu
was upgraded from version 4.0 to 5.0 on a new server. The
Once a user is logged in as an admin user, the admin menu
project team consists of four final-year computer science
becomes available at the left side. The major change is that
students. They aim to maintain existing system functionality the menus are grouped into major functionality shown in
in the new system and improve the user interface and the shown Figure 5.
logical flow of pages and add possible new functionality
[14]. In the following subsections, the newly developed
Admin menu and Student Menu are presented.
IPMS Module
Login Email Database Register Web_Content User User_Details Project Team Report Former_Student
Login Logout PHPMyAdmi Update_Web_Conten Team_Report

n t
Add_New Delete Details
Check_Login External Update Team _Document User_Report Project_Report
Figure 4. Top-down Module Design for IPMS [15]
Back to the default home page
Contain general web pages

that are meant for everyone
to see
Contain web pages which

displays Team’s personal
information
Contain the assigned

Team(s) to a supervisor with
their contact details
Figure 6. General Information submenus

Contain forms/documents
used to enable a supervisor
to evaluate Team’s
performance
used to enable unit
coordinator to able to
modify user, project and
Logout from the system team information.
Figure 5. Newly developed GUI for Admin Menu [15]

3.2.1 General Information Menu
The general information menu consists of six submenus,
namely About Industrial Project, Projects, Supervisors,
Sponsors, Industrial Partners and FAQs as shown in Figure
6.
Figure 7. Administrator Menu
3.2.2 Administrator menu
3.2.3 My Team menu
The Administrator Area contains the key menus such as
My Team menu contains three submenus namely:
Overview Information, Reports, Emails, User
RCM3001, RCM3002 and Assessment, each submenu
Administration, Project Administration, Team
contains several submenus as well as shown in Figure 7.
Administration, Database Access and Content Management
as shown in 7.
Q1. Registration/Signup process

Q2. Availability of other students to form a team
Q3. Team formation
Q4. Registration of an available project to a team or
project proposal
Q5. Efficiency from registration to team formation
Q6. Overall experience
The detailed results of the usability test are shown in Figure

11.
Back to the default home page
Contain general web pages

that are meant for everyone
to see
Contain web pages which

displays Team’s personal
information
(a) RCM3001 submenu

Contain the team with their
contact details
used to enable students to
view Team’s performance
criteria
Logout from the system
Figure 9. Student Menu [15]
(b) Assessment submenu Figure 10. My Team submenus

Figure 8. My Team menu
Q1. Re gistration/Signup proce ss

3.3 Student Menu 50%
40%
Once a user logins in as a student, the student menu displays
No. of Users
30% Pap er- Based

as shown in Figure 8. NEW GUI
20%
10%
0%
3.2.1 My Team menu Very Hard Hard Neutral Easy Very Easy
This menu contains five submenus, i.e. Team Details,

(a)
Create, Join, Upload Photo and Leave as shown in Figure
Q2. Availability of other s tudents to form a te am
10. 50%
40%
No. of Users
4. Use Acceptance Test 30% Paper-Based

New GUI
20%
Many usability tests were carried out by the project 10%
coordinator during the course of IPMS re-development. 0%
Improvement and changes were made to enhance the GUI. Very Hard Hard Neutral Easy Very Easy
In this paper, the usability test conducted is in the form of (b)

student user experience survey on the new developed GUI
on the same aspects [12]:
Q3. Te am form ation Design through Risk Analysis”, Proceedings of 43rd ACM
60% Southeast Conference, pp. 2-300- 2-305, 2005.
No. of Users
50%
[3] G. Jones, “One Solution for Project Management”,
40% Paper-Based
30% NEW GUI Proceedings of SIGUCCS (The Special Interest Group on

20% University and College Computing Servies) Fall Conference.
10% pp. 65-69, 2001.
0% [4] H. Shi, “IPMS: A Web Portal for Industry Project Team
Very Hard Hard Neutral Easy Very Easy
Management”, International Journal of Communication, Vol.
(c) 7 No. 4, April 2007, pp. 111-116.
Q4. Re gis tration of an available project to a te am or project pr oposal
[5] Source-Forge, http://sourceforge.net [Accessed: Feb. 12,
50% 2010]
40% [6] E. R. Haley, G. B. Collins, and D. J. Co, "The wonderful
No. of Users
30% P aper-B as ed world of wiki benefits students and instructors", IEEE

IP M S P o rtal
20% Potentials, Volume: 27, Issue: 2, pp. 21-26, 2008.
10% [7] Blackboard, http://blackboard.com [Accessed: Feb. 12, 2010]
0% [8] Moodel, Open-source course management system
http://moodle.com [Accessed: Feb. 12, 2010]
(d) [9] A. N. Norita and P. A. Laplante, “ Software Project
Management Tools: Making a Practical Decision Using
Q5. Efficiency from r egis tration to team form ation
60%
AHP”, Proceedings of the 30th Annual IEEE/NASA
50% Software Engineering Workshop, 24-28, 2006.
No. of Users
40% Paper-Based [10] G. V. Seshagiri and S. Priya, "Walking the Talk: Building
30% NEW GUI
Quality into the Software Quality Management Tool",
20%
10%
Proceedings of the Third International Conference On
0% Quality Software (QSIC), pp. 67 – 74, 2003.
Very Hard Hard Neutral Easy Very Easy [11] N. Hanakawa and K. Okura, "A project management support
tool using communication for agile software development",
(e)
Proceedings of the 11th Asia-Pacific Software Engineering
Q6. Over all e xpe rie nce
Conference (APSEC), pp. 316 - 323, 2004.
70%
60% [12] R. Martin and H. Shi “Design and Implementation of IPMS
Web Portal”, Proceedings of International Conference on
No. of Users
50%
Paper-Based
40%
30%
NEW GUI Computers and Advanced Technology in Education (CATE),
20% pp. 16-21, 2007.
10% [13] H. Shi, "Reshaping ICT Industry Projects - My Three-Year
0%
Experience", Proceedings of AusWIT06 Australian Women
in IT Conference, 4-5 December, Adelaide, Australia, pp.36-
(f) 46, 2006
Figure 11. User experience survey [14] R. El Tabbal, L. H. It, J. Toke and D. Tu, “Redevelopment of
Industry Project Management System”, Final-year Industry
5. Conclusions Project Design Report, School of Engineering and Science,
Victoria University, November 2009.-
IPMS has been upgraded to PHP5.0 after one-year [15] R. El Tabbal, L. H. It, J. Toke and D. Tu, “Redevelopment of
development. The usability test has proved the new Industry Project Management System”, Software Design
developed GUI efficient and user friendly. The new Document and User Manual, School of Engineering and
upgraded IPMS removes tedious manual process and Science, Victoria University, June 2009.
provides smooth management functionalities for students,
supervisors, and coordinator and industry sponsors. It is Author Profile
concluded that the newly developed IPMS meets the user
requirements and is better than the previous version.
Hao Shi obtained her BE in Electronics
Engineering from Shanghai Jiao Tong
Acknowledgements University, China and her PhD at
University of Wollongong. She is now an
The author would like to thank the project team, Riad El
Associate Professor and ICT Industry
Tabbal (team leader), Leang Heng It, Jack Toke and Duncan
Project coordinator at School of
Tu and the project supervisor, Associate Professor Xun Yi
Engineering and Science, Victoria
for their contributions in revitalising the IPMS GUI.
University. She has established Industry- Based Learning
program at the School and won a number of Teaching and
References Leaning grants and awards. She is currently managing more
than a dozen of ICT university scholarships with local
[1] J. Ceddia and J. Sheard, “Evaluation of WIER – A Capstone industry partners via her grants from Victorian Government,
Project Management Tool”, Proceedings of the International
Australia.
Conference on Computers in Education (ICCE), pp. 777-781,
2002.
[2] J. L. Smith, S. A. Bohner, D. S. McCrickard, “Project
Management for the 21st Century: Supporting Collaborative
Analysis of Searching Techniques and Design of

Improved Search Algorithm for Unstructured
Peer – to – Peer Networks
Dr. Yash Pal Singh1 , Rakesh Rathi2, Jyoti Gajrani3 , Vinesh Jain4
1
Bundelkhand Institute of Engg. and Tech.
Jhansi India
yash_biet@yahoo.co.in
2
Govt.Engg.College,Ajmer
Badliya Circle, NH 08, Ajmer
rakeshrathi4@rediffmail.com
3
t_jyoti1@rediffmail.com
4
vineshjain1280@rediffmail.com
topology. Unstructured, Loosely structured and Highly

Abstract: We study the performance of several search structured are various categories of P2P networks based on
algorithms on unstructured peer-to-peer networks, both using the control over data location and network topology. In this
classic search algorithms such as flooding and random walk, as paper we are mainly concerned on comparative study of
well as a new hybrid algorithm proposed in this paper. This
various available search Algorithms in unstructured P2P
hybrid algorithm uses two level random walks for the adaptive
probability search (APS). We compare the performance of the
systems also it present the design of new proposed search
search algorithms on several graphs corresponding to common algorithm for unstructured P2P systems.
topologies proposed for peer-to- peer networks. In this paper it
is found that Local Indices algorithm gives the average
performance. Intelligent search and Routing Indices have higher
2. Unstructured P2P systems
bandwidth. Further work can be done on reducing the size of the In unstructured networks, the placement of data (files) is
query message subsequently it will reduce the bandwidth. APS is completely unrelated to the overlay topology. Since there is
the efficient technique among all. Further it can be improved by no information about which nodes are likely to have the
proposed search algorithm which uses two-level k-walker
relevant files, searching essentially amounts to random
random walk with APS instead of k-walker random walk.
Advantages of two level walk will further reduce collision of
search, in which various nodes are probed and asked if they
nodes and can help in searching the distant nodes in the have any files matching the query. These systems differ in
network. But it may slightly increase the response time. the way in which they construct the overlay topology, and
the way in which they distribute queries from node to node.
Keywords: peer-to-peer networks, adaptive probability search. The advantage of such systems is that they can easily
accommodate a highly transient node population. The
1. Introduction disadvantage is that it is hard to find the desired files
without distributing queries widely. For this reason
P2P network is a distributed network composed of a large unstructured p2p systems are considered to be unscalable.
number of distributed, heterogeneous, autonomous, and However work is done towards increasing the scalability of
highly dynamic peers in which participants share a part of unstructured systems. Napster, Gnutella, Kazaa,
their own resources such as processing power, storage Morpheus[1] are various unstructured P2P systems.
capacity, software and files. The participant in the P2P
network can act as a server and a client at the same time.
P2P systems constitute highly dynamic networks of peers 3. Searching in unstructured Systems [4]
with complex topology. This topology creates an overlay
Initially for the purpose of searching specific data item,
network, which may be totally unrelated to the physical
flooding which is basically BFS was used but it generates a
network that connects the different nodes (computers). P2P
large number of duplicate messages and also does not scale
systems can be differentiated by the degree to which these
well so a number of alternative schemes have been proposed
overlay networks contain some structure or are created ad-
to address the above problem.
hoc. Network structure here means the way in which the
content of the network is located with respect to the network
These works include iterative deepening, k-walker k-walker BFS, subset of High High
random walk, modified random BFS, two-level k-walker random neighbor
Blind
random walk, directed BFS, intelligent search, local indices walk
based search, routing indices based search, attenuated bloom 2 Lvl k- BFS, subset of Low Low
filter based search, adaptive probabilistic search, and walker neighbor
dominating set based search. Blind
random
walk
Searching strategies in unstructured P2P systems are APS BFS, subset of Medium Medium
either blind search or informed search. In a blind search neighbor
such as iterative deepening, no node has information about Informed
the location of the desired data. In an informed search such
as routing indices, each node keeps some metadata about the Based on scalability, response time (RT), success rate(SR)
data location. To restrict the total bandwidth consumption, and bandwidth various searching methods are compared as
data queries in unstructured P2P systems may be terminated follows-
prematurely before the desired existing data is found;
Algorithm Search Query Message Node
therefore, the query may not return the desired data even if
method forward- over-head dupli-
the data actually exists in the system. An unstructured P2P
ing cation
network can not offer bounded routing efficiency due to lack
Flooding No High Medium Low
of structure.
The searching schemes in unstructured P2P systems can Iterative Yes High Medium Medium
also be classified as deterministic or probabilistic. In a Deepning
deterministic approach, the query forwarding is Local Yes Medium Medium Medium
deterministic. In a probabilistic approach, the query Indices
forwarding is probabilistic, random, or is based on ranking. Directed Yes Medium Medium High
Another way to categorize searching schemes in BFS
unstructured P2P systems is regular-grained or coarse-
grained. In a regular-grained approach, all nodes Intelligent Yes Medium Medium High
participate in query forwarding. In a coarse-grained scheme, Search
the query forwarding is performed by only a subset of nodes
in the entire network. Routing Yes Medium Medium High
indices
4. Comparison of Existing Search Algorithms
Std. Yes High Medium Low
Based on search method, Query forwarding, Message random
Overhead and node duplication various searching methods walk
are compared as follows- k-walker Yes Medium Medium low
random
Algo- Search Query Message Node walk
rithm method forward- over-head dupli- 2 Lvl k- Yes Medium Medium low
ing cation walker
Flooding BFS, Broadcast High High random
Blind walk
Iterative BFS, Broadcast High High APS Yes Low High Medium
Deepning
Blind
Local BFS, Broadcast Medium Mediu
Indices Among those algorithms, Adaptive Probability Search
Informed m
(APS) is the most efficient algorithm. APS is based on k-
Directed BFS, Partial Medium High walker random walk and probabilistic (not random)
BFS Broadcast forwarding. Another interesting algorithm is Two-Level
Informed
Intelligen BFS, subset of Medium Mediu Random Walk in which walkers are searching for an object
t Search neighbor in two levels. So it reduces the redundancy of nodes.
Informed m
Routing BFS, subset of Medium Mediu 5. Adaptive Probability Search (APS) [6]
indices neighbor In the Adaptive Probabilistic Search (APS) [6], it is
Informed m
assumed that the storage of objects and their copies in the
Std. BFS, One Low Low network follows a replication distribution. The number of
random neighbor query requests for each object follows a query distribution.
Blind
walk The search process does not affect object placement and the
P2P overlay topology.
The APS is based on k-walker random walk and

probabilistic (not random) forwarding. The querying node Figure 1 shows an example of how the search process
simultaneously deploys k walkers. On receiving the query, works. Node A initiates a request for an object owned by
each node looks up its local repository for the desired object. node F using two walkers. Assume that all index values
If the object is found, the walker stops successfully. relative to this object are initially equal to 30 and the
Otherwise, the walker continues. The node forwards the pessimistic approach is used. The paths of the two walkers
query to the best neighbor that has the highest probability are shown with thicker arrows. During the search, the index
value. The probability values are computed based on the value for a chosen neighbour is reduced by 10. One walker
results of the past queries and are updated based on the with path (A,B,C,D) fails, while the second with path
result of the current query. The query processing continues (A,E,F) finds the object. The update process is initiated for
until all k walkers terminate either successfully or fail (in the successful walker on the reverse path (along the dotted
which case the TTL limit is reached). To select neighbors arrows). First node E, then node A increase the value of
probabilistically, each node keeps a local index about its their indices for their next hops (nodes F, E respectively) by
neighbors. There is one index entry for each object which 20 to indicate object discovery through that path. In a
the node has requested or forwarded requests for through subsequent search for the same object, peer A will choose
each neighbor. The value of an index entry for an object and peer B with probability 2/9 (= 20 20+40+30), peer E with
a neighbor represents the relative probability of that probability 4/9 and peer G with probability 3/9.
neighbor being selected for forwarding a query for that
object. The higher the index entry value the higher the APS requires no message exchange on any dynamic
probability. Initially, all index values are assigned the same operation such as node arrivals or departures and object
value. Then, the index values are updated as follows. When insertions or deletions. The nature of the indices makes the
the querying node forwards a query, it makes some guess handling of these operations simple: If a node detects the
about the success of all the walkers. arrival of a new neighbour, it will associate some initial
The guess is made based on the ratio of the successful index value with that neighbour when a search will take
walkers in the past. If it assumes that all walkers will place.
succeed (optimistic approach), the querying node pro- If a neighbour disconnects from the network, the node
actively increases the index values associated with the removes the relative entries and stops considering it in
chosen neighbors and the queried object. Otherwise future queries. No action is required after object updates,
(pessimistic approach), the querying node proactively since indices are not related to file content. So, although our
decreases the index values. Using the guess determined by algorithm actively uses information, its maintenance cost on
the querying node, every node on the query path updates the any of these events is zero, a major advantage over most
index values similarly when forwarding the query. current approaches.
Upon walker termination, if the walker is successful,
5.1 Discussion on APS
there is nothing to be done in the optimistic approach. If the
walker fails, index values relative to the requested object Each node stores a relative probability (an unsigned
along the walker’s path must be corrected. Using integer) for each of its neighbours for each requested object.
information available inside the search message, the last So for R such objects and N neighbours, O(R x N) space is
node in the path sends an “update” message to the preceding needed.
node. This node, after receiving the update message, For a typical network node, this amount of space is not a
decreases its index value for the last node to reflect the burden. On nodes with limited storage capacities, index
failure. The update procedure continues along the reverse values for objects not requested for some time can be erased.
path towards the requester, with intermediate nodes This can be achieved by assigning a time-to-expire value on
decreasing their local index values relative to the next hops each newly-created or updated index. Each search or update
for that walker. Finally, the requester decreases its index message carries path information, storing a maximum of
value that relates to its neighbour for that walker. If we TTL peer addresses. Alternatively, each node can associate
employ the pessimistic approach, this update procedure the search and requester node IDs with the preceding peer in
takes place after a walker succeeds, having nodes increase the path of the walker. Updates then follow the reverse path
the index values along the walker’s path. There is nothing back to the requester. This information expires after a
to be done when a walker fails. certain amount of time.The number of messages exchanged
by APS method to terminate in the worst case will be (2 x k
x TTL) where all walkers (k walkers) travel TTL hops and
then invoke the update procedure, so the method has the
same complexity with its random counterpart. The only
extra messages that occur in APS are the update messages
along the reverse path. This is where the two index update
policies are used.
Along the paths of all k walkers, indices are updated so
that better next hop choices are made with bigger
probability. Learning feature includes both positive and
Figure 1. Searching object using pessimistic approach of negative feedback from the walkers in both update
APS with walkers. approaches. In the pessimistic approach, each node on the
walker’s path decreases the relative probability of its next

hop for the requested object concurrently with the search. If
the walker succeeds, the update procedure increases those
index values by more than the subtracted amount (positive
feedback). So, if the initial probability of a node for a certain
object was P, it becomes bigger than P if the object was
discovered through (or at) that node and smaller than P if
the walker failed. Conversely, if many of our walkers hit
their targets on average, the optimistic approach should be
considered. This is the only invariant we require from our
update process. Figure 4. Hits per Query vs. number deployed walkers for
The learning process in the optimistic approach operates APS and Random walk algorithms
in an opposite fashion, Learning is important to achieve
both high performance and discovery of newly inserted
objects. Unlearning helps our search process adjusts to
object deletions and node departures, redirecting the walkers
7. Two- Level Random Walk[7]
elsewhere. All the nodes participating in the search get It’s an efficient search algorithm which increases the total
benefited from the process. number of nodes searched for a certain total number of
search step, and reduces the redundancy or average number
Besides standard resource-sharing in P2P systems, APS of times a particular node is searched. It works in the
achieves the distribution of search knowledge over a large following manner. When a node wishes to send a query with
number of peers. a certain search key, it composes a search message and
broadcasts it to k1 randomly selected neighbours. The
6. Performance of APS [6] message has an initial TTL1 = l1 hops. When an
The main metrics used to evaluate the performance of a intermediate node receives this message, it checks the TTL1
search algorithm are the success rate, the number of timer. If the latter is still more than 0 then it decrements the
discovered objects (Hits per Query) and the number of timer by one, selects one random neighbour and forwards
messages produced. the message to it. This process continues until one of the
nodes, say node E, receives the message with an expired
TTL1 timer (i.e. TTL1 = 0). We call such a node an edge
node. The message will then “explode” into k2 search
messages forwarded from this node. Specifically, node E
will compose a message with TTL1=0, and a second timer
TTL2=l2. It will then randomly select k2 of its neighbours,
excluding the one it just received the message from, and
broadcast the message to them. Figure 1 shows an example
illustrating this process. At level one, a source node sends
k1 random messages to a set of k1 randomly selected nodes
of its neighbours. This constitutes k1 threads (or random
walks) which travel from the source node to the edge nodes
Figure 2. Success rate vs. number deployed walkers for APS (a node where TTL1 expires). Each of the k1 threads will
and Random walk algorithms then explode into k2 threads (with TTL2 = l2 ) at each of
the k1 edge nodes. This algorithm reduces redundancy by
decreasing the average number of times a node is searched.
In the one-level k-walk algorithm k random threads are
generated from the source and they are likely to have
“thread collisions” (i.e. threads run into each other)
especially near the source. This results in having redundant
hits in the same nodes (nodes being searched multiple
times). On the other hand, the two-level algorithm sends
fewer threads from the source node which results in a
smaller probability of thread collisions near the source. Each
Figure 3. Message production vs. number deployed walkers of the k1 threads will then explode into k2 threads once it is
for APS and Random walk algorithms ”sufficiently” away from the source and the other threads.
This way, the same number of search threads can be
generated (k=k1*k2) but with a larger number of nodes
searched and a smaller probability of redundant searches to
the same nodes using the same number of total search steps.
8. Enhancing Performance of APS [2] V. Vishnumurthy and P. Francis. On heterogeneous

overlay construction and random node selection in
Proposed algorithm uses two-level random walk for the unstructured P2P networks. In Proc. IEEE Infocom,
existing APS algorithm[6] instead of k-walker random walk. 2006.
Advantage of two-level walk[7] over one-level walk is that it [3] .MA. Jovanovic, Modelling large-scale peer-to-peer
increases the total number of nodes searched for a certain networks and a case study of gnutella. Master's thesis,
total number of search step, and reduces the redundancy or Department of Electrical and Computer Engineering
average number of times a particular node is searched. So and Computer Science, University of Cincinnati, June
collision of nodes can be further reduced and also distant 2000.
objects can also be search efficiently. Two level walk will [4] .Xiuqi Li and Jie Wu, Searching Techniques in Peer-to-
also help in further reducing message overhead. Only Peer Networks, Department of Computer Science and
disadvantage will be increased in response time. Engineering, Florida Atlantic University.
[5] V. Vishnumurthy and P. Francis. A comparison of
structured and unstructured P2P approaches to
9. Algorithm of proposed Technique heterogeneous random peer selection. In Proc. Usenix
Assumptions Annual Technical Conference, 2007.
[6] D. Tsoumakos and N. Roussopoulos. Adaptive
k1 = k2 = k = k3 – Number of walkers in each level Probabilistic Search (APS) for Peer-to-Peer Networks.
ttlcount – counter for ttl value Technical Report CS-TR-4451, Un. of Maryland, 2003.
l1 = ttl2 = ttl - Time to live for each level [7] Imad Jawhar and Jie Wu, A Two-Level Random Walk
level – variable for level number Search Protocol for Peer-to-Peer Networks, Department
kcount – counter for k3 i.e. number of walkers of Computer Science and Engineering, Florida Atlantic
Select a querying node University.
Kcount = 0 [8] Beverly Yang Hector Garcia Molina, Improving Search
level = 1 in Peer-to-Peer networks, Computer Science
Department, Stanford University.
while (level <= 2)
{
while (kcount <= k3)
{
while (ttlcount <= ttl)
{
select a neighbouring node by
applying APS and Process the node;
if object is not found
then
increment ttlcount by one
continue;
else
come out of the loop (exit); }
increment kcount by one;
}
increment level by one;
}
10. Conclusion
In this research work, various searching techniques in
unstructured p2p networks are studied. Comparative study
of these techniques is done. A new Search Technique is
proposed which helps in further enhancing the performance
of APS.
References
[1] Stephanos Androutsellis-Theotokis, ‘A Surver of Peer-

To-Peer File Sharing Technologies’, White Paper,
ELTRUN, Athens University of Economics and
Business, Greece, 2002.
Steganography Security for Copyright Protection of

Digital Images Using DWT
K.T.Talele1, Dr.S.T.Gandhe2, Dr.A.G.Keskar3
1
Electronics Engineering Dept.,Sardar patel Institute of Technology,
Andheri(w)Mumbai, India
kttalele@yahoo.co.uk
2
Electronics Engineering Dept.,Sardar patel Institute of Technology,
Andheri(w)Mumbai, India
stgandhe@gmail.com
3
Electronics Engineering Dept , Visvesvaraya National Institute of Technology,
Nagpur,India
avinashkeskar@yahoo.com
Abstract: The proposed system combines cryptography and Signal to Noise Ratio (PSNR) and Correlation Coefficient
steganography for copyright protection of digital images using (CC) values of the extracted logo for different attacks[7].
DWT. The proposed algorithms tested on various attacks such as
median, wavelet compression, fading and resizing by comparing
different performance parameters such as mean square error,
peak signal to noise ratio, correlation coefficient and the results
2. Proposed System
are very encouraging. The sensitivity is least observed in DWT The block diagram for proposed system is as shown in
method where the watermark maintains a fair level of resistance figure 1.First the logo is encrypted and then it is inserted in
to noise and other attacks. The proposed system can be used for a given image using DWT and the logo is extracted and
enhanced copyright protection, detection of misappropriated then it is decrypted to get original logo.
images; detect alternation of images stored in a digital library.
Keywords: Cryptography, encryption, Decryption,
Steganography.
1. Introduction
Security is one of the major concerns in today’s age. Unlike
the past, most of the transactions between people take place
over the internet. But internet itself is not a secure medium.
So, when it comes to sending highly important documents
over the internet, an extra precaution has to be taken. In
other cases, authenticity of digital data is a big concern.
With the widespread usage of digital media, demand for
copyright protection has increased manifold as it is
evidently seen in the audio records industry. The extra
precaution for copyrighting digital media is required here as
well. Figure 1. Block Diagram of Proposed System
One of the ways to take this extra precaution is to use
Steganography. Steganography helps to hide the content of 3. Algorithm
interest which is to be protected, inside any image, audio or
video file. To further ensure that interception of content 3.1 Cryptography Algorithm
does not happen, the content can be encrypted using one the
The encryption algorithm works on the approach of
popular Cryptographic algorithms[1][2]. Fragile
swapping pixel values of randomly generated 128
Watermarking is used in the case where tamper detection
locations in the row of every logo. It is important that the
and authenticity have a higher priority whereas Robust
set of 128 locations so generated are done with the help of
Watermarking deals with copyright protection[3][4] [5][6].
a password and that they are all unique. The steps of the
The proposed system hides a logo in images using DWT .
algorithm are as follows:
The watermark should be imperceptible to anyone and
sensitive to any kind of tampering done on the image under a) Take the input logo.
consideration. The system is compared for various b) Ask the user to enter a 8 bit key.
algorithms for embedding the logo. The algorithms are c) Generate 8 random vectors of size 1X128.
compared on the basis of Mean Square Error (MSE), Peak
d) Specify the ‘state’ of the random number generator h) Perform the two level inverse discrete wavelet
by giving the ASCII value of each character in the transform of host image (all three planes) by using
key for every random vector. This will generate 8 approximation coefficients of three planes of host
random vectors. image.
e) Generation of Input vector: i) Find the mean square error (MSE) and peak signal
Ø The outer loop controls the column to noise ratio (PSNR) and the correlation
traversal of the watermark logo. coefficient (CC) between the original host image
Ø The middle loop controls the selection and invisible watermark image by using the related
of random vectors previously generated. formulae as these are the important performance
Ø The inner loop controls the row parameters.
traversal of the watermark logo.
Ø Inside the innermost loop, we select a 3.3 Performance Parameters
random vector based on the value of the middle
loop. With every turn of the innermost loop we 3.3.1 Peak Signal to Noise Ratio(PSNR) and
take two consecutive values, r(1,k,j) and Mean Square Error(MSE).
The imperceptibility of a watermark is measured by the
r(1,k+1,j) from this random vector and swap
watermarked image quality in terms of Peak-Signal-to-Noise
the corresponding location values from the Ratio (PSNR) (in dB). Most common difference measure
watermark image in the same column and between tow images is the mean square error. The mean
store it in another array called ‘encrypted’ at square error measure is popular because it correlates
the same position. reasonably with subjective visual tests and it is
mathematically tractable.
f) Every time the middle loop finishes, the random
Consider a discrete image A(m, n) for m=1,2,……M and
vectors are considered corresponding to the first n=1,2,……N, which is regarded as a reference image.
character of the password. Consider a second image Â(m, n), of the same spatial
g) This cycle continues till the last column is covered. dimension as A(m, n), that is to be compared to the
h) Thus the watermark image is encrypted into a new reference image.
image file ‘encrypted.bmp’. Under the assumption that A(m, n) and Ã(m, n) represent
samples of a stochastic process,
3.2 Proposed Steganography in color images MSE is given as
Algorithm
a) Consider any color image having size 512 X 512 as

a host image. If size of host image is not 512 X 512 Where E (·) is the expectation operator.
then make it 512 X 512.
b) Split the image into three planes viz. Red, Green and The normalized Mean Square Error is given as
Blue.
c) Decompose the host image (all three planes) by
using discrete wavelet transform. Store the first
Normalized mean square error for deterministic image
level coefficients i.e. LL1, LH1, HL1, HH1 as first
level watermark key coefficients of host arrays is defined as
image.[8][9]
d) Approximation coefficient of first level is LL1 which
is further decomposed into new coefficients i.e. Image error measures are often expressed as signal-to-noise
LL2, LH2, HL2, HH2 as second level watermark ratio,
key coefficients of host image.
e) Consider the gray scale image having size 128 X
128 as a logo to be hidden. If size of watermark
logo is not 128 X 128 then make it 128 X 128. We use PSNR to determine the difference between original
image A (m, n) and the watermarked image Ã (m, n).
f) Decompose the watermark logo by using discrete
wavelet transform. Store the first level The value of mean square error should be minimum and the
approximation coefficients i.e. LL1, LH1, HL1, value of peak signal to noise ratio should be as maximum as
HH1 as first level watermark key coefficients of possible.
watermark logo.
g) Insert coefficients of LH2 part of host image by LL1
part pixel by pixel.
3.3.2 Correlation Coefficient (CC) Figure 3(a) shows original logo and figure 3(b) shows
The robustness performance of watermark extraction is watermark logo .Figure 4(a) shows watermarked image and
evaluated by normalized correlation coefficient, r, of the figure 4(b) shows extracted logo.
extracted watermark A and the original watermark B.
4.3 Comparative Study Of Different Watermarking
Algorithms for different original images
The algorithm of insertion is applied to five different input
images as shown in figure 5 through figure 9 and compared
Where A and B respectively, the normalized original and the result using different performance parameters. The
watermark image by subtracting its corresponding means comparison is as shown in table 1.. We can insert secret
value. The magnitude range of r is [0, 1], and the unity information in these images for copyright protection.
holds if the extracted image perfectly matches the original
one.
The correlation coefficient is used to compare original
image and the watermarked image, and also for comparing
original watermark and the retrieved watermark.
4. Results and Discussion

The algorithms with various attacks are implemented (a) (b)
using MATLAB. The results of various algorithms are Figure 5: (a) Original Image (b) Watermarked Image for
shown through figure 2 to figure 19. lena image.
4.1 Cryptography
(a) (b)
(c) (a) (b)
Figure 2: (a) Original Logo (b) Encrypted Logo (c) Figure 6: (a) Original Image (b) Watermarked Image for
Decrypted Logo medical image
Figure 2 (a) shows original logo and figure 2(b) shows

encrypted logo and figure 2(c)shows decrypted logo.
4.2 Steganography
4.2.1 Steganography in images using DWT
(a) (b)
Figure 7: (a) Original Image (b) Watermarked Image for
satellite image.
(a) (b)
Figure 3: (a) Original Image (b) Watermarked Logo (to be
hidden)
(a) (b)
satellite Scene.
(a) (b)
Figure 4: (a) Original Image (b) Extracted Logo
(a) (b)
text image. (c)
Figure 10: Graph of (a) PSNR, (b) MSE, (c) CC for
Table 1: PSNR, MSE, and CC for original image and steganography in images using DWT
watermark image for Invisible Watermarking for five
different images Graphically these values are shown in figure 10.
4.4 Attacks
The algorithm is tested for various attacks[7] such as

median filter, wavelet compression, fading, noise, resizing
etc.
4.4.1 Median Filter
The value of PSNR is sufficiently high, MSE is very low and The figure 11 through figure 14 shows the algorithm is
CC is nearly equal to 1.So this algorithm has created tested for median filter under four mask sizes 3X3, 5X5,
minimum disturbance to host image and perceptually both 7X7, 9X9. For each of these cases peak signal to noise ratio,
the images are alike. mean square error and correlation coefficient are calculated
and are as shown in figure 15.
(a) (b)
Figure 11: (a) Median filtered Watermark Image with mask
(a)
size 3X3, (b) Extracted Watermarked Logo
(a) (b)
Figure 12: (a) Median filtered Watermark Image with mask
size 5X5, (b) Extracted Watermarked Logo
(b)
4.4.2 Wavelet Compression.
The algorithm is tested for wavelet compression and is as

shown in figure 16. The performance parameters are
MSE=0.0060, PSNR=46.02619, CC=0.9907
(a) (b)
Figure 13: (a) Median filtered Watermarked Image with
mask size 7X7, (b) Extracted Watermark Logo
(a) (b)
Figure 16 (a) Wavelet compressed Watermarked image (b)
Extracted Watermark Logo
(a) (b) 4.4.2 Fading

Figure 14: (a) Median filtered Watermarked Image with
mask size 9X9, (b) Extracted Watermark Logo The algorithm is tested for fading where each pixel value of
image is increased by 50 and extracted watermark logo and
is as shown in figure 17. The performance parameters are
MSE=0.0561, PSNR=36.5775, CC=0.9954
(a) (a) (b)

Figure 17: (a) Faded Watermarked Image (Original+50),
(b) Extracted Watermark Logo
4.4.3 Noise
(b)
(a) (b)
Figure 18: (a) Noise added Watermarked image,
(b) Extracted Watermark Logo
The algorithm is tested for noise and is as shown in figure

18. The performance parameters are MSE=0.0322
PSNR=38.9902 CC=0.8103
4.4.4 Resizing
The algorithm is tested for resizing where watermark image
(c) resized by a scaling factor of 2.and is as shown in figure
Figure 15: Graph of (a) PSNR, (b) MSE, (c) CC for 19.The performance parameters are MSE=0.0028
watermarked image for different mask size of median filter PSNR=49.5556 CC=0.9972.
conference INDICON07 organized by IEEE India

council at Bangalore on 6th,7th,8th September 2007.
[9] S.T.Gandhe, K.T.Talele, A.G.Keskar “Intelligent Face
Recognition: A Comparative Study” ICGST’s
Graphics Vision & Image Processing Journal,
Vol.7,Issue 2,2007, page no.53-60.
(a) (b) Authors Profile

Figure 19: (a) Resized Watermarked Image, (b) Extracted
Watermark Logo
K.T.Talele is aAssistant Professor in
Electronics Engg Dept, S. P. Institute of
5. Conclusion Technology, Mumbai., India He is a member
of IEEE. His area of interest includes DSP,
Thus we have developed a system which successfully Image Processing and Multimedia Comm. He
embeds a logo imperceptibly in a given cover image. The has published twenty papers in National
logo is sensitive to tampering and compression. On the other Conferences and three papers in International
hand, the sensitivity is least observed in DWT method journal and 14 papers in international
where the watermark maintains a fair level of resistance to conference
noise and other attacks. In order to further prevent the
watermark logo from detection, we encrypt it before Dr.S.T.Gandhe is Professor in Electronics
embedding it in an image. The algorithm used is based on Engg Department, S. P. Institute of
swapping the pixel values of the watermark logo by using Technology, Mumbai, India.. He is a member
password generated random vectors. The system is not of IEEE. His area of interest includes Image
immune to compression techniques. Whenever the Processing, Pattern Recognition and Robotics.
watermarked is stored in compressed format, a considerable He has published 12 papers in National
loss of information is observed. The system can be extended Conferences and 10 papers in International
to possible counter the compression effect. Extracted conference and 3 papers in international
journal.
watermark face image can be recognized by using
automated intelligent face detection algorithm[11].The
proposed system can be used for enhanced copyright
Dr. A.G.Keskar is a Professor in Electronics
protection, detection of misappropriated images, detect Engg. Dept,Visvesvaraya National Institute of
alternation of images stored in a digital library. Technology, Nagpur. He is a senior member
of IEEE. He has published 10 Journal papers
References and published 25 papers in International
Conferences. His area of interest includes
[1]W.Diffie and M.E.Hellman, “New Directions in Fuzzy logic, Embedded System and Machine
Cryptography”, IEEE trans. On Information Vision.
Theory, Vol.IT-22, No.6, Nov.1976.
[2] B.M.Macq, J.J.Quisquater, “Cryptography for Digital
TV Broadcasting”, Proc. of the IEEE, Vol.83, No.6,
Jun1995, pp 944-957.
[3] J. M. Acken, “How Watermarking Value to Digital
Content”, Comm. of ACM, July 1998, Vol 41, No.7, pp
75-77.
[4] Yongjian Hu, Sam Kwong, Jiwu Huang, “An Algorithm
for Removable Visible Watermarking”; IEEE
transactions on circuits and systems for video
technology, vol. 16, no. 1, January 2006.
[5] Zhang Fan, Zhang Hongbin, “Capacity and Reliability of
Digital Watermarking”, International Conference on
the Business of Electronic Product Reliability and
Liability 2004.
[6] Fan Zhang, Hongbin Zhang, “Digital Watermarking
Capacity and Reliability”, Proceedings of the IEEE
International Conference on E-Commerce Technology
2004.
[7] Chun-Hsiang Huang and Ja-Ling Wu, “Attacking
Visible Watermarking Schemes”, IEEE transactions on
multimedia, vol. 6, no. 1, February 2004
[8] S.T.Gandhe and K.T.Talele and Dr. A.G. Keskar “ Face
Recognition Using DWT+PCA” International
Design and Implementation of A GUI Based on

Neural Networks for the Currency Verification
1
Ajay Goel , 2O.P.Sahu, 3 Rupesh Gupta and 4Sheifali Gupta
1
Department of CSE, Singhania University, Rajasthan, India,
goelajay1@gmail.com
2
Department of ECE, N.I.T. Kurukshetra, India,
opsahu_reck@yahoo.co.in
3
Department of ME, Singhania University, Rajasthan, India,
rup_esh100@yahoo.co.in
4
Department of ECE, Singhania University, Rajasthan, India,
sheifali@yahoo.com
a content-based watermarking technique. When the photo is

Abstract: The technological development in the era of image
processing and machine vision has two faces. One face is to help changed, the image with the watermark is of course lost and
the society by automation and the other side has serious this just requires that the watermark hidden in the owner’s
implications on the society like cyber crimes e.g. web hacking, passport is robust to one cycle of print and scan.
cracking, etc. One of the emerging crimes is preparing fake Considering the special characteristic of FFT on rotation,
legal documents in now days. These documents have social scaling and cropping, Lin [7][8] has carried out the research
values, like a degree certificates certifies the educational
on fragile watermarking rather early and obtained many
qualification of a person. The legal documents contain lots of
symbols like kinegrams, hologram, watermark etc. by which we useful conclusions on the distortion brought by print and
can verify the authenticity of these documents. Digital scan. Re-searchers in China [9] began to hide some
watermarking emerged as a tool for protecting the multimedia information in printing materials, using the function offered
data from copyright infringement. In this paper an attempt has by PhotoShop. All these are focused on the watermark
been made to verify the legal document on the basis of robust to one cycle of print and scan.
watermark. In this work the correlation mapping with neural
network is used for extracting the watermark to verify the legal
documents. This technique gives elevated accuracy rate with 2. Basic Concept
fewer times to extract watermark. This method can be 1. Since water mark making requires highly efficient
implemented also in additional applications like stamp technique and the water mark can be seen only by its
verification, currency verification etc.
shadow, the water mark can be effective key to certify the
Keywords: Watermarking, Multilayered-Network, Certifying, currency note.
Epochs. 2. In certifying the currency note, since normally using
currency note is folded, sometimes noise occurred, it
1. Introduction needs feedback learning of water mark of used currency
With the increasing use of internet and effortless copying, note. The back propagation neural network is suitable to
tempering and distribution of digital data, copyright certify the water mark, because it can design many layers
protection [1] for multimedia data has become an important for many nodes network, that it is used to recognize the
issue. There are lots of symbols present on the printed complicate pattern [3].
document for their identification but in this work watermark
has been chosen for the verification. Basically legal They consider correlation as the basis for finding matches of
documents can be verified by two methods: first-line a sub- image w(x,y) of JxK within an image f(x,y) size
inspection methods and second-line inspection methods MxN, we assume that J≤M and K≤N. They prepare the
.First-Line Inspection Methods are Watermarks, Ultraviolet template of each type of note then apply correlation on each
Fluorescence, Intaglio Printing are further divided in Micro stored note with on which we are testing. Zero value of
text & Holograms and Kinegrams and second is Second- correlation coefficient gives the location of the
Line Inspection Methods Isocheck / Isograms. Recent watermark[10].
public literatures show that some researchers have tried to
apply watermarking into printing system. In geometric 3. Certifying
transform. Pun [5] has devised a watermarking algorithm To certify the watermark it is inputted to back-propagation
robust to printing and scanning. The PhotoCheck software neural network. Result of neural network is used to certify
developed in AlpVision Company by Kutter [6] is mainly the currency note. First neural network must be trained by
fo-cused on authentication detection of passports. As a sending the ideal watermark to it. Size of input is sent to
passport belongs to the owner with his photo, this belongs to neural network about 4225 nodes.
4. Trainable Multilayered- network
BP method is given by eq(5.2) as [6]
∆ W ij = ηδ i x j + α ∆ W ij ( k − t ) (1.1)
Where Wij is the weight connecting an output of unit j into

an input of unit i, η is the step size, a is a momentum
Figure 1. Flowchart of a system
coefficient, and xj is an input signal from unit j. The
The edge information from the shadow of water mark is
quantity δi is an error term, computed as
derived from shining image, and it is inputted to neural
network for certifying. The edge information all of currency
 O ' i ( ti − Oi )
δi =  O 'i
∑ W mi δ m (1.2)
note is inputted to neural network for checking.

6. The Proposed Scheme
Where ti is a desired signal to unit I, Oi is the actual output
We are applied an algorithm which tries to find the
of unit i, and O’i is the derivative of O’i. Weight is adjusted
watermark more effectively. First step to break the image
according to what is the target. When there is much
into different pieces having size of each piece is equal to
difference between input and target then neural network
ideal watermark size. Then it finds the correlation among
needs more training. When epochs are increased neural
each piece with ideal watermark. Correlation process gives
network gives best response as the number of epochs
correlation coefficients then the method picks up the piece
increased training is better, although with increasing
which gave greatest correlation coefficient, because this
number of time taken to train the network is more but target
piece has most probably of containing watermark. The piece
is achieved with less mean squared error [4].
found in the previous step has been given to the neural
network for verification. We trained the neural network with
5. Methodology ideal watermark as a target.
In this methodology first the searching of watermark is done This is implemented in matlab 7.0 and vb.net. GUI has been
for that a document is split into blocks. Block size of each created in vb.net and main implementation has been done in
block is equal to the size of ideal watermark. After splitting matlab 7.0. Then GUI is linked with matlab7.0 with the
each block is stored in different variable. After splitting help of M-files.
document into blocks further process is that to correlate each
blocks with ideal watermark. This correlation will give the
correlation coefficient. Block which contain watermark
gives the highest correlation coefficient. Now we extracted
the block which gives us highest correlation coefficient and
give it to neural network. Correlation coefficient[3] is given
by the equation (1.3)
c( s, t ) = ∑∑ f ( x, y)w( x − s, y − t ) (1.3)
Figure 2. Main form of GUI, “Document verification system”
Where x, y are co-ordinates of selected block and s=0, 1, In fig 2 photo of the currency note 1000 has been shown.
2…. M, t= 0,1,2,……N Where M×N is size of ideal After acquiring image the neural network has been trained.
watermark. Each pixel of the selected block is matched with Following figure 3 shows the training of neural network.
each corresponding pixel of ideal watermark block.
Difference between these is calculated called correlation
coefficient. Correlation coefficient of each block has been
calculated with the same procedure. Now select the block
which contains the highest correlation coefficient. The
system software flowchart is shown in figure 1 can be
described; the location of the water mark is detected[7].
Figure 3. TrainingoOf Neural Network

After clicking ok main form displayed again then we have to We have implemented our technique on the Indian currency
verify it. By selecting verify from the menu bar will start and Indian postage stamp, but their technique is
process of verification. Following graph in figure 4 shows implemented on Thai currency. However output of this
the watermark accuracy present on the document. technique is also different it did not show the accuracy of
watermark. This technique searches the watermark into
whole image while our technique will split the image into
blocks and apply correlation on each block with ideal
watermark, which gives us a correlation coefficient. The
value of correlation coefficient gives us an idea of similarity
between two images. This technique takes shorter time to
find the watermark in the note.
Figure 4. Accuracy of two watermark present on the Reference

document. [1] Ingermar J. Cox, Matthew L. Miller, and Jeffrey
A.Bloom, Digital Watermarking, Morgan
KaufmannPublishers, 2002
7. Result [2] Francisco J. Gonzalez-Serrano, Harold. Y. Molina-
The note is divided into two categories one is called training Bulla, and Juan J. Murillo- Fuentes,” Independent
set other is called testing set. After inputting the note’s component analysis applied to digital image
image of currency notes, the location of watermark has been watermarking,” International Conference on Acoustic,
detected by correlation mapping by splitting image into Speech and Signal Processing (ICASSP), vol. 3, pp.
blocks then finding correlation of each block. Then block is 1997-2000, May 2001.
given to neural network for certification. Neural network is [3] Dan Yu, Farook Sattar, and Kai-Kuang Ma,
trained through training set. After the goal is accomplished “Watermark detection and extraction using
we test the neural network by tested data. In this neural independent component analysis method,” EURASIP
network Mean squared error has been used. So training Journal on Applied Signal Processing, vol. 1, pp. 92–
graph has been plotted target versus input. So the neural 104, 2002.
network has been trained with 1000 epochs. [4] Minfen Shen, Xinjung Zhang, and Lisha Sun, P. J.
Beadle, F. H. Y. Chan, “A method for digital image
watermarking using ICA,” 4th International
Symposium on Independent Component Analysis and
Blind Signal Separation (ICA 2003), Nara, Japan,
April 2003, pp. 209-214.
[5] Ju Liu , Xingang Zhang, Jiande Sun, and Miguel
Angel Lagunas, “A digital watermarking scheme based
on ICA detection,” 4th International Symposium on
Independent Component Analysis and Blind Signal
Separation, (ICA 2003), Nara, Japan, April 2003, pp.
Figure 5. Training neural network with 1000 epochs 215-220.
[6] Stephane Bounkong, Boremi Toch, David Saad, and
This bar chart is plotted on scale 1. First watermark shows David Lowe, “ICA for watermarking digital images,”
the accuracy of 99% and second watermark shows the Journal of Machine Learning Research 4, pp. 1471-
accuracy of 100% which are above threshold so we can say 1498, 2003.
that the currency note is real. This accuracy is checked on [7] Viet Thang Nguyen and Jagdish Chandra Patra,
the basis of percentage matching with real watermarks “Digital image watermarking using independent
because neural network has been trained with the real component analysis,” PCM 2004, Lecture Notes in
watermark and gives mean squared error when tested. Two Computer Science 3333, pp. 364-371, Springer-
neural network for two watermark has been created Verlag, 2004.
differently. If accuracy is below threshold then the note is [8] Thai Duy Hien, Zensho Nakao, and Yen-Wei Chen,
rejected. “Robust multi-logo watermarking by RDWT and
ICA”, Signal Processing, Elsevier, vol. 86, pp. 2981-
2993, 2006.
[9] Aapo Hyvarinen, “Survey on Independent Component
Analysis”, Neural Computing Surveys,vol. 2, pp. 94-
128, 1999.
[10] Hyvarinen, Karhunen, and Oja, “Introduction,”
Chapter 1 in Independent Component Analysis,John
Wiley, pp. 1-12, 2001.
Figure 6. Bar chart showing percentage accuracy of two

watermarks.
An Approach to Protection at Peer-To-Peer

Networks
*
Dr.G.Srinivasa Rao, *Dr.G. Appa Rao, *S. Venkata Lakshmi, *D. Veerabhadra Rao,
**
B. Venkateswar Reddy,*P. Venkateswara Rao, *K. Sanjeevulu.
*
GITAM University, **CVSR Engineering College
giduturisrinivasarao@yahoo.co.in
systems (IDS). Considerable importance is also placed on

Abstract: Open networks are often insecure and the topology of the network being protected [43], as well as
provide an opportunity for viruses and DDOS activities to its fault tolerance to ensure that its operation will
to spread. To make such networks more resilient against continue even if a part of it is damaged.
these kind of threats, we propose the use of a peer-to-
A significant increase in the spread of viruses, worms
peer architecture whereby each peer is responsible for:
(a) detecting whether a virus or worm is uncontrollably
and Trojan horses over the Internet has been observed in the
propagating through the network resulting in an recent years. Recent evidence shows that older boot sector
epidemic; (b) automatically dispatching warnings and viruses, as well as viruses transmitted over floppy disks no
information to other peers of a security-focused group; longer constitute a considerable threat [12]. At the same
and (c) taking specific precautions for protecting their time, though, modern viruses have become more dangerous,
host by automatically hardening their security measures employing complex mutation, stealth and polymorphism
during the epidemic. This can lead to auto-adaptive techniques [37] to avoid detection by anti-virus software
secure operating systems that automatically change the and intrusion detection systems. These techniques are
trust level of the services they provide. We demonstrate particularly advanced and, combined with the fact that
our approach through a prototype application based on antivirus software is often not properly updated with the
the JXTA peer-to-peer infrastructure.
latest virus definitions, can lead to uncontrollable
situations.
Keywords: Peer-to-peer, Antivirus, Intrusion Detection,
JXTA
In the last two years it has been proven both theoretically
[38, 23] but mainly practically that the infection of
hundreds of thousands of computers within a matter of
1. Introduction hours -or even minutes is feasible. At the theoretical level
The rapid evolution of the Internet, coupled with
Staniford [38] presented scanning techniques (random
the reduction in the cost of hardware, have brought forth
scans, localized scans, hit-list scans, permutation scans)
very significant changes in the way personal computers are
which, used by a worm, can perform attacks of this order.
used. Nowadays, the penetration of the Internet is wide, at
Indeed such worms are often referred to as Warhol worms or
least in the developed world, and high percentage of
Flash worms due to their potential velocity of transmission.
connectivity is handled through broadband technologies
A similar confirmation was obtained practically in the
such as DSL, cable modems, satellite links and even 3G
cases of the worms Code Red [31], Code Red (CRv2) [5],
mobile networks. Many companies have permanent
Code Red II [13], Nimda [26, 27, 20], and Slammer [22],
connections to the Internet through leased lines and optical
which were characterized as epidemics by the scientific
fibers, and many home users through the aforementioned
community [44] (although a more appropriate
broadband connections. If one also takes into account the
epidemiological term would be pandemics). Recently the
significant development of wireless networking technologies
Blaster-worm [24, 21] caused significant disruption in the
(such as Wireless LAN, HyperLAN), the immediate result is
Internet, although the infection rate of the specific worm
an almost universal connection of most users on a 24-hour
was relatively slow in comparison with the previously
basis. Although the potential benefits arising from these
mentioned worms. The reason for the effectiveness of the
developments are various and important, so are the dangers
Blaster-worm was the exploitation of the Windows DCOM
that follow from the possibility of malicious abuse of this
RPC interface buffer overrun vulnerability. This
technology.
vulnerability affects all unpatched Windows NT /2000/ XP
The proliferation of viruses and worms, as well as the
systems, as opposed to Code Red worms variations or the
installation of Trojan horses on a large number of
Slammer worm which were focused on machines acting as
computers aiming at Denial of Service (DoS) attacks
Web Servers or SQL Servers respectively.
against large servers, constitute one of the major current
All of the above is evidence that rapid malcode is
security problems. This is due to the extent to which critical
extremely hard to confront using the “traditional” way of
infrastructures and operations such as hospitals, airports,
isolating and studying the code to extract the appropriate
power plants, aqueducts etc. are based on networked
signature and update the IDS in real time.
software-intensive systems. The measures taken for
We now propose to the reader to consider human
protection against such threats include [45] the use of
behavior during a flu epidemic. Obviously a visit to a doctor
firewalls, anti-virus software and intrusion detection
and the use of vaccines is essential, however there is also The Notifier is a daemon responsible for monitoring the
need for an increased awareness and use of hygiene rules: computer on which it runs and collecting any information
avoiding crowded spaces, increasing the ventilation of our relevant to probable security attacks. There is a plethora of
working area etc. Once the epidemic subsides, these different approaches to incorporate in the Notifier; for
measures can be suspended; a person showing symptoms of simplicity in our preliminary implementation we only
the disease, of course, should still visit a doctor to receive monitor the log files of several security related applications,
medical care, regardless of whether the epidemic is still such as firewalls, anti-virus software and IDS systems.
taking place. These are applications that collect information about
The classic computer protection methods can be likened security threats and attacks to the computer system on
to the above medical situation: The vaccination of the which they are running and either notify the user of these
population can be compared to updating the virus signature attacks or take specific measures, while at the same time
files; the lookout for symptoms may be compared to storing information relevant to the attacks into log files. By
detection by an IDS; while the hygiene rules followed, regularly reading the log files generated by these
which are essential for the protection of the larger, still applications, the Notifier detects any recently identified
unaffected population, may be compared to the operation of security attacks to the computer it is running on. At regular
our proposed system, described in the time intervals t, the Notifier of node n will record the
following sections. number of hits (hnt) the node received over the past interval.
It will then calculate and transmit the percentage pnt by
2. Architecture: which this average differs from the average hits in an
Peer-to-peer networks, which we will hereafter reference aggregate of the k latest intervals, given by
as p2p networks, are often considered to be security threats
for organizations, companies or plain users, mainly due to
the use of p2p-based applications for illegal file sharing,
and to the ability of worms to be spread through such
applications (e.g. VBS.GWV.A [41, 40] and W32.Gnuman
[10]). Our work indicates, however, that p2p networks can
also be positively utilized to significantly reinforce network
security, by offering substantial help in the protection
against malicious applications. We propose an effective way
to achieve this by collecting and exchanging information
that will allow us to obtain a global overview of the network
status, with reference to ongoing security attacks. The goal
of our methodology is to select the most appropriate security
policy, based on the level of danger posed by rapid malcode
circulating in the network.
P2p networks leverage the principle that a much
better utilization of resources (processing power, bandwidth,
storage etc.) is achieved if the client/server model is
replaced by a network of equivalent peers. Every node in
such a p2p network is able to both request and offer services
to other peer nodes, thus acting as a server and a client at
the same time (hence the term “servent” = SERVer +
cliENT which is sometimes used).
The motivation behind basing applications on p2p
architectures or infrastructures derives to a large extent
from their adaptability to variable operating environments,
i.e. their ability to function, scale and self-organize in the
presence of a highly transient population of nodes (or
computers/users), hardware failures and network
outages,without the need for a central administrative server.
Our proposed application, which we call
“NetBiotic”, requires the cooperation of several computers
within a common peer group, in which messages are
exchanged describing the attacks received by each
computer. It consists of two independent entities: a Notifier
and a Handler. These entities act as independent daemons Figure 1. The architecture of the NetBiotic system
for UNIX systems, services for Windows NT/2000/XP or within a group of cooperating peer
processes for Windows 9x/Me. From now on we will be computers.
referring to these entities as daemons for simplicity. Figure
1 illustrates the architecture of the proposed system within a
group of cooperating peer computers.
where: approach, the thresholds are selected empirically and

• t is the ordinal number of a fixed time interval. we have:
• n is a node identifier. • if pavg> τ high then increase security policy.
• hnt is the number of attacks node n received in the • if pavg< τlow then decrese security policy.
interval t. • if τlow≤ pavg≤ τ high, do nothing.
• pn t is the percentage increase or decrease in attacks
during the current interval t on node n. We base our decision for modifying the security policy
• k(>0) is the size of the “window” used, in number on the rate of change of attacks, rather than on the actual
of t time intervals, within which the average number of attacks, to normalize the inputs from all peers
attack rate is calculated. with respect to their regular susceptibility to attacks; a peer
whose actual number of attacks during a monitored time
Selecting the appropriate length of the time interval t is
interval has increased from 1000 to 1100 has only
currently a subject of further research. In our current
experienced a 10% change in the attack rate, while a peer
implementation we use a value of 15 minutes, which we
whose number of attacks increased from 50 to 150 within
feel provides a balance between increased network traffic
the same interval has experienced a 200% change in the
and delay in notifying the network of attacks. This will be
attack rate; still, they have both received 100 attacks more
further discussed in the next Section.
than usual. As far as the actual utilization of our
A value of pn t significantly greater than 1.0 is an
architecture for protecting the computer system is
indication that node n is under security attack during the
concerned, the countermeasures taken will depend on many
interval t. The actual threshold used for pnt is set by
factors. A simple personal computer will be requiring
experience, and can vary according to the tolerance for
different protection strategy than the central server of a
false positives/negatives one has. With a small threshold it
large company. The type of operating system is also an
is possible to falsely interpret slightly increased rapid
important factor. The proposed system is not suggested as a
malcode activity as an epidemic (false positive), leading to
replacement for traditional protection software (anti-
an unnecessary activation of the available countermeasures,
viruses, IDS, firewalls etc.). The aim of NetBiotic is to
which in turn can disrupt some non critical useful services
assemble an additional, overall picture of the network status
and cause inconvenience to the users. A very large
and suggest the basic security measures to be taken in the
threshold on the other hand, would probably fail to identify
event of an epidemic. The NetBiotic architecture might not
a rapid malcode epidemic (false negative) leaving the
be capable to protect against a specific attack, however it
system protected only by its standard built-in security
will engage the standard measures that in many cases are
mechanisms. We tend to believe that is much better to tune
crucial (such as disabling HTML previewing in several
the NetBiotic system towards a large threshold because
mail clients, not allowing Active X controls in various web
rapid malcode epidemics cause a number of side-effects
browsers, disabling macros in some office application etc.).
which are difficult to remain unnoticed. For us it is more
important to ensure the timely recognition of these
In our prototype design, the recommended measures for
symptoms, in order to increase the security level of the
a simple personal computer running Microsoft Windows
protected system before a circulating worm may manage to
would be to increase the security level of the default mail
launch an attack against it.
client and web browser. It would be additionally helpful to
The Handler is also a daemon, responsible for
alert the user of the increased threat, in order to minimize
receiving the messages sent from the Notifiers of other
threats of automated social engineering attacks. Servers can
computers, and for taking the appropriate measures when it
similarly disable non-critical networked services (e.g. by
is deemed necessary. More specifically, it records the hit
modifying the inetd.conf file in the case of Linux/Unix
rates ht and percentage changes pt received from the
based operating systems). Figure 2 illustrates the operation
different nodes in the peer group within a predefined period
and interaction of the Notifier and Handler daemons.
of
time t, and calculates the overall change in attack rate,
averaged for all n nodes of the peer group that transmitted a 3. Implementation
message during the last interval: The prototype system we present here was developed
using the JXTA protocol [15]. JXTA is a partially
centralized p2p protocol implementation introduced in early
2001, designed for maximum peer autonomy and
independence. It allows applications to be developed in any
The architecture supports countermeasures based language, it is independent of operating system type and is
upon predefined thresholds for pavg,, which are again set by not limited to the TCP/IP protocol for data transfer. This
experience. If pavg, exceeds an upper threshold, the security allows an application such as NetBiotic to be easily ported
level of the computer is raised. If, on the other hand, it to various operating systems, which is crucial to its
drops below a lower threshold for a large period of time, the operation, as its effectiveness will depend on the size of the
security level at which the computer functions is reduced. peer group that will adopt it. An additional benefit of JXTA
is its availability under an open source software license
Selecting the appropriate thresholds τhigh and τlow for agreement, similar to the Apache License [1].
increasing or decreasing the security levels is crucial. In our Due to the nature of our application, security issues are
of particular interest. Security provisions are usually System independence cannot be achieved in the case of the
incorporated in p2p architectures by means of various countermeasures taken, which will depend on the operating
cryptographic mechanisms such as the information system. Different scripts have to be used to modify the
dispersal algorithm [30] or Shamir’s secret sharing code security levels of applications in different operating
[33], anonymous cryptographic relays [32], distributed systems.
steganographic file systems [11], erasure coding [19],
SmartCards or various secure routing primitives [7].
JXTA peers function under a role-based trust model,
whereby individual peers function under the authority of
third-party peers to carry out specific tasks. Public key
encryption of the messages exchanged, which may be in
XML format, as well as the use of signed certificates are
supported, providing confidentiality to the system. The use
of message digests provides data integrity, while the use of
credentials — special to-kens that authenticate a peer’s
permission to send a message to a specific endpoint —
provide authentication and authorization. JXTA also
supports the use of secure pipes based on the TLS protocol.
Further work is being carried out based on the security
issues of the JXTA system, notably the implementation of a
p2p based web of trust in the Poblano Project [4], which
will be discussed in the future work Section.
Our system was implemented in Java (java2
version 1.4.0 02) using JXTA version 1.0, and uses the
winreg [36] tool to administer the windows registry and
modify the security settings of the various applications. The
main advantages of Java are its compatibility with most
operating systems as well as the fact that it is one of the
most secure programming languages.
In our preliminary implementation, the Handler
modifies the security settings of the Microsoft Outlook mail
client and the Microsoft Internet Explorer web
browser.These two applications were selected as they are
often the target of viruses. The simple operation of Figure 2. Operation of the Notifier and Handler daemons
increasing their security settings is therefore enough to
provide effective protection to a large number of users. Our system has been tested in laboratory
Most anti-virus programs can be adjusted to environment as well as in a peer group that was set up for
produce log files with the attacks they intercept. By this purpose, in which virus attacks were simulated on
regularly monitoring such log files, the Notifier daemon is some peers, resulting in the modification of the security
able to detect a security attack and notify the peers. To test settings of Microsoft Outlook and Internet Explorer on
our prototype system, we created a software tool which other peer computers. No real viruses were deployed. A
randomly appends supposed security attack entries to these program was running on each of the peer computers and
log files. periodically edited the log file of the antivirus software,
The NetBiotic architecture is compatible with any IDS simply changing its size to simulate a security attack event.
or anti-virus software that can be setup to record the The average frequency with which these events were
security attacks against the system it is protecting in a log simulated was random and different for each computer. The
file. Our aim is to make the NetBiotic system as exchange of messages, individual and overall average hit
independent as possible from the IDS with which it rates as well as the resulting changes in the security settings
cooperates and the underlying operating system. This of the application were recorded and verified against our
independence, however, theoretical expectations.
cannot be total, as the following factors will be unavoidably Finally, since our system consists of two independent
system dependent: daemons, it is possible to only install one of the two on
certain peer computers. For instance, the Notifier daemon
¯ Log files would be particularly useful running on a large company
In its simplest form, the system can simply check the size of server, and supplying the peers with information about the
the log file. For a more sophisticated operation, though, it security threats it faces. The administrators of such a server
would be necessary to incorporate a parser that would may prefer not to install the Handler daemon, and instead
extract specific information from the log files. Such a parser manually take action in the event of security attacks.
has to be specific to each different type of log file used. Similarly, for a personal computer user who may not have
adequate security measures and antivirus software installed
¯ Countermeasures taken (for either financial or other reasons), installing the Handler
daemon itself may provide an adequate level of protection. Emerald, like Hummingbird, can be invaluable in
In this case, the Handler daemon would modify the local protecting a computer system or network against
security level based on information received by the security distributed and targeted attacks. NetBiotic may not be in
focused peer group. The Handler would therefore operate the position to affront such attacks with the same
relying on the trustworthiness of the information received effectiveness, as its goal is the seamless and automated
from the peer group only, which may in some cases be a creation of a network of peers for the fast exchange of
disadvantage. information regarding rapid spread malcode activity,
leveraging the benefits of peer-to-peer architectures and
4. Related Work topologies, and providing basic protection to the
The research that is most relevant to our proposed participating peers.
system has been carried out within the framework of project Bakos and Bert [2] presented a system for the
Indra [14], with which we partially share a common detection of virus outbreaks. The fastest spreading worms
philosophy. We agree on the basic principle of using p2p use scanning techniques for identifying potential target
technology to share security attack information between computers. As a result, they also scan a large number of
computers in a network in order to activate security addresses that do not correspond to actual computers. The
countermeasures if necessary. routers that intercept such scanning messages usually reply
We differ however in the circumstances under which with a ICMP Destination Unreachable (also known as
specific countermeasures should be taken. According to the ICMP Type 3 or ICMP-T3) message. The authors propose
Indra project team, in the event that a security attack is that a carbon copy message be sent by the routers to a
detected countermeasures should be immediately initiated, central collector system, which will be responsible for
by using the appropriate plugins to protect the computer collecting, correlating and analyzing this data. Bakos and
system. A single security attack anywhere in the network is Bert have implemented such a system by modifying the
enough for them to generate a response. In short, Indra is kernel of the Linux operating system to act as a router. The
designed to respond to every single security attack. central collector receives the messages and forwards them
In contrast, our system’s goal is to determine if there to an analyzer system, which extracts the valuable
is a general increase in the virus or worm attacks in the information. It should however be examined whether the
network, or more importantly a virus or worm epidemic time required for the entire processing prohibits the use of
outbreak. Measures taken in this case, such as the increase this system for fast spreading worms, as described by
in security settings of mail clients, web browsers and anti- Staniford [38].
virus programs will only be effective during the epidemic, Systems that use an extended network to gather
and the system will return to its original state after it is information yet rely on a centralized client/server model
finished. In our design, individual virus or worm attacks in were also examined. DeepSight [6] is a system developed
the network are not considered separately. Furthermore, we by Symantec based on a client/server architecture, whereby
believe that our design can be expanded to very large centralized servers collect and re-distribute security attack
network sizes without considerably increasing the overall information. Since it is a commercial system it is not
network traffic. available for scientific research, however it does include a
A number of highly distributed systems rely on peer very widespread data collection network.
communications. The Hummingbird system [28] is based An approach similar to DeepSight is taken by
on a cooperative intrusion detection framework that relies DShield, in which hundreds of computers communicate
on the exchange of security related information between with central servers and transmit their IDS log files. The
networks or systems in the absence of central servers process the data and announce in a web site
administration. The structure of the Hummingbird system information about the currently active malware, the IP
is significantly more complex and advanced than addresses from which most attacks originated and other
NetBiotic, using a combination of Manager-Hosts, useful information. Through the incorporation of different
Managed Hosts, Slave Hosts as well as Peer, Friend and parsers, DShield supports various different IDS systems.
Symbiote relationships for the exchange of security related DShield has been active for more than two years, with a
information. The Hummingbird system includes advanced significant number of users. A disadvantage of the system is
visualization tools for its configuration and monitoring of that the large volume of data collected requires considerable
log files, and although it may require considerable effort processing time for extracting useful information. The
and expert knowledge for fine tuning the cooperation of theoretical times taken by the Flash and Warhol worms as
each host with the others, it is particularly effective for well as the measured times for the Slammer worm [22, 38]
distributed security attacks (such as doorknob, chaining, to spread through the Internet are probably beyond the
loopback attacks etc.). A potential secondary use of the ability of DShield to react.
Hummingbird system, in our view, could also be in the Both DeepSight and DShield aim at providing a
detection of malcode. global view of the Internet security status, however they are
Emerald [29, 25] is a system targeted towards the both subject to the disadvantages of the client/server
exchange of security incident related information between architecture they follow: their dependence on a single
different domains or large networks. It consists of a layered server for their operation and their lack of adaptability
architecture that provides a certain abstraction, and makes them vulnerable to targeted attacks. An original
requires the adjustment of parameters relevant to the trust approach taken by the AAFID [35], whereby agents are
relationships between cooperating parties. We believe that used to collect virus attack information also follows a
centralized control structure. The same holds for the GrIDS purposes.
system [39], which uses activity graphs to control large 3. Other research groups are involved in the creation of a
scale networks and identify suspicious activities, based on p2p-based web of trust. We intend to study these systems to
the judgment of a System Security Officer. examine to what extent they can be used to enhance the
Finally, the following two approaches propose NetBiotic architecture.
different ways of monitoring the overall security state and ¯ Use of epidemiological models
threat level of a network: In the DIDS system [34], the We believe that the incorporation of mathematical
overall security state of a network under observation is epidemiological models for the detection of epidemic
represented by a numerical value ranging between 0 (safest) outbreaks in the network and determining the threshold for
and 100 (least safe), while a clearly visual approach to initiating security level modifications should significantly
representing the network security state has been proposed enhance the robustness of our system. A key point in our
[42, 8]. We find both approaches very descriptive and future research will be the selection of the thresholds for
useful to a System Security Officer. In our prototype modifying security policies. These thresholds will be
NetBiotic implementation, however, we are currently variable and will depend on each system’s characteristics
adopting a much simpler approach which consists of and on an analysis of the attack data collected. Studies [9,
choosing between three different security states (regular, 18, 16, 17] show that there is a correlation between the
low risk and high risk), as described in Section 2. patterns of spread of biological viruses and computer
viruses. These studies were mainly limited to closed local
5. Future Work area networks. P2p models are ideal for gathering large
The NetBiotic system is an evolving research scale network virus information, which can subsequently be
prototype. It is currently being extended in a number of processed and adapted to epidemiological models, leading
ways as discussed below, in order to subsequently be to decision tools for concluding, or perhaps even predicting,
released as open source software to allow the collaboration whether there is — or is likely to be — an epidemic
with other research groups working in similar directions. outbreak in the network.
At this stage, our goal is to propose an architecture, ¯ Choice of appropriate security policy
accompanied by a basic implementation for proof-of- In conjunction with other factors, such as the role of the
concept purposes, which, based on a p2p network system being protected, our system should be able to
infrastructure can provide security services for computer effectively choose the most appropriate security policy for
systems. Although our prototype performed well in the the specific period of time. In this way, single incidents of
situation in which we tested it, it is not suitable for virus attacks
performing large-scale testing. may not be the cause of any concern, yet the detection of
We expect that, before more advanced versions of epidemic outbreaks would initiate a modification of the
our application will be implemented, the scientific security policies.
community will examine the use p2p networks in security ¯ Platform porting
applications from a theoretical standpoint and provide In porting our system to Unix/Linux platforms, the
insight into the advantages and disadvantages of such an operating system could be instructed to launch or halt
approach. applications, or automatically request updates. The
The following conceptual and implementation configuration of these operating systems can be edited
improvements are currently being considered: through plain text files, which is an additional benefit for
our system.
¯ Vulnerability to malicious attacks
A major drawback of our current design is its 6. Conclusions
inability to effectively verify theinformation transmitted in Even the best protected organizations, companies
the network. If one or more malicious users manage to or personal users are finding it difficult to effectively shield
introduce in the peer network a large number of false hit themselves against all malicious security attacks due the
rate indications, the result may be the unwanted decrease of increasing rate with which they appear and spread.
the security measures of the computers in the network, Antivirus applications, as well as IDS systems, identify the
rending them vulnerable to virus attacks. unknown malware by employing behavioral based heuristic
We propose that all members of the security peer group will algorithms. These algorithms are particularly effective
have to be authenticated and verified, probably through the under a strict security policy, however they tend to produce
use of certificates, to enforce a consistent authentication and an increased number of false alarms, often disrupting and
authorization policy. upsetting the smooth operation of a computer system and
At the implementation level, to confront the the organization or users it supports. On the other hand, if
problem of malicious users introducing false information we the security policy is relaxed, the threat of a virus infection
further propose the following approaches, based on the becomes imminent.
capabilities offered by JXTA: We propose a platform based on p2p technology in
which the computers participating as peers of a network
1. JXTA supports the exchange of encrypted messages automatically notify each other of security threats they
based on the TLS algorithm secured pipes [3], which will receive. Based on the rate of the warning messages
be used for the transmission of warning messages. received, our system will increase or decrease the security
2. JXTA message digest will be used for data integrity measures taken by the vulnerable applications running on
the computer. Our approach automates elements of the IEEE WET ICE Workshop on Enterprize Security,
process of choosing the appropriate security policy, based Linz, Austria, June 2003.
on data useful for adjusting the security levels of both the [15] Project jxta v2.0 java programmer’s guide: Current
operating system (by launching and terminating related on-line (June 2003):
applications) and the security applications (by modifying http://www.jxta.org/docs/jxtaprogguide v2.pdf.
the security parameters of the heuristic algorithms they [16] J Kephart. How topology affects population dynamics.
employ). In Proceedings of Artificial Life 3, Santa Fe, New
An important aspect of our design is that the traffic Mexico, June 1992.
introduced in the network by the peer nodes as a result of [17] J. Kephart, D. Chess, and S. White. Computers and
the transmission of hit rate information is minimal. We epidemiology. IEEE Spectrum, May 1993.
believe that, with the inclusion of the future extensions we [18] J. Kephart and S. White. Directed-graph
are currently working on, our approach may lead to epidemiological models of computer viruses. In
operating systems, antivirus programs, IDS software and Proceedgings of IEEE Computer Society Symposium
applications that will be able to self-adjust their security on Research in Security and Privacy, pages 343–
policies. 361, Oakland, CA, 1991.
[19] J. Kubiatowicz, D. Bindel, Y. Chen, P. Eaton, D.
References: Geels, S.R. Gummadi, H. Weatherspoon, W.
[1] Apache license: Current on-line (June 2003): Weimer, C. Wells, and B. Zhao. Oceanstore: An
http://httpd.apache.org/docs/license. architecture for global-scale persistent storage. In
[2] G. Bakos and V. Berk. Early detection of internet Proceedings of ACM ASPLOS. ACM, November
worm activity by metering icmp destination 2000.
unreachable messages. In Proceedings of the the [20] A. Mackie, J. Roculan, R. Russell, and VanVelzen M.
SPIE Aerosense, 2002. Nimda worm analysis - incident analysis report
[3] Wilson B.J. JXTA. New Riders, Indianapolis, IN, USA, version ii. September 2001.
June 2002. [21] J. Miller, J. Gough, B. Konstanecki, J. Talbot, and J.
[4] R. Chen and W. Yeager. Poblano: A distributed trust Roculan. Deepsight threat management system
model for peer-to-peer networks. Technical report, threat alert - microsoft DCOM RPC worm alert.
Sun Microsystems. Current on-line (August 2003):
[5] Code Red CRv2. Current on-line (June 2003): https://tms.symantec.com/members/analystreports/030811-
http://www.caida.org/analysis/security/code- alert-dcomworm.pdf.
red/coderedv2 analysis.xml. [22] D. Moore, V. Paxson, S. Savage, C. Shannon, S.
[6] Deepsight threat management system: Current on-line Staniford, and N. Weaver. The spread of the
(June 2003): http://www.securityfocus.org. sapphire/slammer worm. Current on-line (June
[7] P. Druschel and A. Rowstron. Past: A large-scale, 2003):
persistent peer-to-peer storage utility. In Proceedings http://www.caida.org/outreach/papers/2003/sapphire/
of the Eighth Workshop on Hot Topics in Operating sapphire.html. Technical
Systems, May 2001. report, 2003.
[8] R. Erbacher, K. Walker, and D. Frincke. Intrusion and [23] D. Moore, G. Voelker, and S. Savage. Internet
misuse detection in large scale systems. IEEE quarantine:requirements for containing self-
Computer Graphics and Applications, 22(1), 2002. propagating code. In Proceedings of the 2003 IEEE
[9] S. Forrest, S. Hofmeyr, and A. Somayaji. Computer Infocom Conference, San Francisco California,
immunology.Communications of the ACM, USA, April 2003.
40(10):88–96, 1997. [24] Microsoft security bulletin ms03-026. Current on-line
[10] W32.gnuman.worm: Current on-line (June 2003): (August 2003):
http://service1.symantec.com/sarc/sarc.nsf/html/w32. http://www.microsoft.com/technet/treeview/default.as
gnuman.worm.html. p? url=/technet/security/bulletin/ms03-026.asp.
[11] S. Hand and T. Roscoe. Mnemosyne: Peer-to-peer [25] P. Neumann and P. Porras. Experience with
steganographic storage. In Proceedings of the 1st EMERALD to date. In First USENIX Workshop on
International Workshop on Peer-to-Peer Systems Intrusion Detection and Network Monitoring, pages
(IPTPS ’02), MIT Faculty Club, Cambridge, MA, 73–80, Santa Clara, California, April 1999.
USA, March 2002. [26] Current on-line (June 2003):
[12] Icsa labs 2002 computer virus prevalence survey. http://www.incidents.org/react/nimda.pdf.
Current on-line (June 2003): [27] Current on-line (June 2003): http://www.f-
http://www.trusecure.com/download/dispatch/vps200 secure.com/v-descs/nimda.shtml.
2.pdf. [28] Polla, D., J. McConnell, T. Johnson, J. Marconi, D.
[13] Code Red II. Current on-line (June 2003): Tobin, and D. Frincke. A framework for cooperative
http://www.eeye.com/html/research/advisories/al200 intrusion detection. In Proceedings of the 21st
10804.html. National Information Systems Security Conference,
[14] R. Janakiraman, M. Waldvogel, and Q. Zhang. Indra: pages 361–373, October 1998.
A peer-to-peer approach to network intrusion [29] P. Porras and P. Neumann. EMERALD: Event
detection and prevention. In Proceedgings of 2003 monitoring enabling responses to anomalous live
disturbances. In Proceedings of the National

Information Systems Security Conference, October Authors Profile
1997.
[30] M.O. Rabin. Efficient dispersal of information for Dr. G.Srinivasa Rao, M.Tech, Ph.D,
security, load balancing and fault tolerance. Journal Sr.Asst.Professor. four years industrial
of the ACM, 36(2):335–348, April 1989. experience and over 8 Years of teaching
[31] Code Red. Current on-line (June 2003): experience with GITAM University, handled
http://www.eeye.com/html/research/advisories/al200 courses for B.Tech, M.Tech. Research areas
include Computer Networks And Data
10717.html.
Communications. published 6 papers in
[32] A. Serjantov. Anonymizing censorship resistant various National and International
systems. In Proceedings of the 1st International Conferences and Journals.
Workshop on Peer-to-Peer Systems (IPTPS ’02), MIT
Faculty Club, Cambridge, MA, USA, March 2002. Dr. G.Appa Rao., M.Tech., M.B.A.,Ph.D.,
[33] A. Shamir. How to share a secret. Communications in computer science and Engineering form
of the ACM, 22:612–613, November 1979. Andhra Universiy. Over 12 Years of
[34] S. Snapp, J. Brentano, G. Dias, T. Goan, T. teaching experience with GITAM
Heberlein, C. Ho, K. Levitt, B. Mukherjee, S. Smaha, University, handled courses for B.Tech,
M.Tech. Research areas include Data
T. Grance, D. Teal, and D. Mansur. DIDS
Mining and AI. Published 8 papers in
(distributed intrusion detection system) - motivation, various National and International
architecture, and an early prototype. In Conferences and Journals.
Proceedings of the 14th National Computer Security
Conference, pages 167–176, Washington, DC, 1991. Mrs. S. Venkata Lakshmi M.Tech in
[35] E. Spafford and D. Zamboni. Intrusion detection Information Technology from Andhra
using autonomous agents. Computer Networks, University. Asst.Prof in GITAM University.
(34):547–570, October 2000. Over 2 years of teaching experience with
[36] D. Spinellis. Outwit: Unix tool-based programming GITAM University and Andhra University
meets the windows world. In Proceedings of the handled courses for B.Tech, and M.C.A. and
USENIX 2000 Technical Conference, pages 149– 2 years of industry experience as a software
engineer. Published 2 papers in various
158, San Diego, CA, USA, June 2000.
International Conferences and Journals.
[37] D. Spinellis. Reliable identification of bounded-length
viruses is np-complete. IEEE Transactions on Mr..D.Veerabhadra Rao., M.Tech., in
Information Theory, 49(1):280–284, January 2003. Information Technology. Over 7 Years of
[38] S. Staniford, V. Paxson, and N. Weaver. How to own teaching experience with GITAM University,
the internet in your spare time. In Proceedings of the handled courses for B.Tech and M.Tech .One
11th USENIX Security Symposium, 2002. research paper was published in international
[39] S. Staniford-Chen, S. Cheung, R. Crawford, M. journal and one conference.
Dilger, J. Frank, J. Hoagland, K. Levitt, C. Wee, R.
Yip, and D. Zerkle. GrIDS – A graph-based
intrusion detection system for large networks. In B. Venkateswar Reddy M.Sc (Maths) from
Osmania University and M.Tech(CS) from
Proceedings of the 19th National Information
Satyabhama University, having one year
Systems Security Conference, 1996. teaching experience, handled courses for
[40] VBS.Gnutella. Current on-line (June 2003): B.Tech and M.Tech in CVSR Engineering
http://service1.symantec.com/sarc/sarc.nsf/html/vbs. college, Hyderabad.
gnutella.html.
[41] VBS.Gnutella. Current on-line (June 2003):
http://vil.nai.com/vil/content/v 98666.html.
[42] G. Vert, J. McConnell, and D. Frincke. A visual
mathematical model for intrusion detection. In P. Venkateswara Rao M.Sc(Physics) from
Proceedings of the 21st National Information Acharya Nagarjuna University, Pursuing
M.Tech (CST) from GITAM University
Systems Security Conference, pages 329–337,
October 1998.
[43] C. Wang, J.C. Knight, and M.C. Elder. On computer
viral infection and the effect of immunization. In
Annual Computer Security Applications Conference
(AC-SAC), pages 246–256, December 2000. K. Sanjeevulu M.Sc(Maths) from Osmania
[44] V. Yegneswaren, P. Barford, and J. Ullrich. Internet University., Pursuing M.Tech (CST) from
intrusions: Global characteristics and prevalence. In GITAM University.
Proceedings of ACM SIGMETRICS, June 2003.
[45] R.L. Ziegler. Linux Firewalls. New Riders
Publishing, Indianapolis IN, USA., 2002.
New Power Aware Energy Adaptive protocol with

Hierarchical Clustering for WSN
Mehdi Golsorkhtabar1 , Mehdi Hosinzadeh2, Mir Javad Heydari1, Saeed Rasouli1
1
Islamic Azad University, Tabriz Branch, Department of Computer Engineering,
Tabriz, Iran
{m.golsorkhtabar;m.heydari;s.rasouli}@iaut.ac.ir
2
Science and Research Branch, Islamic Azad University,
Tehran, Iran
hosseinzadeh@srbiau.ac.ir
Abstract: In recent years, progress in wireless communication

has made possible the development of low cost wireless sensor 2. Related work
networks. Clustering sensor nodes is an effective topology In recent years, many protocol and algorithm for clustering
control approach. In this paper, we propose a new routing
algorithm which can increase the network lifetime. We assume
have been proposed. Usually there are many differences
that each node can estimate its residual energy and then a new between CH selection and clustering organism. But
clustering method will be proposed for increase of network generally some of these clustering schemes applied in
lifetime. This assumption is similar to many other proposed homogeneous networks and some others clustering
routing algorithms for WSN and is a possible assumption. In the algorithms applied in heterogeneous networks. Therefore,
new algorithm, the predefined numbers of nodes which have the most of the current popular clustering algorithms are not
maximum residual energy are selected as cluster-heads based on
fault tolerant, such as LEACH [3], PEGASIS [4] and HEED
special threshold value first and then the members of each
cluster are determined based on the distances between the node [5].
and the cluster head and also between the cluster head and base LEACH is the most popular clustering algorithm. Many of
station. At last, the simulation results show that our method CH selection algorithms are based on LEACH’s
achieves longer lifespan and reduce energy consumption in architecture. [6] is proposed to elect the CHs according to
wireless sensor networks. the energy remaining in each node. We call this clustering
Keywords: wireless sensor networks; clustering algorithm; protocol LEACH-E. In the rest of this section, we review
energy adaptive; network lifetime LEACH algorithm and discuss its limitations, because
LEACH is very popular in wireless sensor network
1. Introduction clustering protocols.
Wireless Sensor Network (WSN) comprises of micro
2.1 LEACH (Low Energy Adaptive Clustering
sensor nodes, which are usually battery operated sensing
Hierarchy)
devices with limited energy resources. In most cases,
In LEACH protocol, energy efficiency are achieved by
replacing the batteries is not an option. [1-2].
being CH in turn, and then distributing impartially the total
Wireless sensor network, usually, are heterogeneous. The
networks energy to unique node, thus lowing energy
protocols should be design for the typical of heterogeneous
consuming and increasing network lifespan. CH election
wireless sensor networks. Most of the clustering algorithms
depends on the whole numbers of CH in networks and times
are designed for homogeneous wireless sensor networks and
that nodes have been CH until now. Principles scenarios for
they are not optimized when network's nodes is in
this protocol are:
heterogeneous state such as [3-4].
• The base station fixed in nowhere near of the sensor
In this paper, we propose and evaluate PEAP, (new Power
nodes.
aware Energy adaptive Protocol with hierarchical clustering
• All the nodes in the wireless sensor network have the
for WSN). In considered wireless sensor network, nodes
same initial battery power and are homogeneous in all other
send sensing information to a cluster-head and then the CH
ways.
transmit data to base station. The certain clustering
In first phase, algorithm chooses a node stochastically,
algorithms with special method periodically electing cluster-
the principal will be explained in the following: all sensor
heads then cluster-heads aggregate the data of their cluster
nodes compute a value T(n ) according to the following
nodes and send it to the base station. We assume that all the
formula at the beginning of all rounds:
nodes of the network are spread heterogeneous, at first all
 p
nodes battery power is equal, all sensor nodes have limited
 n∈G
energy and the base station is fixed and not located between T ( n) = 1− p(r mod(1 p)) (1 )
sensor nodes and most of them are static and only a few are 0
 others
mobile.
Where in this equation P describes desired percentage of
CHs (e.g. P=0.05) current round, and G is the set of nodes Where in this equation P = the desired percentage of CHs
that have not been CH in the last 1/P rounds, r is the (e.g. P=0.05) the current round, and E n_current is the current
number of the current round. energy and En_max the initial energy of the node, with r s as
For each node, a random number between 0 and 1 the number of consecutive rounds in which a node has not
is generated. If this random number is less than T(n ) , this been CH. Thus, the chance of node n to become cluster head
sensor node will become a cluster head in this round and increases because of a higher threshold. A possible blockade
broadcast an advertisement message to other sensor of the network is solved. Additionally, r s is reset to 0 when a
node becomes CH. Thus, we ensure that data is transmitted
nodes near it.
to the base station as long as nodes are alive [6].
When each node has elected as cluster head itself for the
current round broadcasts an advertisement message to the Our clustering model is based on confidence value
rest of the nodes in the network. All the non-cluster head associated with broadcast from CHs. Confidence value of a
nodes, after receiving this advertisement message, decide on CH is a function of some parameters (1) distance between
the cluster to which they will belong for this round. This the CH, the node and (2) the CH current battery power and
decision is based on the received signal strength of the (3) number of nodes already were a member of this CH.
advertisement messages. After cluster head receives all the Basically, our model checks first if, with the current battery
messages from the nodes that would like to be included in power the CH has, it would be able to support the current
members at maximum data broadcast rate. A node decides
the cluster and based on the number of nodes in the cluster,
to join a CH if the head can still support the node with its
the cluster head creates a TDMA schedule and assigns each
rest power. Confidence value given by:
node a time slot when it can transmit [3].
Despite many advantages in using of the LEACH protocol
for cluster organization, CH selection and incising network Cv(i) = Bp
(3)
lifetime, there are a few features that the protocol does not Cm* Dc
support. LEACH assumes nodes power energies Where in this equation BP is the battery power of given
homogeneously. In a real, wireless sensor networks node, Cm is number of nodes already a member of given CH,
scenario, sensor nodes energy spread in heterogeneous Dc is distance between the CH and the node.
manner. Like LEACH, in order to reduce the probability of
collision among joint-REQ messages during the setup phase,
3. The New Protocol CSMA (Carrier Sense Multiple Access) is utilized as the
In this section, the details of PEAP are introduced. The MAC layer protocol. When a cluster head has data to send,
major application of a wireless sensor network is to it must sense the channel to see if anyone else is
monitoring of a remote environment. Data of individual transmitting using the BS spreading code. If so, the cluster
nodes are usually not very important. Since the data of head waits to transmit the data. Otherwise, the cluster head
sensor nodes are correlated with their neighbor nodes, data sends the data using the BS spreading code [3].
aggregation can increase reliability of the measured
parameter and decrease the amount of traffic to the base 4. Simulation Results
station. PEAP uses this observation to increase the efficiency
of the network. In order to develop the PEAP, some In order to evaluate the performance of the PEAP protocol,
assumptions are made about sensor nodes and the the simulator, specific to the needs of our model, was coded
underlying network model. For sensor nodes, it is assumed in PHP with Apache HTTP server version 2.2 and uses
that all nodes are able to transmit with enough power to PHP/SWF Charts for its graphical needs.
reach the BS if needed, that the nodes can adjust the amount We assume a simple model for the radio hardware energy
of transmit power, and each node can support different dissipation where the transmitter dissipates energy to run
Medium Access Control (MAC) protocols and perform the radio electronics and the power amplifier, and the
signal processing functions. These assumptions are receiver dissipates energy to run the radio electronics, as
reasonable due to the technological advances in radio shown in Fig. 1. For the experiments described here, both
hardware and low-power computing [3]. For the network, it the free space (d2 power loss) and the multi path fading (d4
is assumed that nodes have always data to send to the end power loss) channel models were used, depending on the
user and the nodes located close to each other have distance between the transmitter and receiver [7]. Power
correlated data. control can be used to invert this loss by appropriately
Such as LEACH, in first phase, PEAP chooses a node setting the power amplifier. If the distance is less than a
stochastically, , the principal will be explained in the threshold do, the free space (fs) model is used; otherwise, the
following: all sensor nodes compute a value T(n ) multi path (mp) model is used. Thus, to transmit l-bit
according to the following formula at the beginning of all message a distance, the radio expends
rounds.
lEelec +lεfs d 2 , d <d o
E  E n _ current  ETx(l,d) = ETx- elec(l ) + ETx- amp(l, d ) =  (3 )
 n _ current  lEelec +lεmpd ,
4
d ≥ do
T(n) = p* * 1 −  (2)
 E  E n _ max 
 n _ max  
And to receive this message, the radio expends:
Fig. 3 presents the number of nodes dead when using of

ERx(l) = ERx - elec (l ) = lEelec (4) clustering protocols. This result is closely related to the
network lifetime of the wireless sensor networks.
The electronics energy, Eelec, depends on factors such as

the digital coding, modulation, filtering, and spreading of
the signal, whereas the amplifier energy, εfs d2 or εmp d4 ,
depends on the distance to the receiver and the acceptable
bit-error rate.
We consider a wireless sensor network with N=100 nodes
randomly distributed in a 300m * 300m field. We assumed
the base station to be fixed and located at the origin (0, 0) of
the coordinate system. The radio parameters used in our
simulations are shown in Table (1).
Figure 3.Total nodes transmitting (N= 100).
Figure 1.Total nodes transmitting (N= 100).

5. Conclusion
In wireless sensor networks, the energy consumption and
the network lifetime are important issues for the research of
Table 1: Parameters used in simulations
the route protocol. This paper introduces PEAP, a new
Parameter Value power aware energy adaptive protocol with hierarchical
E elec 50 nJ/bit clustering for wireless sensor networks that distributes loads
among more powerful nodes. Compared to the existing
ε£s 10 pJ/ bit/m2
clustering protocols, PEAP has better performance in CH
εmp 0.0013 pJ/bit/m4 election and forms adaptive power efficient and adaptive
E0 3J clustering hierarchy. The simulation results presented that
PEAP significantly improves the lifespan and the energy
EDA 5 nJ/bit/message consumption of the wireless sensor networks compared with
d0 70 m existing clustering protocols. Further directions of this study
will be deal with clustered sensor networks with more than
Message size 8192 bits
three parameters with in threshold calculating and more
parameters to confidence value calculating.
Fig. 2 present the energy consumption of the clustering
protocols when the amount of nodes spread in network is
100. The x-axis indicates the number of rounds while y-axis
References
indicates the mean residual energy of each node. The results [1] C. Buratti, A. Conti, D. Dardari,R. Verdone, “An
demonstrate that the energy consumption of our algorithm is Overview on Wireless Sensor Networks Technology
and Evolution” , Sensors 2009, 9, 6869-6896;
generally smaller than LEACH and LEACH-E. doi:10.3390/s90906869.
[2] I. F. Akyildiz, W. Su, and Y. Sankarasubramaniam, “A
survey on sensor networks”, IEEE Communications
Magazine, 2002, 40(8), pp.102-114.
[3] W.R. Heinzelman, A.P. Chandrakasan, H.
Balakrishnan, “An application-specific protocol
architecture for wireless microsensor networks”,
IEEE Transactions on Wireless Communications 1
(4) (2002) 660–670.
[4] S. Lindsey, C.S. Raghavenda, “PEGASIS: power
efficient gathering in sensor information systems”,
Proceeding of the IEEE Aerospace Conference, Big
Sky, Montana, March 2002.
[5] O. Younis, S. Fahmy, “HEED: A hybrid, energy-
efficient, distributed clustering approach for ad hoc
sensor networks”, IEEE Transactions on Mobile
Computing 3 (4) (2004) 660–669.
[6] M.J. Handy, M. Haase, D. Timmermann,“Low energy
clustering hierarchy with deterministic cluster head
selection”, Proceedings of IEEE MWCN, 2002.
[7] T. Rappaport, Wireless Communications: Principles &
Figure 2.Total network energy (N= 100) Practice. Englewood Cliffs, NJ: Prentice-Hall, 1996.
Hexagonal Coverage by Mobile Sensor Nodes

G.N. Purohit1, Seema Verma2 and Megha Sharma3
1
Department of Mathematics, AIM & ACT, Banasthali University,
Banasthali-304022
gn_purohitjaipur@yahoo.co.in
2
Department of Electronics, AIM & ACT, Banasthali University,
Banasthali- 304022
seemaverma3@yahoo.com
3
Department of Computer Science, AIM & ACT, Banasthali University,
Banasthali-304022
edify44@yahoo.com
Abstract: Before the advent of mobile sensor nodes, static of the hexagon. They move in a systematic manner on
nodes have been used to provide coverage, which focuses on rectangular and triangular parts of hexagon. Previous work
repositioning of sensors to achieve coverage. But mobile sensor on Mobile traversal has been done using triangulation based
nodes provide a dynamic approach to cover age. Targets that
coverage [3], but the hexagonal approach proves to be more
might never be detected in a stationery sensor network can be
detected by moving sensors. Mobile sensors can compensate for efficient as the total distance traveled and time taken is
lack of sensors and improve network coverage. Here, we focus comparatively less. Deploying a good topology is also
on coverage of a rectangular region which is divided into beneficial to management and energy saving, and the
regular hexagons. The region is covered with mobile sensor hexagonal topology provides 2-coverage, as we wish to
nodes, where a group of four MSNs position themselves on four ensure optimal and energy efficient coverage. A
vertices of a hexagon. We can employ N≥4 MSNs for this deterministic energy-efficient protocol for sensor networks is
purpose, although basically only 4 MSNs are needed but extras
used in [1] that focuses on energy efficient coverage of ROI.
are employed in case of failure of any MSN.
Energy efficient distributed algorithms for sensor target
Key Words: coverage, mobile sensor nodes, energy coverage based on properties of an optimal schedule is
efficiency, hexagonal coverage. included in [2]. Power efficient organization of wireless
sensor networks is done in [4]. A coverage-preserving node
1. Introduction scheduling scheme for large wireless sensor networks is
discussed in [5].
The coverage problem is a fundamental issue in WSN,
which mainly concerns with a fundamental question: How The proposed objectives of our approach are:
well a sensor field is observed by the deployed sensors? To
optimize network coverage, the traditional approach (i) is to Covering the sensing area by minimum
deploy a large number of stationary sensor nodes and then to number of sensors, N≥4, as well as providing highly reliable
schedule their sensing activities in an efficient way [6]. and long
Recently, mobile sensor nodes have received much attention system lifetime, which is the main design challenge in
since network performance can be greatly improved by sensor networks.
using just a few of mobile nodes. Mobile sensor nodes have (ii) Upon a failure, the remaining MSN’s (N-4) efficiently
the movement capability to collaboratively reinstall the complete the coverage of the targeted area, otherwise they
network coverage. They are extremely valuable in situations remain in sleeping mode.
where traditional deployment mechanisms fail or are not
suitable, for example, a hostile environment where sensors We assume the following:
cannot be manually deployed or air-dropped. It is well (i) The sensing range of a sensor x, is a disc of radius
known that mobility increases the capacity of networks r centered at x and defined by
(MANETs) by reducing the number of relays for routing, Sx (r) = {a € R2 : | x-a| ≤ r}
prolonging the lifespan of wireless sensor networks (WSNs) where |x-a| stands for the Euclidian distance
and ensuring network connectivity in delay-tolerant between x and a.
networks (DTNs), using mobile nodes to connect different (ii)A location in region A is said to be covered by sensor x if
parts of a disconnected network. In this paper we present it is within x’s sensing range. A location in A is said to be
Mobile Traversal Algorithm (MTA) where the region of covered if it is within at least K sensor’s sensing
interest [ROI], considered as a rectangular area, is covered range.
by regular hexagons. The MSNs are placed at four vertices
(iii) All sensors are considered to have identical

sensing range.
This paper is organized as follows. The problem is

formulated in Section 2. In Section 3 the MTA is presented,
and the total traveling distance for various sides ‘a’ of
MSNs is determined. In Section 4 we conclude the paper.
Figure 2. K-Coverage of a point
2. Problem Formulation 3. MTA (Mobile Traversal Algorithm)

We consider a rectangular field of length ‘L’ and breadth The side of regular hexagons is taken less than or equal to
‘B’ which is divided into regular hexagons of sides a, We the sensing range of the sensors. i.e., a ≤ r. Given N ≥ 4
consider all sensors having uniform sensing range r and side MSN’s, we take four sensors, which are placed at a (0, 0), b
of the regular hexagon a is taken less than r. The hexagon is (0, a), c (a√3/2, a), d (0, a√3/2), and the sensors move as
further subdivided into three parts: two isosceles triangles indicated in Figure 3. Considering only the hexagons in
and a rectangle, as shown in Fig1. The whole rectangular the ROI i.e., the rectangular region, the 4 MSNs a, b, c, d
area is covered by m*n hexagons. move towards the right until the last column is reached.
Figure 1. Division of rectangular field into regular

hexagons Figure 3. Traversal schemes of MSNs
Out of these four MSNs only three MSNs cover the triangles
However, the area is over covered by some hexagons below the rectangles of the hexagon in first row. After
covering the perimeter of the ROI, shown as the shaded area reaching to the leftmost triangle of the first row these three
in Figure 1. The length (L) and breadth (B) of the targeted MSNs cover the top triangles of the hexagons in the second
region are related with m, n and side a of the hexagon by row until they cover up to the right most top triangle of
following relations: second row. Then these three MSNs along with the
remaining MSNs position on the four vertices of the right
L = a (m-1) √3, B = a (2n-1) most half rectangle of the second row and cover all the
Thus there are n rows of hexagons and in each row there are rectangles as done earlier in first row. This process
m hexagons. continues until the whole unshaded area in the Figure 3 is
covered. The distance traveled by these MSNs covering
The rows are numbered 1 to n and the columns are
rectangles, and triangles, rectangles in the indicated
numbered 1 to m. For sake of convenience we consider the
directions are detailed in the next paragraph. The time taken
center of the top most left hexagon as origin of reference, x-
is proportional to distance covered by the MSNs.
axis along horizontal line and y-axis as the vertical line
downwards In the first movement only two MSNs ‘a’ and ‘b’ together
The coordinate positions of the centers of hexagons dividing travel a distance of 3√3a to cover the first full rectangle, Fig
the rectangular field are represented as (x, y) = (a√3i, aj), 4(a). To cover the remaining full rectangles the 4 MSNs
which represent the position of the centre of the hexagon in together move a total distance of 4√3a in each move, to
ith row and jth column, where x= 1…n, y = 1…m and 1≤i≤n, cover a rectangle, Fig 4(b). To cover the last right most top
1≤j≤m. half-rectangle the MSNs ‘a’ and ‘b’ together travel a
distance of 3√3a, Fig 4(c). To cover the right-angled
We can see that each point in the ROI is covered by at least (right-most half triangle in the first row) triangle, two
2 sensors, which is depicted by the figure shown below MSNs (namely ‘d’ and ‘c’ in figure) out of these 4 MSNs
are kept stationery and the other two move along the
indicated directions and travel together a distance 2a, Fig 3.2.1 Total traveling Distance for different sides of
4(d). To cover the isosceles triangle one MSN (namely ‘c’ in varying length of the hexagons
figure) moves along the indicated direction a distance of
√13/2a, Fig 4(e). To cover the next isosceles triangles MSN To compare our results, with the results of [3], we have
‘b’ travels a distance of √7a, Fig 4(f). To cover the right- considered the area of ROI as a rectangular plot of size
angled triangle(right most half triangle) in the next row 4500*2000 units of measure. The side of the hexagon (taken
MSNs ‘a’, ‘b’, ‘c’ travel a distance of 3a, Fig 4 (vii) To less than the sensing range) is considered 45, 50,55,60,65
units. The Total traveling distance of MSNs for their
cover the rightmost half-rectangle in the next row the four
varying sensing ranges is determined. The traveling
MSNs (namely ‘a’, ‘b’, ‘c’, ‘d’ in figure) travel a total
distance of the MSNs decrease as the length, a, of the side is
distance of 5a , Fig 4 (g).This way the MSNs cover the
increased. i.e., we can say that the traveling distance of the
rectangular region (ROI) upto m columns and n rows. MSNs is inversely proportional to the length, a, of the side
of the hexagon. The data is graphically represented in
Figure 5, and in tabular form in Table 1.
Figure 4. MSN movements between hexagons
3.1 MTA with failure tolerance

In order to provide failure tolerance to MTA described
above, we add a few extra MSN’s which in the failure of a
particular MSN would occupy its position, otherwise these
extra MSN’s stay in the sleep mode.
Sides of regular hexagon
Figure 5. Total distance covered versus side lengths
of the hexagon
Table 1 Total distance covered by MSNs for

different side lengths of the hexagon
Length of side of
hexagon (a) Total traveling distance
Figure 5. Extra sensors to provide coverage in case of (D)
sensor failure 45 690677.78
50 624499.33
If a sensor fails at say i th row and jth column, then the 55 569478.6
sensor staying at the nearest corner to the coordinate 60 536880.36
position (i, j) will move to cover that point. 65 465344.62
3.2 Total traveling Distance
Based on the number of moves and individual traveling
In [3] the authors have taken equilateral triangle of sides of
distance of the MSNs the total traveling distance, D, is
length, a=50 units and distance traveled by the MSNs
calculated as:
covering the rectangular region and starting at arbitrary
D = n [6 3 a + 4 (m − 2 ) × 3 3 ] + points in the region varies from 7.38*105 to 7.54*105 which
[5 a + 2 13 a + 2 7 (m − 2 )a ](n − 1) + 5 a (n − 1) is much more than 6.91*105 obtained in our case for the
(1) regular hexagon having sides of length a=50.
4. Conclusions
In this paper, we present a MTA for the coverage of a Mrs. (Dr.) Seema Verma obtained her M.Tech and Ph.D degree
rectangular field by N≥4 mobile sensors. Though only 4 from Banasthali University in 2003 and 2006 respectively. She is
sensors are sufficient to cover the region, however, in case a associate Professor of Electronics. Her research areas are VLSI
of failure, extra MSN’s kept in reserve/sleeping mode are Design, communication Networks. She has published around 18
activated. Thus the system is reliable for coverage. We also research papers in various Journals.
observe that as side ‘a’ of the hexagon is increased the total
traveling distance covered by MSNs decreases. The
hexagonal topology provides an efficient and reliable
coverage as each point in the ROI is covered at least by 2-
sensors.
References
[1] A. Dhawan, S. K. Prasad, “Energy efficient distributed
algorithm for sensor target coverage based
Performance Computing, 2008.
[2] A. Khan, C. Qiao, S.K. Tripathi, “Mobile Traversal
Schemes based on Triangulation Coverage,” Mobile
Netw Appl, Vol. 12, pp. 422-437, 2007.
[3] Wang, H. B. Lim, D. Ma, “A survey of movement
strategies for improving network coverage in wireless
sensor networks,” Computer Communications, Vol.
32, pp. 1427- 1436, 2009.
[4] D. Brinza, Al. Zelikovsky, “Deeps: Deterministic
energy- efficient protocol for sensor networks,”
Proceedings of the International Workshop on Self
Assembling Wireless Networks (SAWN), pp. 261–
266, 2006.
[5] D. Tian, N. D. Georganas, “A coverage-preserving
node scheduling scheme for large wireless sensor
networks,” In WSN Proceedings of the 1st ACM
international workshop on Wireless sensor networks
and applications, New York, NY, USA, ACM, pp. 32–
41, 2002.
[6] S. Slijepcevic, M. Potkonjak, “Power efficient
organization of wireless sensor networks,” IEEE
International Conference on Communications (ICC),
Vol. 2, pp. 472– 476, 2001.
Authors Profile
Megha Sharma received the B.C.A and
M.C.A degree from I.G.N.O.U in 2004 and
2008, respectively. She is currently working
towards a Ph.D degree in computer Science at
the Banasthali University of Rajasthan. Her
research interests include wireless sensor
networks with a focus on the coverage of
wireless sensor networks.
Prof. G. N. Purohit is a Professor in

Department of Mathematics & Statistics at
Banasthali University (Rajasthan). Before
joining Banasthali University, he was
Professor and Head of the Department of
Mathematics, University of Rajasthan, Jaipur.
He had been Chief-editor of a research journal and regular
reviewer of many journals. His present interest is in O.R., Discrete
Mathematics and Communication networks. He has published
around 40 research papers in various journals.
Powers of a Graph and Associated Graph

Labeling
G. N. Purohit, Seema Verma and Usha Sharma
Centre for Mathematical Sciences,

Banasthali University, Rajasthan 304022
usha.sharma94@yahoo.com
two coloring problem and is equivalent to the proper

Abstract: - Graph coloring is a classical topic in graph theory coloring of the square of a graph, [11] includes labeling of
and vertex coloring of a graph is closely associated with channel many important graphs. Another generalization of labeling
assignment in wireless (sensor) network. Unit Disk graph is a (coloring) is L(h,1,1)- labeling, in which the labels on
suitable model for connectivity in sensor network. This paper is adjacent vertices differ by atleast h and labels on vertices at
concerned with the power of a graph in general and power of distance 2 or 3 are distinct [10]. This concept is applied in
Unit Disk graph in particular. L(1,1,1)- Labeling is used to channel assignment problem and in wireless (sensor)
avoid interference between communicating channels. We
network.
develop L(1,1,1)-Labeling of a UD graph. For this we make use
of cellular partition algorithm. We have proved that cube of any
UD graph can be properly colored by at most 25ω colors, where Unit Disk graph [8] is another class of a graph, which finds
ω is the maximum clique size. application in modeling a wireless (sensor) network. Since
the radio coverage range of sensors is based on Euclidean
Keywords: Graph Labeling, wireless network. distance between the nodes. So we utilize the concept of
Euclidean distance in a graph. This concept of Euclidean
1. Introduction distance in a graph has given rise to a new branch termed as
geometric graph theory. One can extend the concept of
A graph G = (V, E), where V is the set of vertices and E is power of graphs to the UD graph to obtain square and cube
the set of edges. Each edge i.e. element of E is an unordered of graphs and also Euclidean distance two graph [8] and
pair of element of V. Out of many induced graphs from a Euclidean distance three graph. Chromatic number of UD
graph; power graph finds a special place. Powers of a graph graph and square of UD graph is considered in [8]. These
have been considered in [1]. Square of a graph is a graph results are useful in the wireless sensor network technology.
with the same vertex set in which vertices at distance 2 are In this paper we describe some powers of a graph and
connected through an edge. Cube of a graph is also the powers of a unit disk graph. We develop a L(1,1,1)- labeling
graph on the same set of vertices; however, additionally of a UD graph by using cellular partition algorithm.
there is an edge between two vertices whenever they are at
most distance 3. This paper is organized as follows. In Section-2 we have
provided some auxiliary definitions. In particular we have
Graph coloring is a classical problem in graph theory and obtained some results related to powers of a cycle and a
proper coloring of a graph means assigning distinct colors complete bipartite graph. In Section-3 we have defined Unit
(labels) to adjacent vertices. The minimum number of colors Disk graph and its powers. Some results have been proved
required to color a graph G properly is called chromatic for powers of a UD graph. In Section-4 we have given
number of G and denoted as χ (G). A lot of research has cellular partition algorithm [8]. The main result of the paper
been done on the chromatic number of graphs. χ is bounded is theorem (4.1). This theorem shows that using the
by ω ≤ χ ≤ ∆+1 [5], where ω is the maximum clique size in developed cellular partition algorithm, cube of any UD
the graph and ∆ is the maximum degree of graph. The graph can be properly colored using 25ω colors, where ω is
chromatic number of powers of a graph has been studied in the maximum clique size. In last Section, we have given
[1]. conclusion.
Besides proper coloring there are many types of coloring 2. Auxiliary Definitions
(labeling) of vertices. One such generalization is L(p,q)-
labeling, in which the labels at adjacent vertices should 2.1 Graph Powers
differ by atleast p and labels at vertices at distance 2 should
differ by atleast q [6]. L(p,q)- labeling problem has attracted In this section we consider different powers of graph, which
attention of many researchers in the past [7]. Particular finds application in channel assignments, L(p, q)- coloring
cases of L(p,q)- labeling (i) L(1,1)- labeling and (ii) L(2,1)- of graphs etc.
labeling have been defined and a lot of research has been
done in this area. L(1,1)- labeling is also known as distance
2.1.1 Square of a graph (G2) - The Square G2 of a graph G in Gk and hence Gk will be a complete graph on n vertices
= (V, E) is the graph, whose vertex set is V itself and there and thus (C n)k =Kn.
is an edge between two vertices vi and vj if and only if their
graph distance (length of shortest path between vi and vj) in Case (ii): Let G = Cn (n≥2k+1) and let vi be an arbitrary
G is at most 2. Examples of a graph and its square graphs vertex of graph G. We have to show that deg (vi) = 2k ∀ vi є
are given in following figures: C n. We know that deg (vi) = 2, ∀ vi є C n . From the
definition of Gk, two vertices will be adjacent if the distance
between them is at most k.
There are exactly 2k vertices which are at most at a distance

k from vi. On one side of vi these k vertices vi+1, vi+2,
vi+3………. vi+k are at distance 1, 2, 3…….k from vi
respectively. Similarly on the other side of vi these k vertices
vi-1, vi-2, vi-3………. vi-k are at distance 1, 2, 3…….k from vi
respectively. Out of these 2k vertices vi+1 and vi-1 are already
Figure 1. (a) Cycle C6 Figure 1.(b) Square of Cycle C6 adjacent to vi in Cn and remaining 2k-2 vertices will be
made adjacent to vi in (Cn) k. Therefore deg (vi) = 2+2k-2 =
2.1.2 Cube of a graph (G3) - The Cube G3 of a graph G (V, 2k. Thus (Cn) k will be a 2k-Regular graph on n vertices.
E) is the graph, whose vertex set is V and there is an edge
between two vertices vi and vj if and only if their graph Theorem 2.2 If G = Km,n, then
distance in G at most 3. Examples of a graph and its cube Gk = Gk-1= Gk-2 =……………= G3 = G2 = Km+n .
graphs are given in following figures:
Proof: Let G = Km,n be a bipartite graph. Let V1 and V2 be

two partitions of vertex set V of G with m and n number of
vertices respectively. Let vi be an arbitrary vertex of V1.
Then all the vertices of V2 are at distance 1 from vi.
Moreover all other vertices of V1 are at distance 2 from vi.
Since vi is an arbitrary vertex therefore this is true for all vi’s
Figure 2. (a) Cycle C6 Figure 2. (b) Cube of Cycle C6 in V1 as well as for all vertices in V2. Thus all the pairs of
vertices are adjacent in G2. Thus G2 will be a complete
graph on m+n vertices. Gk (k>2) will not change Km+n.
We can generalize the above definitions as follows:- Thus Gk = Gk-1= Gk-2 =……………= G3 = G2 = Km+n.
2.1.3 Kth power of a graph (Gk) - The kth power Gk of a 2.2 Labeling of a graph G (V, E)
graph G (V, E) is the graph, whose vertex set is V and there
is an edge between two vertices vi and vj if and only if their 2.2.1 L (p, q) – Labeling - For two positive integers p and
graph distance in G is at most k. q, an L(p,q)- Labeling of a graph G is a function
C:V(G) → N such that C(vi) - C(vj) ≥ p if vertex vi and
As a special case we prove the following results for cycle Cn
and complete bipartite graph Km,n. vj are adjacent and C(vi) - C(vj) ≥ q if vertex vi and vj
are at distance 2.
Theorem 2.1 If G = Cycle with n vertices (Cn), Then
In particular L(1, 1)- labeling and L(2,1)- labeling are well
known examples of L(p,q)- Labeling.
2.2.2 L (1,1) – Labeling - It is also called the proper

Proof: Case (i): Let G = Cn (4≤n≤2k) and let vi and vj be labeling of a graph G. It is the labeling of the vertices with
any two arbitrary vertices of G. The maximum distance non negative integers such that the labels on adjacent
between vi and vj could be k in this case and thus any pair of vertices differ by at least 1.
vertices could have a graph distance at most k.
Thus max {d (vi, vj)│ ∀ vi, vj є V}≤k. 2.2.3 L (2,1) – Labeling - It is a labeling of the vertices
with non negative integers such that the labels on adjacent
From the definition of Gk, two vertices will be adjacent if d vertices differ by at least 2 and the labels on vertices at
(vi, vj) ≤ k. Since this condition is satisfied by all pairs of distance 2 differ by at least 1.
vertices in G. Therefore all pairs of vertices will be adjacent
We can generalize the above definition as follows:-
2.2.4 L (p,q,r) – Labeling - For three positive integers p, q V and there is an edge between two vertices vi and vj if and
and r, an L(p,q, r)- Labeling of a graph G is a function only if their graph distance in GUD is at most 3.
C:V(G) → N such that C(vi) - C(vj) ≥ p if vertex vi and
3.5 Euclidean distance three graph of a Unit Disk graph
vj are adjacent, C(vi) - C(vj) ≥ q if vertex vi and vj are at (GUDED3) - Euclidean distance three graph of a unit disk
graph GUD (V, E) is the graph whose vertex set is V and
distance 2 and C(vi) - C(vj) ≥ r if vertex vi and vj are at there is an edge between two vertices vi and vj if and only if
distance 3. their Euclidean distance in GUD is at most 3.
In particular L(1,1,1)- Labeling is more useful in channel
assignment problem and in wireless (sensor) network than
the others.
2.2.5 L (1,1,1) – Labeling - It is the labeling of the vertices

with non-negative integers such that the labels on adjacent
vertices, on vertices at distance 2 and 3 are different.
We can generalize it as follows:-
2.2.6 L (d1,d2,d3,…..di,…..dk) – Labeling:- It is a labeling

of the vertices with non-negative integers such that the
labels on vertices at distance i from each other differ at least
by di.
3. Powers of a Unit Disk graph

4(a)
For the sake of completeness, we first define unit disk
graph.
3.1 Unit Disk Graph- A graph G is a Unit Disk graph if

there is an assignment of unit disks centered at its vertices
such two vertices are adjacent if and only if one vertex is
within the unit disk centered at the other vertex. We denote
a unit disk graph by GUD.
3.2 Square of a Unit Disk Graph (GUD2) - The Square

GUD2 of a Unit Disk graph GUD (V, E) is the graph whose
vertex set is V and there is an edge between two vertices vi
and vj if and only if their graph distance in GUD is at most
2.
3.3 Euclidean distance two graph of a Unit Disk graph

4(b)
(GUDED2) - Euclidean distance two graph of a unit disk
graph GUD (V, E) is the graph whose vertex set is V and Figure 4. (a): Cube of a UD graph (GUD3)
there is an edge between two vertices vi and vj if and only if Figure 4. (b): ED-3 graph of a UD graph (GUDED3)
their Euclidean distance in GUD is at most 2.
Now we discuss some results relating to GUD and GUDED.
Theorem 3.1 For any Unit disk graph GUD, GUD2 ⊆ GUDED2.
Proof: The proof of this theorem is given in [8].
Theorem 3.2 For any Unit disk graph GUD, GUD3 ⊆ GUDED3.
Proof: Let GUD be a Unit Disk graph. GUD3 be the cube of

Figure 3. (a): Figure 3. (b): ED-2
GUD and GUDED3 be the Euclidean distance three graph of
Square of a UD graph of a UD graph
GUD. Since both the graph are on the same vertex set. So it
graph
is sufficient to prove this theorem that edge set of GUD3 is the
subset of edge set of GUDED3. Let (c, w) be an edge in GUD3.
3.4 Cube of a Unit Disk graph (GUD3) - The cube GUD3 of a There must exist two vertices u & v such that (c, u), (u, v),
Unit Disk graph GUD (V, E) is the graph whose vertex set is
(v, w) are three edges in GUD. Since GUD is a Unit Disk Now in order to label the nodes we adopt the Cellular
graph. Partition algorithm. In this algorithm first of all we partition
the whole plane in unit hexagonal cells with a side length ½
If we consider dEd(c, w) denotes the Euclidean distance thus the diagonal length of each cell is 1. If there is any UD
between c and w, then graph in this plane, vertices of the graph inside the cell will
form a clique, since no two vertices in the same hexagon
dED(c, w) ≤ dED(c, v) + dED (v, w) have a Euclidean distance greater than 1. Let the maximum
clique size be ω then there can not be more than ω vertices
≤dED(c, u) + dED (u, v) + dED (v, w) in the same hexagonal cell. Since we know that ω colors are
sufficient to color each hexagonal cell. Therefore we can
= 1+1+1 = 3. color the whole graph properly.
Hence dED (c, w) ≤ 3. Thus the edge (c, w) is an edge in Using the above Cellular Partition algorithm we prove the
GUDED3. following theorem:
Hence GUD3 ⊆ GUDED3 Theorem 4.1 Euclidean distance three graph GUDED3 for any
UD graph GUD can be properly colored by at the most 25ω
Further GUD3 may be proper subgraph of GUDED3 in some colors where ω is the maximum clique size.
instances. We will show that there might be an edge in
GUDED3 but not in GUD3. As shown in figure 4(b), there might Proof: We partition the whole plane into hexagonal cells
be a vertex x in GUDED3 such that 1<dED(c, x) ≤ 3 and but with side ½ and diagonal 1. All vertices included in any
there are no two vertices u' and v' such that (c, u'), (u', v') hexagon would form a clique. Since ω is the maximum
and (v', x) are edges in GUD. Thus (c, x) is an edge in GUDED3 clique size, so we could place at most ω vertices into each
but not in GUD3. Similarly there might be a vertex y in cell.
GUDED3 such that 1<dED(c, y) ≤ 2 but there are no vertex w'
such that (c, w') and (w', y) are edges in GUD. Thus (c, y) is Next we construct a patch of 25 hexagons* and use 25ω
an edge in GUDED3 but not in GUD3. color to color the patch. An example of the patch is shown
in figure [5] and keeps the same orientation of patches of 25
Theorem 3.3 For any UD graph GUD, a coloring scheme χ to cover the whole plane as shown in figure [6]. Now we
(GUDED3 ) for coloring GUDED3 would also color GUD3 which is prove that a vertex in ith hexagon in a patch would be at a
equivalent to L(1,1,1)- labeling of GUD. Euclidean distance of at least 3 to any other vertex in the ith
hexagon in any other adjacent patch.
Proof: Since we’ve proved in previous theorem, any GUD3 be
a subgraph of GUDED3 then ∃ a coloring scheme χ (GUDED3)
to color GUDED3 properly could be sufficient to any of its
subgraph. Therefore it would also color GUD3. Since L(1,1,1)
labeling of GUD is equivalent to proper coloring of GUD3 .
Thus χ (GUDED3) fulfill L(1,1,1) labeling of GUD.
4. Cellular Partition Algorithm
The concept of UD graph as well as labeling can be applied

in wireless sensor networks since we can model a wireless
sensor network as a UD graph. In this modeling sensors are
denoted as vertices. The sensing coverage area of a sensor is
represented by a unit disk centered at the corresponding Figure 5. A patch of 25 hexagons
vertex. The connectivity between two sensors is determined
as if one sensor is within the sensing coverage area of We maintain the same numbering orientation in a patch for
another sensor. If GUD represent a model of a wireless sensor all patches in the whole plane. So the distance between two
network, then GUD2 and GUD3 provide possible interfering vertices in ith hexagons in adjacent patches is constant. As
sensor nodes. To avoid this interference we need a proper an example let A, B and C be the centers of centre hexagons
labeling for GUD3 which is equivalent to L(1, 1, 1)- Labeling in the three adjacent patches as shown in figure [6]. Their
of GUD. distance can be computed as:
In order to cover the targeted area by sensors, we have to We know that AB = 10* 3 /4 = 4.33 > 4.
divide the whole area in smaller cells (area). We have
chosen regular hexagons to cover the whole plane based on Also we have AD = ½ + ½ + 1+ ½ + 1+ ¼
the observation that hexagon is the most suitable polygon = 15/4
which could cover the plane efficiently. It is the most
suitable tile that could cover the plane with no overlap and
thus it is the most efficient way to cover the plane.
References
[1] N. Alon and B. Mohar, “The chromatic number of graph

powers”, Comb. Probab. Comput. 11, 1 (2002), 1–10.
[2] G. J. Chang and D. Kuo, “The l(2,1)-labeling problem
on graphs”, SIAM J. Discret. Math. 9, 2 (1996), 309–
316.
[3] B. N. Clark, C. J. Colbourn and D. S. Johnson, “Unit
disk graphs”, Discrete Math. 86, 1-3 (1990), 165–177.
[4] A. Graf, M. Stumpf and G.Weisenfels, “On coloring
unit disk graphs”, Algorithmica 20, 3 (1998), 277–293.
[5] D. B. West, “Introduction to Graph Theory”, Second
edition Prentice Hall, 2001.
Figure 6. Cover the whole plane with the patches of 25 [6] B. M. K. Q. Peter Bella, Daniel Kral, “Labeling planar
hexagons graphs with a condition at distance two,” in
Proceedings 2005 European Conference on
CD = 5* 3 /4 Combinatorics, Graph Theory and Applications, 2005.
AD2 + CD2 = AC2 [7] M. Hall dórsson, “Approximating the l(h, k)-labelling
problem,” Engineering Research Institute, University of
2 Iceland Technical, Tech. Rep. Report No. VHI 03-2005,
Therefore AC =  15 
2
 5* 3  = 4.33 > 4. Since the
  +  

Available: citeseer.ist.psu.edu/252952.html
 4  4  [8] T. Ren, K. L. Bryan, and L. Thoma, “On coloring the
vertices in both the hexagons are at distance ½ only from square of unit disk graph,” University of Rhode Island
their centers A, B and C then the distance from any vertex Dept. of Computer Science and Statistics, Tech. Rep.,
in hexagon with center A to any vertex in hexagon with 2006.
center B will be greater than 3. Similarly the distance from [9] Kevin L. Bryan, Tiegeng Ren, Lisa DiPippo, Timothy
any vertex in hexagon with center A to any vertex in Henry, Victor Fay-Wolfe, “Towards Optimal TDMA
hexagon with center C will also be greater than 3. Therefore Frame Size in Wireless Sensor Networks”, University of
our patch fulfills the coloring of GUDED3. Rhode Island Dept. of Computer Science and Statistics,
Tech. Rep.
In the above coloring scheme each hexagon with a certain
[10] T. Calamoneri, E.G. Fusco, R.B. Tan and P. Vocca, “L
color pattern is far away from its sibling hexagon with same
(h,1,1)- labeling of outerplanar graphs”, Mathematical
color pattern. So it is a valid coloring for GUDED3. We can
Methods of Operations research, Volume 69, Number 2,
properly color any GUDED3 graph with 25ω colors. Using
May 2009, 307-312.
Theorem 3.2, it follows that it is valid color scheme for GUD3
also. [11] T. Calamoneri, “The L(h, k)-Labelling Problem: A
Survey and Annotated Bibliography”, The Computer
5. Conclusion Journal Vol. 49 No. 5, 2006.
Using the developed cellular partition algorithm, cube of

any UD graph can be properly colored using 25ω colors,
where ω is the maximum clique size. This is equivalent to
L(1,1,1)- labeling of unit disk graph and can be used to
avoid interference between communicating channels in
wireless (sensor) network. The number 25ω is the upper
bound and we are looking for obtaining a suitable lower
bound also.
Acknowledgement
Ms USHA SHARMA, one of the authors of this paper

acknowledges the grant received from Department of
Science & Technology (D.S.T.), Government of India, New
Delhi for carrying out this research.
A Method of Access in a Dynamic Context Aware

Role Based Access Control Model for Wireless
Networks 1
Dr. A.K. Santra, 2Nagarajan S
1
Director and Professor, MCA Department, Bannari Amman Institute of Technology, Sathayamangalam, TamilNadu.
2
Research Scholar, Bharathiar University, Coimbatore and Selection Grade Lecturer, Alliance Business Academy, Bangalore, Karnataka.
Abstract: This paper address security in dynamic context available at that point of time. This reduces the data storage
aware systems. Context awareness is a emerging as an important and transmission for using only the bits making it very
element in wireless systems. Security challenges in context much easy to complement in networks where the bandwidth
aware systems include integrity, confidentiality and availability of the network is very low.
of context information as well as end user’s privacy. The paper The anytime, anywhere access infrastructures is to enable a
addresses the dynamic changes happening in the mapping new generation of applications that can leverage
between the roles and permissions depending on context
continuously manage, adapt and finally optimization is
information. The paper presents a access control method using
artificial neural networks. It represents the data in terms of bits
required.
to express the roles and permissions which helps in reducing the The major challenge faced in Wireless applications is
data transmission and is a good fit for wireless networks with managing the security of the system using Access Control
lower bandwidth. It also introduces a novel method for storing Lists. ACL's is a very common mechanism used in Access
the information in a reduced format. Instead of accessing the Control. It has been observed that the ACL's are used to
access control tables the machine is learning it, which in turn check for permission to access resources or services.
reduces the time required to access the tables. Being dynamic in Another point to be noted at this juncture is such type of
nature there is no requirement for changes, any change is taken approach is very inadequate for wireless applications, since
care by the machine learning itself. Further, the algorithm is most proposed models do not take care of context
simple and easy to implement in wireless networks. information into consideration.
There is a need for giving control in a dynamic way as the
Keywords: Dynamic Context, Wireless Networks.
context changes according to location, time, system
resources, network security configuration etc., Therefore,
1. Introduction access control mechanism that changes the permission of a
user dynamically based on context information is very much
It has been proved that Dynamic Role Based Access Control essential.
can manage Access Control and security, more and more In this direction [3] have proposed a GRBAC Model and
mobile devices are incorporating this feature. Pervasive representing the system using State Machines. Using this
communication technology is becoming a everyday feature model, It is representing the information for the new
and it is changing the way of communicating with the algorithm proposed and show how it can be stored and
external world. This type of DRBAC requires the following retrieved. Then finally, show how this can be used to train
tables: 1. User Location Table 2. User Role Table 3. Role – the system without accessing the matrix.
Permission Table and 4. Mutual Exclusive role table.
Each time anybody accesses the system the first three tables
2. Background
are searched.
Further, there is a very complex mapping of Location, users,
Location, User, Role and Permission are the major
roles and permissions. It has been observed that frequently
components of a DRBAC which are represented as follows:
searching the tables reduces the efficiency of access control.
L = {L1, L2, ........................Li}
An disadvantage of wireless devices are that they have less
U = {U1, U2, .......................Ui}
power, storage, computing and transmission abilities.
R = {R1, R2, .......................Ri}
Hence, performing access control in wireless environments
P = {P1, P2, ........................Pi}
is actually more complex than that I wired environments.
T = {T1, T2, T3}
Therefore, any approach to access control must be relatively
The permission only directly maps to one role. In case many
simple and very efficient.
roles want to own the same permission, this need to be done
This paper addresses the following points:
using role inheritance. Since conflicted permissions also
It gives a access control algorithm and storage is reduced
needs to be addressed.
using the EAR decomposition and is retrieved accordingly.
It also uses a ANN to train the system so that this procedure
is learnt by the system, rather than searching the tables.
This algorithm assigns the user with different permissions
in different sessions depending on the context aware data
3. Dynamic Context Aware Role Based Access L1 = Campuses Abroad

Control L2 = Campuses coming under the home country
L3 = Campuses in each City
DRBAC addresses the dynamic requirement of applications L4 = Campuses within the city
in pervasive environments. It extends the traditional RBAC L5 = Residence
model to use dynamic context information while making Time
access control decisions. The DRBAC addresses the T1 = 8:00 AM to 8:00 PM (Office Hours)
following: T2 = 5:30 AM to 7:59 AM (Morning)
1. A user's access privileges must change when the T3 = 8:01 PM to 5:29 AM (Night)
Roles
user's context changes.
For Time T1
R1 = Professor
2. A resource must adjust its access permission when
R2 = Associate Professor
its system information changes. R3 = Assistant Professor
R4 = Teaching Assistant
R5 = Professor Remote
4. DRBAC Definitions R6 = Associate Professor Remote
R7 = Assistant Professor Remote
The DRBAC definitions are taken from the RBAC R8 = Teaching Assistant Remote
formalisms presented in [3] and [4] For Time T2
USER: A user is an entity whose access is being controlled. R9 = Professor
USERS represents a set of users. R10 = Associate Professor
ROLES: A role is a job function within the context of an R11 = Assistant Professor
organization with some associated semantics regarding the R12 = Teaching Assistant
authority and responsibility conferred on the user assigned R13 = Professor Remote
to the role. ROLES represents a set of roles. R14 = Associate Professor Remote
PERMS: A permission is an approval to access one or more R15 = Assistant Professor Remote
RBAC protected resources. PERMS represents a set of R16 = Teaching Assistant Remote
permissions. For Time T3
LOCATIONS: Locations is the set of points from where the R17 = Professor
user accesses the resources. LOCATIONS is the set of points R18 = Associate Professor
of access. R19 = Assistant Professor
TIMES: Times is the time at which the user access the R20 = Teaching Assistant
resources. Times is the set of time at which the user has the R21 = Professor Remote
access. R22 = Associate Professor Remote
SESSIONS: A session is a set of interactions between R23 = Assistant Professor Remote
subjects and objects. A user is assigned a set of roles during R24 = Teaching Assistant Remote
each session. The active role will be changes dynamically Permission
among the assigned roles for each interaction. SESSIONS P1 = Append
represents a set of sessions. P2 = Create.
UA : UA is the mapping that assigns a role to a user. In the P3 = Execute.
session, each user is assigned a set of roles, the context P4 = Get attribute.
information is used to decide which role is active. The user P5 = I/O Control.
will access the resource with the active role. P6 = Link.
PA : PA is the mapping that assigns permissions to a role. P7 = Lock.
Every role that has a privilege to access the resource is P8 = Read.
assigned a set of permissions, and the context information is P9 = Rename.
used to decide which permission is active for that role. P10 = Unlink.
Definition of the Agent: A Central Authority checks for the P11 == Write.
user's access rights. And gives the privileges that are active The Access Control Algorithm for wireless applications.
for him in that session. For the sake of this study it is considered that static IP
addresses are used. The wireless infrastructure
5. Explanation of the DRBAC Model implementing a WLAN is used for the logins inside the
campus; While Broadband wireless internet is used to login
The environment considered is an educational institute. The remotely.
designations are Professor, Associate Professor, Assistant Step 1: Using IPSec Labeling the process of authentication
Professor and Teaching Assistant. At office they will have is done as described in [5].
both read and write permissions. For this we represent the Step 2: Using the IP address associated with the user the
locations, roles and Time in the following way: location of the user is determined.
Step 3: Depending on the user's location a role is assigned
Locations which is further associated with permissions.
Using the following information we try to ascertain whether Based on the permission rights for that user the access is
a user is permitted to login from a particular location or not allowed. These two matrix are represented in the form of a
using matrix1. If the said user has access rights from that graph and then use the open ear decomposition technique to
location the step 2 of the algorithm is executed i.e., is reduce this information and store it.
mapping the IP address to a role else the access right is
denied. 6. Performance test of the algorithm
Matrix1 The test bed was created as a kernel program in SeLinux. It
L1 L2 L3 L4 L5 is allowed to run with the same modules that Se Linux has
U1 1 1 1 1 1 in addition to the modules created for this purpose.
Whenever somebody logins into the system it uses the
U2 0 1 1 1 1 authentication methods presently provided by the operating
U3 0 1 1 1 1 system. Using this to our advantage we put our static
U4 0 0 0 1 0 addresses specific to the location based on the labeling of
IPSec object called labeled IPSec. This particular feature is
. 0 0 0 1 0
available in mainline Linux version 2.6.16 itself. This does
U5 1 1 1 1 1 the authorization process as described in [5] and also we use
the same information to determine the location of the user.
The function of the second matrix defines the relationship Once the user's location is ascertained the next step is to
between the Location and roles for the time the user logs in. look out for the time at which this login has been requested.
Depending on the time the user logs in the roles are This is done with the help of the system clock. With this
assigned. This is used to check whether a role has access context information that is generated, access roles are
rights at various locations are not. Further, the permission accordingly assigned.
for the roles are defined during the time the role is created. The SELinux user identities are different from UNIX
If the role column in the matrix is 1 it means that role can identities. Here, for experimentation the normal roles
be provided access for that location and further step 3 of the defined are R1, R2, R3, R4, ............R24 and the
algorithm is executed else the access to that role is denied. corresponding Selinux roles defined are R1_r, R2_r, R3_r,
R4_r, ......R24_r. These roles are associated with the user.
Matrix2 The normal user are U1, U2, U3, U4, ........Un and the
For Time T1 corresponding Selinux users defined are U1_u, U2_u, U3_u,
R1 R2 R3 R4 R5 R6 R7 R8 U4_u, ........... Un_u.
Here _r identifies the roles while _u identifies the user.
L1 1 0 0 0 0 0 0 0 SELinux user identities are different from UNIX identities.
L2 1 1 1 0 0 0 0 0 They are applied as part of the security
label and can be changed in real time under limited
L3 1 1 1 0 0 0 0 0 conditions. SELinux identities are not primarily
L4 1 1 1 1 0 0 0 0 used in the targeted policy. In the targeted policy, processes
and objects are system_u, and the default
L5 1 1 1 0 1 1 1 0 for Linux users is user_u. When identities are part of the
policy scheme, they are usually identical to
For Time T2 the Linux account name (UID), and are compiled into the
R9 R10 R11 R12 R13 R14 R15 R16 policy. In such a strict policy, some system
accounts may run under a generic, unprivileged user_u
L1 1 0 0 0 0 0 0 0
identity, while other accounts have direct
L2 1 1 1 0 0 0 0 0 identities in the policy database
L3 1 1 1 0 0 0 0 0 _t identifies type. SELINUX_SRC/rbac is the place in which
roles are allowed to attain which other roles.
L4 1 1 1 1 0 0 0 0 Types are the primary security attribute Selinux uses in
L5 1 1 1 0 1 1 1 0 making authorization decisions as defined in permissions
above. This is defined in /etc/security/selinux/src/policy.
For Time T3 Depending on this roles can be assigned.
R17 R18 R19 R20 R21 R22 R23 R24 7. Representation of the Matrix and
decomposition / retrieval
L1 1 0 0 0 0 0 0 0
Using the three Matrix defined in the above method, the
L2 1 1 1 0 0 0 0 0 next step is to apply the well known Hungarian Algorithm
L3 1 1 1 0 0 0 0 0 to represent the matrix in the form of a graph. The Steps in
the Hungarian Algorithm is as follows:
L4 1 1 1 1 0 0 0 0
L5 1 1 1 0 1 1 1 0
Step 1 Similarly, the graphs for the other two matrix is drawn and
Generate initial labeling L and matching M in EL. reduced as shown.
Now, using the two graphs we apply the path ear
Step 2 decomposition algorithm. The steps of the path Ear
If M perfect, stop. decomposition algorithm is as follows:
Otherwise pick free vertex U such that it belongs to X. An ear decomposition D = [ P0 , P1 , P2 , ………., Pr-1 ] of an
Set S = { U } , T = Null. undirected graph G = (V, E) is a partition of E into an
ordered collection of edge-disjoint simple paths P0 , P1 , P2 ,
Step 3 ………., Pr-1 such that P0 is an edge, P0 U P1 is a simple
cycle, and each end point of Pi , for i > 1, is contained in
If NL (S) = T, Update labels (forcing NL(S) ≠ T) some Pj , j < i, and none of the internal vertices of Pj are
contained in any Pj , j < i. The paths in D are called ears. An
αl = mins € S, y does not belong to T. ear is open if it is non-cyclic and is closed otherwise. A
trivial ear is an ear containing a single edge. D is an open
l(v) – αl if v € S ear decomposition if all ears are open.
l’(v) = l(v) + αl if v € T Let D = [ P0 , P1 , P2 , ………., Pr-1 ] be an ear
l(v) otherwise decomposition for a graph G = (V, E). For a vertex v in V,
we denote by ear(v), the index of the lowest numbered ear
that contains v; for an edge e = (x,y) in E, we denote by
Step 4 ear(e) (or ear(x,y)), the index of the unique ear that contains
If Nl (S) ≠ T, Pick y € Nl (S) – T e. A vertex v belongs to Pear(v).
If y free, u – y is the augmenting path, The path ear decomposition algorithm:
Then Augment M and Go to step 2. Input: A connected graph G = (V, E) with a root r € V, and
Else with V = n.
If y matched, say to z, extend alternating tree: Output : A depth first search tree of G, together with a label
Such that, S = Su { z }, T = T U { y } on each edge in E, indicating its ear number.
Go to step 3. Set T of edges; integer count;
Procedure df s(vertex v);
Matrix1 and its graph representation G1 { * This is a recursive procedure. The call df s(v) of the
main program constructs a depth first search tree T of G
rooted at r; the recursive call df s(w) constructs the sub tree
of T rooted at w. The depth first search tree is constructed
by placing the tree edges in the set T and labeling the
vertices in the sub tree rooted at vertex v in pre-order
numbering, starting with count. The procedure assigns ear
labels to the edges of G while constructing the depth first
search tree. An edge that does not belong to any ear is given
the label (∞, ∞). Initially, all vertices are unmarked. * }
Vertex w;
‘mark’ v;
Pre-order(v) := count; count := count + 1; low(v) := n;
ear(v) := (n,n);
For each vertex w adjacent to v
{ * This for loop performs a depth forth search of each child
of v in turn and assigns ear labels to the tree and non tree
Matrix2 and its graph representation G2
edges incident on vertices in the sub trees rooted at the
children of v. * }
If w is not marked
Add (v,w) to T; parent(w) : = v; df s(w);
If low(w) ≥ pre-order(w)
ear(parent
(w), w) := (∞, ∞)
Low(w) < pre-order(w) ear(parent(w),w) := ear(w)
Fi;
Low(v) := min(low(v), low(w));
Ear(v) := lexmin(ear(v), ear(w))
If w is marked
If w ≠ parent (v)
Low(v) := min(low(v), pre-order(w));
Ear (w, v):= (pre-order(w), pre-order(v)) References

Ear (v) := lexmin(ear(v), ear(w,v)); [1] Efficient Access Control in Wireless Networks, Kun
Fi Wang, Zhenguo Ding, Lihua Zhou, Proceedings of
Fi
IEEE/WIC/ACM International Conference on Web
Rof
End df s; Intelligence and Intelligent Agent Technology. 85-88,
{* Main program *} ISBN:0-7695-2749-3, 2006.
T: = Null set; count: = 0; df s(r);
Sort the ear labels of the edges in lexicographically non- [2] Fast Access Control algorithm in Wireless Network,
decreasing order and relabel distinct labels (expect labels Kun Wang, Zhixin Ma, This paper appears in: Grid
(∞, ∞)) in a order as 1,2,3,,4,………; and Pervasive Computing Workshops, 2008. GPC
Relabel the non tree edge with label 1 as 0 Workshops '08. The 3rd International Conference on,
End.
ISBN 978-0-7695-3177-9, 347p – 351p, 25-28 May
Using the algorithm the graph G1, G2 and G3 reduces to
G11, G21 and G31 respectively. 2008.
Graph G1 reduced to the form G11 [3] Context-Aware Dynamic Access Control for Pervasive
P1 = { < U1, L1 > < U5, L1 > } Applications,. G. Zhang and M. Parashar, Proceedings
P2 = { < U1, L2 > < U5, L2> } of the Communication Networks and Distributed
P3 = { < U1, L3 > < U5, L3 >} Systems Modeling and Simulation Conference (CNDS
P4 = { < U1, L4 > < U5, L4 >} 2004), 2004 Western MultiConference (WMC), pp.
P5 = { < U1, L5 > < U5, L5} 219 . 225, January 2004.
P6 = { < U2, L2 > U3, L2 > < U3, L3 > }
[4] Supporting relationships in access control using role
P7 = { < U2, L3 > }
P8 = { < U3, L4 > < U4, L4 > } based access control. K. Beznosov, J Barkley and J
P9 = { < U3, L5 > } Uppal, Symposium on Access Control Models and
P10 = { <U4, L4 > } Technologies, Proceedings of the fourth ACM
Therefore, G11 = { P1, P2, P3, P4, P5, P6, P7, P8, P9, P10} workshop on Role-based access control, Fairfax,
Graph G2 reduced to the form G21 Virginia, United States, 55p – 65p, ISBN:1-58113-
P1 = { < L1, R1> < L2, R1 > < L2, R2 > < L3, R3 > < L4, 180-1 1999.
R3 > < L4, R4 >}
[5] Leveraging IPsec for Distributed Authorization, Trent
P2 = { < L2, R3 > < L5, R3 > < L5, R5 > }
P3 = { < L5, R6 > < L5, R7 > } Jaeger, David King, Kevin Butler, Jonathan McCune,
Therefore, G21 = { P1, P2, P3} Ramon Caceres, Serge Hallyn, Joy Latten, Reiner
Similar operation is performed on the other two graphs. Sailer and Xiolan Zhang.
G11 and G21 are referred to as the partition matrix and can nsrc.cse.psu.edu/tech_report/NAS-TR-0037-2006.pdf,
be called partition path matrix. The path decomposition is 2006
edge disjoint one Whence the union of the path reduced will
give the entire graph G1 and G2.
Authors Profile
8. Conclusion Dr. A.K.Santra is presently working as
the Director (Computer Applications), at
It has been observed that any dynamic context aware system the Bannari Amman Institute of
needs to search relative tables to get the user permissions. Technology in Sathyamangalam. He has
This paper is presenting a dynamic context aware algorithm close to 40 years of experience both in
using SElinux where the number of tables are reduced. It the industry and Teaching. He published
also shows a way to store it and retrieve. Executing our 17 papers in various International
module the roles are assigned according to the location and Journals and conferences. He is presently guiding a number
time. Hence it can be implemented with ease in a wireless of students for their Ph. D. degrees. He is on the board and
networked environment. a reviewer in various International Journals.
Acknowledgements
Mr. Nagarajan S is presently working
We Would like to thank Prof. K. A Venkatesh, HOD as Selection Grade Lecturer, at the
Department of Computer Applications, Alliance Business Alliance Business Academy, Bangalore.
Academy for all his support and discussions. We would also He is also a Research Scholar at
like to thank Mr. Mahesh M S for the experimental support Bharathiar University at Coimbatore. He
provided in the lab during the preparation of this algorithm has nearly about 13 years of Industry and
and module. teaching experience. He has published
one international paper in an
International Journal and 5 in various conferences.
Design of a Novel Cryptographic Algorithm using

Genetic Functions
Praneeth Kumar G1 and Vishnu Murthy G2
1
C V S R College Of Engineering,
Ghatkesar, Andhra Pradesh, India
praneethkumar@hotmail.com
2
C V S R College Of Engineering,
Ghatkesar, Andhra Pradesh, India
gvm189@gmaill.com
that both have remainder of 3 when divided by 4. That is,

Abstract: Information Security plays a key role in the field of
modern computing. Here, at this paper we present a new p≡q≡3 (mod 4)
cryptographic algorithm which is proven to be resistant to means that (p mod 4)= (q mod 4)= 3. Let n = p X q. Next,
Cryptanalysis, Bruteforce and timing attacks. As the algorithm choose a random number s, such that s is relatively prime to
uses Blum Blum Shub Genrator, A Cryptographically Secure n; this is equivalent to saying that neither p nor q is factor of
Pseudorandom Bit Generator (CSPRBG) for deriving the key s. Then the BBS generator produces a sequence of numbers
and Gentic Funtions in the process of Encryption. A comparison
Xi according to the following algorithm:
of the proposed technique with existing and industrially accepted
RSA and Triple-DES has also been done in terms of resistance X0 = s2 mod n.
to attacks and the various features of the algorithm. for i =1 to infinite
Xi = (Xi-1)2 mod n
Keywords: Encryption, Decryption, Blum, Blum, Shub
Generator, Genetic Functions.
The BBS is referred to as a cryptographically secure
pseudorandom bit generator (CSPRBG). A CSPRBG is
1. Introduction defined as one that passes the next- bit test , Which is
defined as follows: “A Pseudo random bit generator is said
Information Security plays a vital aspect of modern to pass the next-bit test if there is not a polynomial-time
computing systems. With the global acceptance of the algorithm that, an on input of the first k bits of an output
Internet, virtually every computer is connected to every sequence, can predict the (k+1)st bit with probability
other. So at this point of time maintaining secrecy and significantly greater than 1/2”. The security of BBS is based
security of information has become necessity. For these on the difficulty of factoring n. That is, we need to
reasons different types of research works on encryption and determine its two prime factors p and q.
decyption is going on so that various algorithms are 2.2 Genetic Functions
developed in this field. The process of encoding a message In the proposed algorithm we use two genetic functions
so that it can be read only by the sender and the intended “CROSSOVER” and “MUTATION”.
recipient is known as encryption. The encoded version is Crossover is a genetic function which can be described by
known as cipher text and process of decoding the cipher text the following figure: As Illustrated in the figure the Binary
is known as Decryption. representation of key and plain text are Crossected. We have
The Algorithm uses Blum Blum Shub Generator for two forms of crossover: Single and Double Crossover.
generating key and Genetic functions “CROSSOVER” and Taking 1 breaking point for a single crossover and 2
“MUTATION” in the process of encryption and decryption. breaking points for double crossover.
The Algorithm uses a key of four parameters, for security Crossover :
which makes it resistant against Bruteforce attack.
Key = {p, q, s, k}
Where, p, q are two large prime numbers and s is a
randomly chosen number where s is relatively prime to n
(product of p and q) and k is Key Size used.
2. Literature Survey
2.1 Blum, Blum, Shub Generator

A popular approach for generating secure pseudorandom
number is known as the Blum, Blum, Shub (BBS)
generator, named for its developers[1]. The procedure is as Mutation is a genetic function where the bit at a given
follows. First, choose two large prime numbers, p and q , position is inversed (i.e., 0 to 1 and vice versa).
3. Proposed Algorithm AB AD A’ D
è(Crossover) è (Mutation)
The algorithm consists of two phases where the first phase
CD CB C’ B
is of generating random numbers and the other performs
encryption/ decryption.
Double Crossover
3.4 Key Generation Suppose that the Message is AB and Key is CD
The algorithm uses a 4-tuple key {p, q, s, k} where p and
q are large prime numbers, s is a chosen random number Where, A is part of Plain text before first breaking point
which is relatively prime to n, the product of p and q and k, B is part of Plain text between first and second
the key size. The key size is of Variable one. breaking points.
Then, the algorithm uses the Blum, Blum, Shub C is part of Plain text after the second breaking
Generator for generating the random numbers (Which is point.
described in Section 2.1) which are used as keys in each D is part of Key before first breaking point.
iteration for encryption. E is part of Key between first and second breaking
1. Choose p=7 and q=19
point.
2. Implies, n= 7 X 19 = 133
F is part of Key after the second breaking point.
3. Choose s=100, relatively prime with 133
4. Then, X0 =s2mod n= (100)2 mod 133= 25
X1=(X0)2 mod n= (25)2 mod 133= 93 ABC DBF D B’ F
X2=(X1)2 mod n= (93)2 mod 133= 4 è(Double Crossover) è(Mutation)
X3=(X2)2 mod n= (4)2 mod 133= 16 DEF AEC A E’ C
. .
. . Then, the Plain Text’s (Binary representation of ASCII
Here, the key is represented as {7, 19, 100, 8 }. code) is cross-over’d with Key (Binary representation)
generated by BBS (Section 3.1)
3.5 Encryption/ Decryption Algorithm
The proposed algorithm follows the below given method Here, The Cipher text that will be sent consists of 2 numbers
for encryption and decryption. The Random numbers should A’D and C’B instead of AB in the reverse number (if single
be generated concurrently in both the processes. crossover is performed) and DB’F and AE’C instead of
3.2.4 Encryption ABC in the reverse order (if double crossover is performed).
The Encryption process is carried out as :
for every bit in the file until EOF For the plain text “TEXT” the encryption process is as
if random number generated is odd follows:
perform CROSSOVER between plain text(binary
representation of ASCII value) and the random Character ASCII Value Binary Value
number(in binary representations ) where the T 83 01010011
breaking point is xi%k . E 69 01000101
else if the number generated is even X 87 01010111
perform Double CROSSOVER between plain T 83 01010011
text(binary representation of ASCII value) and the
random number(in binary represenations) where the 01010011(83) 01011001 01001001(73)
first breaking point is xi%k and second one is è(Crossover) è(Mutation)
(xi+s)%k. 00011001(25) 00010011 00000011(3)
perform MUTATION at the (2*xi)%k position in the
offsprings.. So the Cipher Text is (3, 73). This process is continued until
all the text in source file (Plain text) is completed.
The set of two numbers from the above output is the cipher 3.2.5 Decryption
text.
Single Crossover The Decryption process is carried out as :
Generate random numbers concurrently.
Suppose that the Message is AB and Key is CD for every bit in the file(cipher text) until EOF
Where, A is part of Plain text before breaking point if random number generated is odd
B is part of Plain text after breaking point. read two characters at a time.
C is part of Key before breaking point. perform CROSSOVER between the second number
D is part of Key after breaking point. read and the xi(binary representations) where the
breaking point is n%k .
crossovered numbers.
perform CROSSOVER between first offspring of the Ease of analysis : The algorithm is explained concisely over
above phase and the first character read(in binary here. Even though it is difficult to cryptanalyze . RSA and
representations). DES lacks in this feature.
else if the number generated is even
perform Double CROSSOVER between the second 5. Conclusion and Future Enhancements
number and the key (binary representation of ASCII
Hence, The paper proposes a new algorithm which is
value)where the first breaking point is xi%k and
equivalently secure with RSA and Triple DES and which
second one is (xi+s)%k .
can be easily implemented on the hardware.
crossovered numbers. Future process will be devoted to extend the algorithm to
perform CROSSOVER between first number of the achieve the other security services like Authentication, Data
above output and the first character read(binary Integrity etc.,
representations).
References
The first number of the above output is the plain text(if
single crossover is to be performed) [1] Lenore Blum, Manuel Blum, and Michael Shub.,
“Comparision of two pseudo random number
generators” Proc. CRYPTO’82, pages. 61-78,
A’ D A’ D A D
Newyork, 1983.
è (Crossover) è(Mutation)
[2] William Stallings, “Cryptography and Network
C D C’ D C’ D
Security”, Prentice Hall, 3rd Edition.
[3] Subramil Som, Jyotsna Kumar Mandal and Soumya
A D A B à Plain Text Basu, “A Genetic Functions Based Cryptosystem
è(Crossover) (GFC)”, IJCSNS, September 2009.
C’ B C’ D [4] Ankit Fadia, “Network Security”, Macmillan India Ltd.
If double crossover is to be performed

Authors Profile
D B’ F D B’ F DBF
è(Double Crossover) è(Mutation) Praneeth Kumar G received the B.Tech
DE F DEF DEF Degree in Computer Science and
Engineering from Progressive Engineering
College in 2008. During May’ 2008 –
DB F ABC à Plain Text
Aug’ 2009, he worked in Concepts in
è(Double Crossover) Computing(CIC) as a Software Engineer.
A E’ C D E’ F He is presently working at C V S R
College of Engineering as an Assistant
4. Analysis Professor. His areas of interest include
software engineering and Information
The proposed algorithm has the following advantages : Security.
Suitable for hardware or software: Algorithm uses only
primitive computational operations that can be easily Vishnu Murthy G received the B.Tech.
and M.Tech. degrees in Computer Science
implemented on hardware in a less economic way. Which is
and Engineering. He is resource person for
not possible with RSA and Triple DES. IEG and Birla Off campus programmes.
Variable-length key : The key length can be varied in the He is presently pursuing his Ph.D in
algorithm which is possible in RSA but not in Triple DES. J.N.T.U. and heading the Department of
Low memory Requirement : A low memory requirement Computer Science and Engineering in C V
S R College Of Engineering. His areas of
makes the proposed algorithm suitable for smart cards and
interest include software Engineering,
other devices with restricted memory which is not possible Information Security and Image
in RSA and Triple-DES. Processing.
Resistant to Known Plain Text, Known Cipher Text and
Bruteforce Attacks
Resistant to Timing Attack : As the algorithm uses Blum,
Blum, Shub generator for Key Generation it is resistant to
timing attacks(Section 2.1). RSA is prone to this kind of
attack but Triple-DES is not.
Computationally Secure : As the proposed algorithm maps
each character in plain text to two characters in cipher text.
It is hard to break the cipher. This feature is present in both
the RSA and Triple-DES algorithms.
Cluster Management using Cluster Size Ratio in Ad

Hoc Networks
D K L V Chandra Mouly, Ch D V Subba Rao and M M Naidu
Department of Computer Science and Engineering
S V University College of Engineering, Tirupati - 517502, India.
dklvcm.mtech@gmail.com, subbarao_chdv@hotmail.com
Abstract: Cluster Management using Cluster Size Ratio delivering messages in a decentralized environment where
(CMCSR) is a completely distributed algorithm for partitioning network topology fluctuates is not a welldefined problem
a given set of mobile nodes into clusters. The proposed algorithm [1].
tries to reduce the amount of computational and information
overhead while maintaining a stable cluster formation. It 1.2 Clustering in Ad Hoc Networks
constructs and maintains a backbone topology based on a A wireless ad hoc network consists of nodes that move freely
minimal dominating set (MDS) of the network. According to this
and communicate with each other using wireless links. Ad-
algorithm, each node determines the membership in the MDS for
itself and its one-hop neighbors based on one-hop neighbor
hoc networks do not use specialized routers for path
information that is disseminated among neighboring nodes discovery and traffic routing. One way to support efficient
using willingness and priority information of the nodes. The communication between nodes is to develop wireless
algorithm then ensures that the members of the MDS are backbone architecture; this means that certain nodes must be
connected into a connected dominating set (CDS), which can be selected to form the backbone. Over time, the backbone
used to form the backbone infrastructure of the communication must change to reflect the changes in the network topology
network to facilitate routing. The algorithm outperforms the as nodes move around. The algorithm that selects the
existing algorithms with respect to stability. Load balancing the members of the backbone should naturally be fast, but also
cluster heads using the cluster size ratio is the heuristic used in should require as little communication between nodes as
this algorithm. possible, since mobile nodes are often powered by batteries.
One way to solve this problem is to group the nodes into
1. Introduction clusters, where one node in each cluster functions as cluster
head, responsible for routing [2].
This section discusses elementary issues of ad hoc networks
and benefits of clustering. 1.3 Benefits of clustering
Ad-hoc networks are suited for use in situations where an
1.1 Ad Hoc Networks infrastructure is unavailable or to deploy one is not cost
In the next generation of wireless communication systems, effective. One of many possible uses of mobile ad-hoc
there will be a need for the rapid deployment of independent networks is in some business environments, where the need
mobile users. Significant examples include establishing for collaborative computing might be more important
survivable, efficient, dynamic communication for outside the office environment than inside, such as in
emergency/ rescue operations, disaster relief efforts, and business meeting outside the office to brief clients on a
military networks. Such network scenarios cannot rely on given assignment.
centralized and organized connectivity, and can be Mobile ad-hoc networks allow the construction of flexible
conceived as applications of ad hoc networks. An ad hoc and adaptive networks with no fixed infrastructure. These
network is an autonomous collection of mobile users that networks are expected to play an important role in the future
communicate over relatively bandwidth constrained wireless wireless generation. Future wireless technology will require
links. Since the nodes are mobile, the network topology may highly-adaptive mobile networking technology to effectively
change rapidly and unpredictably over time. The network is manage multi-hop ad-hoc network clusters, which will not
decentralized, where all network activity including only operate autonomously but also will be able to attach at
discovering the topology and delivering messages will be some point to the fixed networks.
taken care by the nodes, i.e., routing functionality will be
incorporated into mobile nodes. 2. Literature Review
The set of applications for ad hoc networks is diverse, This section emphasizes some of the past clustering
ranging from small, static networks that are constrained by techniques.
power sources, to large-scale, mobile, highly dynamic
networks. The design of network protocols for these 2.1 Types of Topology Management
networks is a complex issue. Regardless of the application, There are two approaches to topology management in ad
ad hoc networks need efficient distributed algorithms to hoc networks:
determine network organization, link scheduling, and
• Power control.
routing. However, determining viable routing paths and
• Hierarchical topology organization.
2.1.1 Power Control C. Reactive Clustering

Different Power control mechanisms adjust the power on a It is on-demand clustering algorithm. There is no periodic
per-node basis, so that one-hop neighbor connectivity is exchange of clustering information in the network. Instead,
balanced and overall network connectivity is ensured [4, 5, whenever there is data traffic, cluster related information is
6]. Li [7] proved that network connectivity is minimally piggybacked in outgoing data packets and extracted out of
maintained as long as the decreased power level keeps at received packets.
least one neighbor remaining connected at every 2π/3 to D. Proactive Clustering
5π/6 angular separation. Ramanathan [7] proposed to Some are proactive clustering algorithms, which require
incrementally adjust nodes power levels so as to keep periodic broadcast of cluster-related information. SPAN [13]
network connectivity at each step topologies derived from adaptively elects coordinators according to the remaining
power-control schemes often result in unidirectional links energy and the number of pairs of neighbors a node can
that create harmful interference due to the different connect.
transmission ranges among one-hop neighbors [9]. The
dependencies on volatile information in mobile networks, 2.3 Scope for Present Work
such as node locations [4], signal strength or angular
positions [8] also contribute to the instability of topology The efficiency of a communication network depends not
control algorithms based on power control. only on its control protocols, but also on its topology. Our
work i.e. CMCSR proposes a distributed topology
2.1.2 Hierarchical topology control management algorithm that constructs and maintains a
This approach to topology control is often called clustering, backbone topology based on a Minimal Dominating Set
and consists of selecting a set of cluster heads in a way that (MDS) of the network. Without topology management each
every node is associated with a cluster head, and cluster and every node should maintain the routing information for
heads are connected with one another directly or by means all the nodes they need. By using topology management a
of gateways, so that the union of gateways and cluster heads subset of nodes are selected called cluster heads and each
constitute a connected backbone [10, 14, 15]. Once elected, cluster head performs the routing work for its members.
the cluster heads and the gateways helps to reduce the
complexity of maintaining topology information, and can 3. System Model
simplify such essential functions as routing, bandwidth
allocation, channel access, power control or virtual circuit 3.1 Assumptions
support. For clustering to be effective, the links and nodes This work assumes that an ad hoc network comprises a
that are part of the backbone (i.e., cluster heads, gateways, group of mobile nodes communicating through a common
and the links that connect them) must be close to minimum broadcast channel using omni-directional antennas with the
and must also be connected. same transmission range. The topology of an ad hoc
network is thus presented by an undirected graph G = (V,E),
2.2 TYPES OF CLUSTER HEAD ELECTIONS where V is the set of network nodes, and E ⊆ U * V is the
set of links between nodes. The existence of a link (u, v) ∈ E
Cluster heads can be elected in four ways.
also means (v; u) ∈ E, and that nodes u and v are within the
• Deterministic Clustering
packet-reception range of each other, in which case u and v
• Non-Deterministic Clustering
are called one-hop neighbors of each other. The set of one-
• Reactive Clustering
hop neighbors of a node i is denoted by Ni. Two nodes that
• Proactive Clustering
are not connected but share at least one common one-hop
A. Deterministic Clustering neighbor are called two-hop neighbor of each other.
Deterministic clustering can determine the cluster heads in a
single round. Different heuristics have been used to form Each node has one unique identifier, and all transmissions
clusters and to elect cluster heads. Several approaches [12] are omni directional with the same transmission range. The
utilized the node identifiers to elect the cluster heads within nodes move with constant mobility. The energy is decreased
one or multiple hops. linearly. Different types of nodes consume energy at
different rates. We ignore the energy consumed due to local
B. Non-Deterministic Clustering computations, but assume that the energy consumption rate
is only dependent on the type of the node. A host consumes
In non-deterministic clustering, negotiations are used.
0.6% of the total energy per minute in these algorithms;
Negotiations require multiple incremental steps, and may
whereas a cluster head consumes 3%.
incur an election jitter during the process, because of the
lack of consensus about the nodes being elected as the
3.2 Model
cluster heads. Examples of this approach are the “core”
extraction algorithm [13] and the spanning tree algorithm In an ad hoc network all nodes are alike and all are mobile.
[14]. SPAN [13] allows a node to delay the announcement There are no base stations to coordinate the activities of
of becoming a cluster head for random amounts of time to subsets of nodes. Therefore, all the nodes have to
attempt to attain minimum conflicts between cluster heads collectively make decisions. All communication is over
in its one-hop neighborhood. wireless links. A wireless link can be established between a
pair of nodes only if they are within wireless range of each
other. We will only consider bidirectional links. It is priorities that consists of two components: (a) Neighboring
assumed the MAC layer will mask unidirectional links and Nodes, (b) Willingness value assigned to a node as a
pass only bidirectional links. Beacons could be used to function of its mobility and energy level.
determine the presence of neighboring nodes. After the We denote the willingness value of node i by Wi, the speed
absence of some number of successive beacons from a of node i by a scalar Mi that ranges from 0 to 1 meters per
neighboring node, it is concluded that the node is no longer second, and the remaining energy on node i as E i in the
a neighbor. Two nodes that have a wireless link will, range of 0 and 1. The willingness Wi is a function that
henceforth, be said to be one wireless hop away from each should be defined according to the following criteria:
other. They are also said to be immediate neighbors.
Communication between nodes is over a single shared 1. To enhance survivability, each node should have the
channel. responsibility of serving as a cluster head with some
nonzero probability determined by its willingness value.
In ad hoc networks the nodes within each neighborhood are 2. To facilitate with the stability and the frequency with
not known a priori. The individual cluster may transition to which cluster head elections must take place, the
spatial TDMA for inter-cluster and intra-cluster willingness value of a node should remain constant as long
communication. All nodes broadcast their node identity as the variation of the speed and energy level of the node do
periodically to maintain neighborhood integrity. Due to not exceed some threshold values.
mobility, a node’s neighborhood changes with time. As the 3. To avoid electing cluster heads that quickly lose
mobility of nodes may not be predictable, changes in connectivity with their neighbors after being elected, the
network topology over time are arbitrary. However, nodes willingness value of a node should decrease drastically after
may not be aware of changes in their neighborhood. the mobility of the node exceeds a given value.
Therefore, clusters and cluster heads must be updated 4. To prolong the battery life of a node, its willingness
frequently to maintain accurate network topology. value should decrease drastically after the remaining energy
of the node drops below the given level.
3.2.1 Attributes of a node
Willingness value (Wi) is as specified below:
The attributes of a node and their functionality are as given log2(Ei+.9)log2(Mi+2)
Wi = 2
in Table I.
Here the constants 0.9 and 2 in Eq. (1) eliminate the
Table 1: Attributes of a node and their functionality
boundary conditions in the logarithmic operations. The
ATTRIBUTE FUNCTION logarithmic operations on the speed and the remaining
energy values render higher willingness values in the high
ID Unique name given to node
energy and low speed field, while giving close to zero values
ENERGY The capacity to work in
in the low energy and high-speed region.
MOBILITY The speed of the node when it
is moving Priority value (Pi) is a function of no.of neighbors and
WILLINGNESS How much the node is willing willingness
to be a cluster head i.prio = 2log2(Wi)/n
PRIORITY Has the priority among other Figure 1 illustrates the effect of the two factors on the
nodes to became a cluster head priority values. From the Figure 2 we can conclude that the
CLUSTER SIZE Cluster size ratio which it is priority is directly proportional to the willingness value and
having number of neighbors.
TYPE Whether it is cluster head or
gateway or door way or
member
NEIGHBORS Number of one – hop neighbors
1
3.2.2 Computing Priorities of Nodes 0.9

0.8
Given that cluster heads provide the backbone for a number 0.7
0.6
of network control functions, their energy consumption is Priority 0.5
more pronounced than that of ordinary hosts. Low-energy 0.4 10
nodes must try to avoid serving as cluster heads to save 0.3 7
0.2
energy. However, to balance the load of serving as cluster 0.1
4
heads, every node should take the responsibility of serving 0 1 Number of
as a cluster head for some period of time with some 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Neighbors
likelihood. Furthermore, node mobility has to be considered Willingness
in cluster head elections. To take into account the mobility

and energy levels of nodes in their election, we define the
two-hop neighbor information needed to assign node Figure 1. Priority Graph
The objective is to develop an enhancement for existing

heuristics to provide a contiguous balance of loading on the
3.2.3 MDS and Cluster Head Election
elected cluster heads. Once a node is elected as cluster head
The approach to establishing a minimal dominating set it is desirable for it to stay as a cluster head up to some
(MDS) is based on three key observations First, using maximum specified amount of time, or budget. The budget
negotiations among nodes to establish which nodes should is a user defined constraint placed on the heuristic and can
begin the MDS incurs substantial overhead when nodes be modified to meet the unique characteristics of the system,
move around and the quality of links changes frequently. i.e., the battery life of individual nodes. Some of the goals of
Hence, nodes should be allowed to make MDS membership the heuristic are:
decisions based on local information. Second, because in an 1. Minimize the number and size of the data structures
MDS every node is one hop away from a cluster head, the required to implement the heuristic,
local information needed at any node needs to include only 2. Extend the cluster head duration budget based on an input
nodes that are one and two hops away from the node itself. parameter,
Third, having too many cluster heads around the same set of 3. Allow every node equal opportunity to become a cluster
nodes does not lead to an MDS. Hence, to attain a selection head in time,
of nodes to the MDS without negotiation, nodes should rank 4. Maximize the stability in the network.
one another using the two-hop neighborhood information
they need. Based on the above, the approach adopted in Data Structures
CMCSR consists of each node communicating to its The data structures necessary for the heuristic consist of one
neighbor’s information about all its two-hop neighbors. local variable: Physical ID (PID). The PID is the initial id
Using this information, each node computes a priority for given and is unique for each individual node. However, this
each node in its two-hop neighborhood, such that no two changes with time to represent the elect ability of a node.
nodes can have the same priority at the same instant of time.
A node can become cluster head if the node has highest Basic Idea
priority in its two hop neighborhood. The node id load heuristic operates on the principle of load
balancing. That is, the ids of each non-cluster head node
3.2.4 Connected Dominating Set Election cycles through the queue at a rate of 1 unit per run of the
load-balancing heuristic. Each node has a minimum value
The CDS [4] of a network topology is constructed in two
of 0 and a maximum value of Max_Cluster Size. Upon
steps. In the first step, if two cluster heads in the MDS are
reaching Max_Cluster Size a node will rotate to a value of 0
separated by three hops and there are no other cluster heads
on the next cluster election heuristic run. As the cluster
between them, a node with the highest priority on the
election heuristics run they will use the priorities to
shortest paths between the two cluster heads is elected as a
determine the cluster heads of the network. A cluster head
doorway, and is added to the CDS. Therefore, the addition
will maintain this value until it has exhausted its cluster
of a doorway brings the connected components in which the
head duration budget. At this point it will set its work to 0,
two cluster heads reside one hop closer. In the second step,
i.e., less than any other node, and become a normal node.
if two cluster heads or one cluster head and one doorway
node are only two hops away and there are no other cluster
heads between them, one of the nodes between them with 4. Performance Evaluation
the highest priority becomes a gateway to connect cluster We have conducted simulation experiments to evaluate the
head to cluster head or doorway to cluster head. After these performance of the proposed heuristic i.e. CMCSR. These
steps, the CDS is formed. simulation results were then compared against Topology
CDS is constructed in two steps Management by Priority Ordering TMPO [15]. We assumed
• Selecting doorway a variety of systems running with 10, 20, 40, 60, 80 and 100
• Selecting gateway nodes to simulate ad hoc networks with varying levels of
node density. Two nodes are said to have a wireless link
(a) Selecting Doorway between them if they are within communication range of
Node i can become a doorway for cluster heads n and j if the each other. Additionally, the span of a cluster, i.e. the
following conditions are satisfied. i) If cluster n and j are maximum number of wireless hops between a node and its
not two hops away. ii) There is no other cluster head m on cluster head (d) was set to 2. The entire simulation was
the shortest path between n and j. iii) There is no other node conducted in a 1150 * 1150 unit region. Initially, each node
m with higher priority than node i. was assigned a unique node id and (x, y) coordinates within
the region. The nodes were then allowed to move at random
Selecting gateway in any direction at a speed of not greater than half of the
wireless range of a node per second. The simulation range is
Node i can become a gateway for cluster head n and j if the
set to 2000 seconds, and the network was sampled for every
following conditions are satisfied.
2 seconds. At each sample time the proposed cluster size
i) If there is no cluster head or doorway between n and j.
ratio and cluster election heuristic was run to determine
ii) If there is no node with higher priority than node i.
cluster heads and their associated clusters. Each simulation
run for 2000 seconds measures several performance metrics.
3.2.5 Computing Cluster Size ratio
The main simulation metric measured was Cluster head
Duration, which provides a basis for evaluating the shaped line indicates the cluster head duration with load i..e.
performance of the proposed load-balancing heuristic. incase of CMCSR.
For the purposes of these simulations we have set the cluster
head budget to be a function of the maximum amount of
4.2 Nodes Vs Number of Cluster Heads
work it performs That is, once a node becomes a cluster
head it will remain a cluster head until it has exhausted its Figure 3 shows the graph for the average number of cluster
maximum work load, or until it loses out to another cluster heads formed during the topology management. The
head based on the rules of the cluster election heuristic. topology management is executed for 1800 seconds for each
of x nodes and the values are noted. The diamond shaped
The proposed CMCSR algorithm makes a noticeable line indicates the number of cluster head formed during
difference in the cluster head duration (ranging from 4% to topology management without load (TMPO). Second the
28%). This shows that the load-balancing heuristics topology management executed for the 600 seconds for each
generates longer cluster head durations; it will also produce x nodes and the values are noted. Totally the program is
much tighter and more deterministic responses (stability). executed for 18000 seconds. The square shaped line
These results are not surprising. Therefore, once a cluster indicates the cluster head formed during topology
head is elected it continues as cluster head for a maximum management with load (CMCSR).
of the programmed budget. This will provide the longer
cluster head durations that we see. The cluster size ratio 16
heuristic is continuously rotating, moving ordinary nodes

14
into the position of becoming a cluster head. Therefore, once
a cluster head budget is exceeded, a different cluster head is 12
elected and the process repeats. This provides the cluster
size ratio effect of distributing the responsibility of being a 10
No.of Clusters
cluster head among all nodes. We present below three
8
graphs for our simulation results. First one is the average
cluster head duration. Second one is the average number of 6
cluster head. And finally the improvement graph for the
cluster head duration. 4
2
4.1 Nodes Vs Cluster Head Duration
0
Figure 2 shows the graph for the average cluster head 10 20 30 40 50 60 70 80 90 100
duration. X-axis takes the number of nodes and y-axis No.of Nodes
shows the cluster head duration in seconds. The topology TMPO CMCSR
management is executed for 1800 seconds for each x nodes
Figure 3. Average no. of clusters
and the values are noted. Totally the program is executed for
18000 seconds. The diamond shaped line indicates the
cluster head duration without load i.e. incase of TMPO.
25
30
25
20
Cluster Head Duration(in Sec)
20
15
15
10
10
5
0
10 20 30 40 50 60 70 80 90 100
No. of Nodes
0
T MPO CMCSR
900 1800 2700 3600
Syst em Executed(Sec)
Figure 2. Average Cluster head duration Vs no. of nodes TMPO CMCSR
Second the topology management is executed for the 600 Figure 4. Average Cluster head duration
seconds for each x nodes and the values are noted. Totally
the program is executed for 18000 seconds. The square
4.3 Improvement in Cluster head Duration Sciences Research 2(6): 336-340, INSInet Publication,
2006.
Figure 4 shows the graph for the average cluster head
[4] L. Hu. “Topology Control for Multihop Packet Radio
duration. X-axis takes system executed in seconds and y-
Networks,”. IEEE Transactions on Communications,
axis shows the average cluster head duration in seconds.
41(10), Oct. 1993.
The below graph is constructed under the following
[5] S. Narayanaswamy, V. Kawadia, R. S. Sreenivas, and
conditions. Both TMPO and CMCSR is run for 900 sec,
P. R. Kumar, “Power Control in Ad-Hoc Networks:
1800 sec, 2700 sec and 3600 sec by taking total number of
Theory, Architecture, Algorithm and Implementation of
nodes into account as 50. The diamond shaped line indicates
the COMPOW Protocol,” Proceedings of the European
the cluster head duration without load and the square shaped
Wireless Conference on Next Generation Wireless
line indicates the cluster head duration with load. The
Networks: Technologies, Protocols, Services and
results related to the above three graphs indicates that
Applications, pages 156-162, Florence, Italy, Feb. 25-
CMCSR outperforms TMPO.
28, 2002.
[6] H. Takagi and L. Kleinrock, “Optimal Transmission
Ranges for Randomly Distributed Packet Radio
5. Conclusions Terminals,” IEEE Transactions on Communications,
32(3),7, Mar. 1984.
The cluster size load balancing heuristics have been [7] L. Li, V. Bahl, Y.M. Wang, and R. Wattenhofer,
proposed for ad hoc networks. The cluster election heuristics “Distributed Topology Control for Power Efficient
favor the election of cluster heads based on node willingness Operation in Multihop Wireless Ad Hoc Networks,”
and number of neighbors. Here the heuristic places a cluster Proceedings of IEEE Conference on Computer
size budget on the contiguous amount of time that a node Communications (INFOCOM), Apr. 2001.
acts as cluster head. As seen from the simulation results, [8] R. Ramanathan and R. Rosales-Hain, “Topology
this heuristic produce larger cluster head durations while Control of Multihop Wireless Networks using Transmit
decreasing the cluster size and enhancing the stability. Power Adjustment,” Proceedings of IEEE Conference
on Computer Communications (INFOCOM), IEEE,
Our proposed CMCSR is a novel energy-aware topology Mar. 26-30, 2000.
management approach based on dynamic node priorities and [9] R. Prakash, “Unidirectional Links Prove Costly in
cluster size load in ad hoc networks. CMCSR consists of two Wireless Ad-Hoc Networks,” Proceedings of the
parts that implement the MDS and CDS elections Discrete Algorithms and Methods for Mobile
respectively. Compared to five prior heuristics of MDS and Computing and Communications - DialM, Seattle, WA,
CDS elections in ad hoc networks, MDS offers four key Aug. 20, 1999.
advantages. i) CMCSR obtains the MDS and CDS of the [10] S. Bandyopadhyay and E. J. Coyle, “An Energy
network without any negotiation stage; only two-hop Efficient Hierarchical Clustering Algorithm for
neighbor information is needed. ii) CMCSR allows nodes in Wireless Sensor Networks”, In Proc. INFOCOM 2003,
the network to periodically re-compute their priorities, so as San Francisco, Apr, 2003.
to balance the cluster head role and prolong the battery life [11] M. Maeda and Ed Callaway, "Cluster Tree Protocol
of each node. iii) CMCSR introduces the willingness value (ver.0.6)",http://www.ieee802.org/15/pub/2001/May01/
of a node, which decides the probability of the node being 01189r0P80215_ TG4-Cluster-Tree-Network.pdf.
elected into the MDS according to the battery life and [12] L. Bao and J.J. Garcia-Luna-Aceves, “Transmission
mobility of the node and iv) MDS introduces doorway Scheduling in Ad Hoc Networks with Directional
concept for the CDS in addition to the well-known gateway Antennas,” Proc. ACM Eighth Annual International
and cluster head concepts. Conference on Mobile Computing and networking,
A key contribution of this work consists of converting the Atlanta, Georgia, USA, Sep, 23-28 2002.
static attributes of a node, such as node identifier, into a [13] B. Chen, K. Jamieson, H. Balakrishnan, and R.
dynamic control mechanism that incorporates the three key Morris, “Span: an Energy-Efficient Coordination
factors for topology management in ad hoc networks -- the Algorithm for Topology Maintenance in Ad Hoc
nodal battery life, mobility, and cluster size load balancing. Wireless Networks,” In Proc. 7th ACM MOBICOM,
Although existing proposals have addressed all these Rome, Italy, Jul, 2001.
aspects, CMCSR constitutes a more comprehensive [14] C.C. Chiang, H.K. Wu, W. Liu, and M. Gerla,
approach. “Routing in Clustered Multihop, Mobile Wireless
Networks with Fading Channel,” IEEE Singapore
References International Conference on Networks SICON'97, pages
[1] http://w3.antd.nist.gov/wahn_mahn.shtml. 197-211, Singapore, Apr. 14-17, 1997.
[2] Tomas Johansson and Lenka Carr-Motyˇckov´. “On [15] L. Bao and J.J. Garcia-Luna-Aceves, “Topology
Clustering in Ad Hoc Networks,” First Swedish Management in Ad Hoc Networks,” Proc of the 4th
National Computer Networking Workshop, ACM Interational Symposium on Mobile Ad Hoc
SNCNW2003, 8-10 September, 2003. Networking and Computing (MOBIHOC), Annapolis,
[3] R.Pandian, P.Seethalakshmi and V.Ramachandran, Maryland, USA, Jun. 2003.
“Enhanced Routing Protocol for Video Transmission
over Mobile Adhoc Network,” Journal of Applied
Authors Profile
Mr D K L V Chandra Mouly received M.Tech (CSE) from S V
University College of Engineering, Tirupati, India in the year
2007. Currently he is pursuing his Ph.D. (Part-time) at S V
University, Tirupati. His areas of interests are Computer Networks
and Distributed Systems.
Dr Ch D V Subba Rao received Ph.D (CSE) from S V University,

Tirupati, India in 2008. He got 18 years of teaching experience. At
present, he is working as Associate Professor, Dept of Computer
Science and Engineering, S V University College of Engineering,
Tirupati, India. His areas of interests are Distributed Systems,
Operating Systems, Computer Networks and Programming
Language Concepts.
Dr M M Naidu received Ph.D (IIT-Delhi) in the year 1988. He got

32 years of teaching experience. Currently he is working as
Professor in the Dept of Computer Science and Engineering, S V
University College of Engineering, Tirupati, India. His areas of
interests include Software Engineering, Enterprise Resource
Planning, Computer Networks and Computer Graphics.
Analysis and Proposal of a Geocast Routing

Algorithm Intended For Street Lighting System
Based On Wireless Sensor Networks
Rodrigo Palucci Pantoni1 and Dennis Brandão1
1
Eng. School of São Carlos of University of São Paulo,
São Carlos, Brazil
rodrigoppantoni@yahoo.com.br and dennis@sc.usp.br
distributed routing for retransmitting information to the

Abstract: This work is part of research project where are
studied and developed efficient technologies highlighted by the final destination. The IEEE 802.15.4 standard was chosen
ReLuz (Brazilian National Program of Public Lightning and because of its minimum consumption, low cost and protocol
Efficient Light Signalization) program. It is proposed, in that simplicity.
context, the development of remote command and monitoring The main challenge for this work is handling the limitations
infrastructure for management purposes for large areas. of the IEEE 802.15.4 with the characteristics of dense
Specifically, it refers the research and analysis of efficient networks, trying to reach a balance among guarantee of
routing algorithms in terms of energy consumption, guarantee
of package delivery and good performance (minimum delay).
delivery, performance (minimum delay) and electric power
Two new geocast routing algorithms are implemented and efficiency for the specific purpose of public lighting.
compared with other candidates applied to street lighting system.
Such algorithms refer to wireless sensor networks IEEE 2. Correlated works of public lighting system
802.15.4 through multi-hop and low reach communication, and
low cost. The results show that the proposed algorithm GGPSRII Several proposals found in the literature related to public
saves more energy, presents a good percentage of delivery lighting use PLC technology (Power Line Communication)
guarantee, good performance than compared ones. [2, 3, 4]. However, there are a few limitations from this
Keywords: Lighting system, Geocast routing algorithm, IEEE technology when it is applied in public lighting [3, 5], such
802.15.4, Wireless sensor network. as noise and impedance variations. Furthermore, Brazilian
infrastructure would have to be in perfect state to achieve a
1. Introduction good operation, but in reality the infrastructure is old and
would have to be rebuilt.
Improvements in the quality of public lighting systems have
A company [6] applied the physical and data link layer
a direct impact in the quality of life for a large part of the
standard of the model ISO/OSI IEEE 802.11 to the wireless
population as well as in the efficiency and rationality of
network protocol, which has higher electric power
using electric power. This work is part of the research
consumption and high communication rates. In this
project that intends to integrate the study and development
solution, each point of light has a network point that
of efficient technologies in the scope of the Brazilian federal
communicates and sends information to an Internet point.
government program ReLuz (National Program of Efficient
In [7], it was proposed the use of ZigBee [8] as the protocol
Public Lighting and Traffic Lights) [1]: a telecommand
for the lighting system. [9] mentioned that the use of ZigBee
system for managing public lighting in large areas. The
is not suitable for public lighting systems, although it has
expected result is the economic operation of a public
not been quantitatively proved. They suggested the use of
lighting system with an economy index superior to the
the network protocol 6LoWPAN [10] for devices IEEE
indexes currently registered in the mentioned program, due
802.15.4 [11] and a GPS. 6LoWPAN is destined to devices
to the efficiency provided by the use of high performance
with low cost, low electric power waste and low
electronic systems together with a telecommand system.
communication rates, and the main goal is to integrate the
The telecommand system is composed by devices attached to
Internet with WSN naturally. However, the use of
the points of light, which are interconnected via network,
6LoWPAN over the layers defined in IEEE 802.15.4 does
and software tools used for monitoring and control.
not define a routing algorithm by itself. Nevertheless, such
This work refers to the research and analysis for choosing
proposal naturally matches the work hereby proposed.
and implementing efficient network routing algorithms in
terms of electric power consumption, that is, besides the
3. Proposed public lighting system
rational use of the electric power, there is also the concern
about saving energy in terms of network communication. To create the routing algorithm, it is necessary to visualize
The routing algorithm refers to IEEE 802.15.4 mesh-based the entire architecture, in brief, where the requirements are
wireless sensor networks (WSN) through multi-hop displayed superficially. The requirements are: points
communication based on low range communication, low supervision (device status, whether it is connected to the
cost and minimum electric power consumption, with network or not; battery power; life time estimation for
battery and lamp; and LEDs luminosity level), control layer level.
(switch the luminosity level lamp on or off; switch on/off a In contrast to reactive algorithms, there are proactive
lamp post, a selected segment, a street, neighbourhood, city, algorithms, such as DSDV (Dynamic Destination Sequenced
etc; automatic programmed actions and freedom to actuate Distance-Vector) [14] and OLSR (Optimized Link State
the devices through a remote tool) and diagnostic and Routing) [15]. Instead of building optimized routes only
alarms (trigger an event when a network, hardware or lamp when they are necessary, proactive algorithms keep a matrix
failure occurs). of connections always updated, which reflects the current
During each device’s initialization, a GPS set attributes the network status and channels available for data transfer.
geographic coordinates of their positions. From the computational and electric power consumption
The user would be able to select a specific area on the map point of view, these algorithms are too costly, especially
besides the previously programmed area (for example, when providing mobility or when a fail occurs.
selecting a segment from a street) to actuate (switch the Geographic routing algorithms use the geographic location
lights on or off). Figure 1 shows the area selected by the user of the devices as a reference, and the location can be
through the system supervision and control tool. obtained from a GPS. The great advantage for this type of
routing is that routing tables are not necessary because the
devices decide where to forward the package according to
the smallest Euclidean distance of the destination
coordinate, for example.
Since those algorithms were designed for mobile devices,
one of the steps for this type of routing is transmitting
“hello” messages to all neighbouring devices (in radio
range), that periodically send packages with the
identification (such as the network address) of the device
and its position. So, the devices store the location of their
neighbours. They apply the greedy routing [16] by
transmitting a message to their neighbours that are
relatively closer in distance to the final destination. For any
variation of the greedy algorithm, it is important to define a
discard criterion to prevent the message from being
Figure 1. System Supervision and Control Screen transmitted uninterruptedly over the network in case the
specified destination is not located. However, in cases where
Mechanisms to automatize input information process in the it is necessary to find a balance between performance and
public lighting system must be applied, in order to become guarantee of delivery, the discard criterion must be defined
the process simplest and fast. Such mechanisms are not in
even if there is a path to the final addressee.
the scope of this paper.
To assure package delivery, greedy algorithms are
frequently used combined with recovering strategies,
4. Correlated works of routing algorithms for providing two operation modes. Such strategies are used
WSN when a package is discarded in “pure” greedy mode, in case
This section intends to review researches found in the there is an obstacle or a non-operating network device, for
literature related to routing algorithms for WSN in general, example.
independent of the application. The most prominent recovering strategy uses planar graphs.
Routing algorithms for mesh-based networks, such as Basically, the idea is to draw the network as a unique graph
AODV (Ad-Hoc On Demand Distance Vector) [12], DSR on a plane and forward the message in the direction of the
(Dynamic Source Routing) [13], and DREAM (Distance adjacent faces, which consequently forward the package to
Routing Effect Algorithm for Mobility) [13] were developed the final destination. Those strategies are extensively
to provide mobility. These algorithms are reactive, that is, studied, as in GFG (Greedy-Face-Greedy) [17], GPVFR
routes are determined by flooding through nodes searching (Greedy Path Vector Face Routing) [18], GPSR (Greedy
the addressee node when a flow of information (triggered by Perimeter Stateless Routing) [19] and GOAFR++ (Greedy
the upper layer) occurs. When the route is determined, it is Other Adaptive Face Routing plus plus) [20].
stored in the memory of the participant nodes. This In the GPSR algorithm, the recovering strategy is named
mechanism causes high energy costs, performance costs and perimeter mode and uses the right hand rule to direct the
guarantee of delivery costs. Besides, devices would have to flow of network packages through the devices. In case the
keep large routing tables in dense networks, which would be distance from the device to the destination is smaller than
impossible considering that such devices have low memory the distance to its neighbours, the algorithm returns to the
available. It is also interesting to keep the minimum possible greedy mode.
overhead in the network package, because it is limited at The term unicast means a point-to-point connection where
127 Kb by the IEEE 802.15.4 specification on the data link data is sent from a sender to a receiver. The most
appropriate type of routing in our context, however, is the messages are responsible to forward the packages; and,
so-called geocast, that also requires the devices to know finally, consultation messages are responsible to obtain the
their geographic positions via GPS. The algorithms deliver location of the addressees from one or more location
network messages to the devices in a specific geographic databases for a certain unique device’s identification (for
area, delivering a message from one device to many devices. example, the network address). Once the scenario does not
There are several routing algorithms developed for that have mobility, consultation messages and location servers
purpose, some based on flooding messages, directed can be removed. Such functionality was implemented
through “hello” messages that send the unique identification
flooding messages, and without flooding. Flooding messages
and the location. Besides, the supervision and control
algorithms find the path to the destination area the same
software requires the location of all devices on the network.
way AODV does. The first package arriving at the
Thus, the package would be sent by the system with the
destination area broadcasts to all nodes in the area. On the geographic coordinate of the destination, instead of having
other hand, directed flooding messages algorithms define only the network address and requiring the current device to
two types of areas, the destination and the routing area. An obtain its location through the location server.
example of directed flooding is the LBM (Location Based The difference between GGPSR and GGPSRII consists only
Multicast) algorithm [21] that is executed as follows: the in the trigger condition of “hello” messages. In GGPSR, the
routing area is defined as an area in the direction of the trigger is invocated according to a pre-determinated
destination area, and packages forwarded outside these two frequency (period). On the other hand, in GGPSRII, the
areas are discarded. trigger is invocated if only the data message does not reach
the destination (geocast region).Thus, it is necessary to
5. Proposed Routing Algorithms implement a confirmation message to inform the data
message forwarding failure. Table II show the GGPSRII
The algorithms proposed in this study, without flooding, are simplified pseudo-algorithm.
named GGPSR (Geocast Greedy Perimeter Stateless Routing
Protocol) and GGPSR II (Geocast Greedy Perimeter Table 1: GGPSR simplified pseudo-algorithm
Stateless Routing Protocol II). They consist of two parts: // Initialization
modified GPSR to find the destination area and geocast to For all devices
Send_Broadcast_Hello_Neighbors ();
broadcast the message to all addressee devices.
Instead of using the specific coordinate of a device as the Start Hello_Timer (period);
destination, the central point of the destination area is If Hello_Timer_Expire
calculated. The package is then forwarded to this point and, Send_Broadcast_Hello_Neighbors ();
when it gets to the destination area, the first device //Send
receiving the message broadcasts it to all devices in the area. If (Packet.Destination_Position != myPosition &&
myPosition = = UNICAST)
As soon as a device receives the broadcasted message, the ModifiedGPSR_Forward (Packet);
device checks if it has already received this message, Else If (Packet.Destination_Position = = myPosition
&& myPosition = = GEOCAST){
checking a sequential number and therefore avoiding If (Packet.seqN_ < ReceivedSeqNo){
Broadcast_Neighborhood_Geocast_Region(Packet);
unnecessary retransmissions. In case the device has not }
received the message yet but it is in the destination area, the }
device receives and rebroadcasts the message to the network. //Receive
Table I shows the pseudo-algorithm GGPSR in a simplified If (Packet.Destination_Position = = myPosition &&
way. ModifiedGPSR_Receive (Packet);
Considering the GPSR part of the proposed algorithm Else If (Packet.Destination_Position = = myPosition
&& myPosition = = GEOCAST){
GGPSR (only), this work suggests some modifications for If (Packet.seqNo¬ _ < ReceivedSeqNo){
lighting systems applications. The first modification is ModifiedGPSR_Receive (Packet);
}
related to the “hello” messages. For the discussed }
application, devices are fixed. Initially, it was assumed that
only one “hello” message when the device is initialized Regarding the destination area, it can have the shape of a
would be enough. However, it is interesting to keep this four-vertex polygon, circle and point (in this last case, the
functionality but with a periodicity much longer than what it communication is unicast).
is used in mobile devices. The reason to still keep this The geographic position is represented through geodesic
periodicity is that the device can simply stop operating coordinates (latitude and longitude). Each coordinate is
because of a permanent or temporary failure caused by an allocated as floating types, which in language C has 4 bytes
obstruction. Information about the neighbors from the and precision of seven decimal places. In relation to the
devices can affect the network reliability because each value ranges, this size is more than enough: the field
device is also a message router. "hours" of the latitude coordinate varies between -180 and
The second modification is related to storing geographic 180, whereas longitude varies between -90 and 90.
positions for the neighbors (applied to GGPSR and Table III shows the header struct in C language of the
GGPSRII). Originally, GPSR has three types of messages: packet types, including “hello” messages (hdr_gpsr_hello),
“hello” messages, consultation to destination locations and and data messages (hdr_gpsr_data) of the proposed protocol.
data messages [19]. “Hello” messages are responsible to
inform the new device’s location to its neighbors; data
Table 2: GGPSRII simplified pseudo-algorithm 6. Simulation and results

// Initialization
For all devices The simulation was analyzed to obtain quantitative data in
order to decide which routing algorithm would be more
//Send appropriate to the problem in discussion. First, the ZigBee
If (Packet.Source_Position = = myPosition &&
myPosition = = UNICAST){ protocol (which uses AODV routing algorithm) was
ModifiedGPSR_Forward (Packet); evaluated because it is a consolidated standard that provides
Start Confirmation_Timer (period);
} interoperability among several manufacturers and reduced
Else If (Packet.Destination_Position != myPosition costs for production. For this reason, the AODV protocol is
&& myPosition = = UNICAST)
ModifiedGPSR_Forward (Packet); simulated and compared to the GPSR, GGSR, GGPSRII and
Else If (Packet.Destination_Position = = LBM protocols.
myPosition && myPosition = = GEOCAST){
If (Packet.seqNo_ < ReceivedSeqNo){ Simulations were performed on a largely used simulator for
Broadcast_Neighborhood_Geocast_Region(Packet) the academic area, the ns-2.33 [22]. Originally, in relation
;
} to routing algorithms, ns-2.33 only implemented the AODV
} algorithm. Hence, the authors implemented GPSR, LBM,
//Receive GGPSR and GGPSRII protocols. Configurations for
If (Packet.Destination_Position = = myPosition && simulating all routing protocols are presented in Table IV.
ModifiedGPSR_Receive (Packet); Devices are equally distributed vertically and horizontally
Else If (Packet.Destination_Position = = along 50 meters to simulate lamp posts in a simplified way.
myPosition && myPosition = = GEOCAST){
If (Packet.seqNo_ < ReceivedSeqNo){ Figure 2 shows the location of the devices. Data always
ModifiedGPSR_Receive (Packet); flows from device 98 to all nodes on the “last line” (devices
}
} 0, 1, 2, 3, 4, 5, 6, 7, 8 and 9). So, device 98 is the network
//Any Time coordinator. In addition, the ns2 oTcl code was
If Confirmation _Timer_Expire implemented to keep the energy of device 98 always as its
initial energy.
Related to the “hello” packet, the field “type_” refers to the Table 4: Configurations for simulation
packet type (whether it is “hello” or data packet). The field Network Interface Phy/WirelessPhy/802_15_4
“x_” and “y_” are the source device geodesic coordinates. MAC Mac/802_15_4
The field “seqNo” is used to control the receiving and IFQ Queue/DropTail/PriQueue
rebroadcast actions, i. e., a device must have to broadcast Link Layer LL
just a once “hello” messages to its neighbours. Antena Antenna/OmniAntenna
Related to the data packet, the fields “sx_” and “sy_” are Dimension X 170
the source device geodesic coordinates. The field “ts_” is the Dimension Y 270
timestamp, used for calculate latency. The fields IFQLEN 50
“sx_GF_Failed” and “sy_GF_Failed_” correspond to the Propagation Propagation/TwoRayGround
coordinates where the greedy mode have failed, that are Phy/WirelessPhy Pt_ 7.214e-3 (100m)
Number of Devices 100
used for the packet whether it can be return to that mode in
Duration 1000 simulation time
case it is in the perimeter mode. The field “seqNo” is used
Transmission Power 0.3 mW
for control the receiving and rebroadcast actions on the
Reception Power 0.28 mW
geocast region. The rest of variables are the polygon 1 Joule
Initial Energy
coordinates’. Packet size (less 64 bytes
header)
Table 3: GGPSR and GGPSRII header packets Flow CBR (Constant Bit Rate)
struct hdr_gpsr_hello {
u_int8_t type_;
float x_; Simulation foresees the basic operation situation: network
float y_;
int seqNo_; traffic is requested every twelve hours (assuming that the
}; unit of simulation is in hours), that is, switch the lights on
struct hdr_gpsr_data {
u_int8_t type_; and off for a street segment, for example.
u_int8_t mode_;
float sx_; GPSR and GGPSR “Hello” messages periodicity were
float sy_; configured to 12 hours, that is the time to send data.
float ts_;
float sx_GF_Failed_; It is important to emphasize that in case of unicast
float sy_GF_Failed_; algorithms, ten messages are sent from device 98 to all
float dst_x1; devices in the “last line”, as mentioned before. In case of
float dst_y1;
float dst_x2; geocast algorithms, only one message is sent to all
float dst_y2; addressees.
float dst_x3;
float dst_y3; Figure 3 shows a comparison of the electric power of all
float dst_x4; devices in the network summed during the time interval. It
float dst_y4;
int seqNo_; was verified that the AODV protocol is the least efficient,
}; and GGPSRII is the most efficient. It can be concluded that
the use of ZigBee protocol is strongly not recommended.
Figure 5. Delay average (performance)
Table V shows the packet header comparison. GGPSRII

presents lesser size of bytes then “LBM – request”, which is
the largest header in terms of bytes.
TABLE 5: Packet header Comparison

Figure 2. Simulated scenario Packet Header Size (bytes)
AODV - request 28
Figure 4 shows the rate in percentage of throughput to each AODV - response 26
evaluated protocol. For the AODV traffic, it was necessary to AODV - error 14
configure the beginning of the information flow in a non- AODV - confirmation 4
simultaneous way, with a short period of time between LBM - request 62
transmissions. If all messages were sent to the ten devices LBM - response 52
simultaneously, no message would be successfully received. LBM - error 14
The scheduling configuration was also applied to the GPSR LBM - confirmation 4
routing for comparison, even though GPSR does not need GPSR – hello 12
this artifice. GPSR - data 34
Observe that because of the AODV flooding characteristic, GGPSRII - hello 12
there is a very significant lost of messages that reassures GGPSRII - data 58
discredit on the ZigBee protocol. For this same reason, it can
be verified in Figure 5 that there is a delay average much
higher than the AODV algorithm. 7. Conclusions and future works
The proposed GGPSRII routing algorithm proved to be
efficient in terms of energy, and provides a good balance in
terms of performance and guarantee of delivery.
Quantitative results showed that the research has a solid
base, that is, resources can be invested in hardware
prototypes for devices (an alternative to ZigBee) and in
implementing the supervision and control software.
Later on, this algorithm will be implemented to transmit a
message to multiple areas, using the Fermat point concept
proposed by [23].
Final validation tests for the protocol will be simulated in
real scenarios obtained from lamp posts mapping, usually
archived in city halls.
The authors will also develop an alarm and network
Figure 3. Comparing the network total energy diagnostic mechanism for the lighting system based on the
protocol of the proposed routing. Further on, application
layer services will be included to switch the lights, trigger
alarms, provide diagnostics, etc.
Acknowledgment
The authors gratefully acknowledge the academic support and
research structure from the Engineering School of São Carlos
- University of São Paulo. The authors also acknowledge the
important technical contributions from Smar International
Corporation and the Prof. Tracy Camp for helping in provide
Figure 4. Delivery rate an implementation of LBM algorithm, which was very
helpful for an implementation of the LBM implementation of
this work.
References [18] Leong, B.; Mitra, S.; Liskov. B. “Path vector face
routing: Geographic routing with local face information”.
[1] Reluz Program (Jan. 2010). “National Program of In Proceedings of the IEEE Conference on Network
Efficient Public Lighting and Traffic Lights”. Available: Protocols, 2005.
http://www.eletrobras.gov.br/ EM_Programas_Reluz [19] Karp, B; Kung, H. T. (2000) GPSR: greedy perimeter
[2] Ei-Shirbeeny, E.H.T.; Bakka, M.E.. “Experimental pilot stateless routing for wireless networks. in Proceedings of
project for automating street lighting system in Abu Dhabi the 6th ACM/IEEE MobiCom. 2000, pp. 243-254, ACM
using powerline communication”. Proceedings of the 10th Press.
IEEE International Conference on Electronics, Circuits [20] Kuhn,F.;Wattenhofer,R.;Zhang,Y.;Zollinger, A.
and Systems. Vol. 2, p.743 – 746, Dec. 2003 “Geometric ad-hoc routing: of theory and practice”. in
[3] Chueiri, I.J.; Bianchim, C.G. “Sistema de Comando e Proceedings of the 22nd annual symposium on principles
Controle de Potência em Grupo para Iluminação Pública”. of distributed computing, 2003.
BR n.PI0201334-7, 2002. [21] Ko, Y.; Vaidya, N.H.”Geocasting in mobile ahoc
[4] Sungkwan C.; Dhingra, V. “Street lighting control networks: Location-based multicast algnrithms” .In
based on LonWorks power line communication Power Proceedings of WMCSA, pages 101-110, 1999.
Line Communications and Its Applications”. IEEE [22] Network Simulator NS2 (Jan. 2010). Available:
Symposium, 2008. http://www.isi.edu/nsnam/ns
[5] Sutterlin, P.; Downey, W. (1999). “A power line [23] Lee,S.;Ko,Y. “Geometry-driven Scheme for Geocast
communication tutorial – challenges and technologies”, Routing in Mobile Ad Hoc Networks”. The 2006 IEEE
Technical Report, Echelon Corporation, 1999. 63rd Vehicular Technology Conference (VTC),
[6] Streetlight Intelligence (Jan. 2010). Available: Melbourne, Australia, 2006.
http://www.streetlightiq.com
[7] Barriquello, C.H. ; Garcia, J.M. ; Corrêa, C. ; Menezes,
C.V. ; Campos, A. ; Do Prado, R.N. “Sistema Inteligente Authors Profile
Baseado em Zigbee para Iluminação Pública com
Lâmpadas de LEDS”. In: XVII Congresso Brasileiro de Rodrigo Palucci Pantoni R&D Systems
Automática. Anais do XVII CBA. Juiz de Fora, 2008. Analyst, received the Computer Science
degree in 2000 and subsequently received
[8] Zigbee. ZigBee PRO Specification, ZigBee Alliance.
the M.S. in 2006 at the University of São
2007. Paulo (USP). He's attending the Ph.D
[9] Denardin, G.W.; Barriquello, C.H.; Campos, A. Do course, in the same university, as part of his
Prado, R.N. “An Intelligent System for Street Lighting job at the Smar R&D department in the area
Monitoring and Control”. 10° Congresso Brasileiro de of software development for automation control and fieldbuses. He
Eletrônica de Potência. Brasil, Bonito, 2009. joined Smar in 2000, working
[10] Kushalnagar, N.; Montenegro, G.; Schumacher, C.. in the Smar R&D department where he conducts research and
“IPv6 over Low-Power Wireless Personal Area Networks development of host systems, including a Fieldbus Foundation
(6LoWPANs): Overview, Assumptions, Problem Asset Management and a Configurator system. He now teaches
Statement, and Goals”. Request for Comments: 4919, computer networks at Information Systems course at University
2007. Dr. Francisco Maeda.
[11] IEEE 802.15.4. “Wireless Medium Access Control
(MAC) and Physical Layer (PHY) Specifications for Low-
Rate Wireless Personal Area Networks”, IEEE Computer Dennis Brandão He received his Ph.D. degree in mechanical
Society, 2006. engineering at the University of São Paulo in
2005. He now teaches “Industrial
[12] Perkins, C. E.; Belding-Royer, E. M.; Das, S. R.. “Ad
Automation” at the Department of Electrical
Hoc On-Demand Distance Vector Routing”, Request for
Engineering of the same university. His
Comments: 3561, 2003. research activities are mainly in the area of
[13] Basagni, S.; Chlamtac, I.; Syrotiuk, V.R.; Woodward, fieldbus technology and application, with a
B.A. “A distance routing effect algorithm for mobility particular interest for distributed systems
(dream)”. In Proceedings of ACM/IEEE MobiCom ’98, and continuous process control.
1998.
[14] Perkins, C.; Bhagwat, P. “Highly Dynamic Destination
Sequenced Distance-Vector Routing for Mobile
Computers”. Comp. Commun. 1994.
[15] Clausen, T.; Jacquet, P. “Optimized Link State Routing
Protocol”. Request for Comments: 3626, 2003.
[16] Finn, G.G. “Routing and addressing problems in large
metropolitan scale internetworks”. Technical Report
ISI/RR-87-180, ISI, 1987.
[17] Bose, P.; Morin, P., Stojmenovic, I.; Urrutia, J.”Routing
with guaranteed delivery in ad hoc wireless networks”. In:
Proceedings of the 3rd International Workshop on Discrete
algorithms and methods for mobile computing and
communications. 1999, ACM Press.
Routing protocols in wireless networks

1
Vineet Agrawal , 2Dr Yashpal Singh,3Manish Varshney ,4Vidushi Gupta
1
Reader Deptt of Computer Science & Engg RBCET Bareilly, India
Email vineetkaga@rediffmail.com
2
Reader Deptt of Computer Science & Engg BIET Jhansi, India
Email yash_biet@yahoo.com
3
Sr Lecturer Deptt of Computer Science & Engg, SRMSWCET Bareilly, India
Email itsmanishvarshney@gmail.com
4
Lecturer Deptt of Computer Science & Engg, SRMSWCET Bareilly, India
Email vidu.leo@gmail.com
Abstract: An ad hoc mobile network is a group of mobile nodes

that are dynamically and arbitrarily located in such a manner
that the interconnections between nodes are capable of changing
on a repetitive basis. In order to facilitate communication within
the network, a routing protocol is used to discover routes
between nodes. The primary goal of such an ad hoc network
routing protocol is correct and efficient route establishment
between a pair of nodes so that messages may be delivered in a Figure 1. A simple ad hoc network of three wireless mobile
timely manner. In this paper we examine routing protocols for hosts
ad hoc networks and evaluate these protocols based on a given
set of parameters. The scope of paper was to test routing The bridges for these networks are known as base stations.
performance of four different routing protocols (AODV, DSR, A mobile unit within these networks connects to, and
DSDV,etc) in variable network sizes up to thousand nodes. communicates with, the nearest base station that is within
Various types of scenarios are generated and each of the its communication radius. As the mobile travels out of range
protocol is simulated on each of these, then their parameters like of one base station and into the range of another, a
throughput, packet delivery ratio and delay will be compared “handoff” occurs from the old base station to the new, and
The performance differentials are analyzed using varying pause the mobile is able to continue communication seamlessly
time, constant nodes and dynamic topology. Based on the
throughout the network. Typical applications of this type of
observations, we make valuable conclusions about which
network include office wireless local area networks
protocol performs better in which condition.
(WLANs). The second type of mobile wireless network is
the infrastructure less mobile network, commonly known as
Keywords: Ad hoc ,Table-Driven, Demand Driven AODV,
DSDV,DSR, , MANET. an ad hoc network. Infrastructure less networks has no fixed
routers; all nodes are capable of movement and can be
connected dynamically in an arbitrary manner. Nodes of
1. Introduction
these networks function as routers which discover and
maintain routes to other nodes in the network. Example
Mobile hosts and wireless networking hardware are
applications of ad hoc networks are emergency search-and-
becoming widely available, and extensive work has been
rescue operations, meetings or conventions in which persons
done recently in integrating these elements into
wish to quickly share information, and data acquisition
conventional networks such as the Internet. Oftentimes,
operations in inhospitable terrain. A community of ad hoc
however, mobile users will want to communicate in
network researchers has proposed, implemented, and
situations in which no fixed wired infrastructure such as this
measured a variety of routing algorithms for such networks.
is available, either because it may not be economically
The observation that topology changes more rapidly on a
practical or physically possible to provide the necessary
mobile, wireless network than on wired networks, where the
infrastructure or because the expediency of the situation
use of Distance Vector (DV), Link State (LS), and Path
does not permit its installation. In networks comprised
Vector routing algorithms is well established, motivates this
entirely of wireless stations, communication between source
body of work . DV and LS algorithms require continual
and destination nodes may require traversal of multiple
distribution of a current map of the entire network’s
hops, as radio ranges are finite. Since Routing Protocols
topology to all routers. DV’s Bellman- Ford approach
emergence in the 1970s, wireless networks have become
constructs this global picture transitively; each router
increasingly popular in the computing industry. This is
includes its distance from all network destinations in each of
particularly true within the past decade, which has seen
its periodic beacons. LS’s Dijkstra approach directly floods
wireless networks being adapted to enable mobility. There
announcements of the change in any link’s status to every
are currently two variations of mobile wireless networks.
router in the network. Small inaccuracies in the state at a
The first is known as the infrastructure network (i.e., a
router under both DV and LS can cause routing loops or
network with fixed and wired gateways).
disconnection [7,9]. When the topology is in constant flux,
as under mobility, LS generates torrents of link status
change messages, and DV either suffers from out-of-date

state [4], or generates torrents of triggered updates. 2.1 Table-Driven Routing Protocols
The two dominant factors in the scaling of a routing Table-driven routing protocols attempt to maintain reliable,
algorithm are: The rate of change of the topology and The up-to-date routing information from each node to every
number of routers in the routing domain. Both factors affect other node in the network. These protocols require each
the message complexity of DV and LS routing algorithms: node to maintain one or more tables to store routing
intuitively, pushing current state globally costs packets information, and they react to changes in network topology
proportional to the product of the rate of state change and by propagating updates throughout the network in order to
number of destinations for the updated state. Hierarchy is maintain a consistent network view. The areas in which they
the most widely deployed approach to scale routing as the differ are the number of necessary routing-related tables and
number of network destinations increases. Without the methods by which changes in network structure are
hierarchy, Internet routing could not scale to support today’s broadcast. The following sections discuss some of the
number of Internet leaf networks. An Autonomous System existing table-driven ad hoc routing protocols.
runs an intra-domain routing protocol inside its borders, and
appears as a single entity in the backbone inter-domain 2.1.1 Destination-Sequenced Distance-Vector Routing
routing protocol, BGP. This hierarchy is based on well- The Destination-Sequenced Distance-Vector Routing
defined and rarely changing administrative and topological protocol (DSDV) described in [2] is a table-driven algorithm
boundaries. It is therefore not easily applicable to freely based on the classical Bellman-Ford routing mechanism [3].
moving ad-hoc wireless networks, where topology has no The developments made to the Bellman-Ford algorithm
well-defined AS boundaries, and routers may have no include freedom from loops in routing tables.
common administrative authority. Caching has come to
prominence as a strategy for scaling ad-hoc routing
protocols. Instead, routers running these protocols request
topological information in an on-demand fashion as
required by their packet forwarding load, and cache it
aggressively. When their cached topological information
becomes out-of date, these routers must obtain more current
topological information to continue routing successfully.
Because of the fact that it may be necessary to hop several Figure 2.Categorization of ad hoc routing protocols
hops (multi-hop) before a packet reaches the destination, a
routing protocol is needed. The routing protocol has two DSDV [2] is a hop-by-hop distance vector routing protocol
main functions, selection of routes for various source- that in each node has a routing table that for all available
destination pairs and the delivery of messages to their destinations stores the next-hop and number of hops for that
correct destination. The second function is conceptually destination Every mobile node in the network maintains a
straightforward using a variety of protocols and data routing table in which all of the possible destinations within
structures (routing tables). This paper is focused on the network and the number of hops to each destination are
selecting and finding routes. recorded. Each entry is marked with a sequence number
This paper examines routing protocols designed for these assigned by the destination node. The sequence numbers
wireless networks by first describing the operation of each of enable the mobile nodes to distinguish stale routes from new
the protocols and then comparing their various ones, thereby avoiding the formation of routing loops.
characteristics. The remainder of the paper is organized as Routing table updates are periodically transmitted
follows. The next section presents a discussion of two throughout the network in order to maintain table
subdivisions of ad hoc routing protocols. Another section consistency. DSDV basically is distance vector with small
discusses current table-driven protocols, while a later section adjustments to make it better suited for ad-hoc networks.
describes those protocols which are classified as on-demand. These modifications consist of triggered updates that will
The paper then presents Simulation parameters and the take care of topology changes in the time between
performance evaluation including, a general comparison of broadcasts. To reduce the amount of information in these
table-driven and on-demand protocols. Finally, the last packets there are two types of update messages defined: full
section concludes the paper. and incremental dump. The full dump carries all available
routing information and the incremental dump that only
2. Ad Hoc Routing Protocols carries the information that has changed since the last
Since the advent of Defense Advanced Research Projects dump. Because DSDV is dependent on periodic broadcasts it
Agency (DARPA) packet radio networks in the early 1970s needs some time to converge before a route can be used.
[1], several protocols have been developed for ad hoc mobile This converge time can probably be considered negligible in
networks. Such protocols must deal with the typical a static wired network, where the topology is not changing
limitations of these networks, which include high power so frequently. In an ad-hoc network on the other hand,
consumption, low bandwidth, and high error rates. As where the topology is expected to be very dynamic, this
shown in Fig.2, these routing protocols may generally be converge time will probably mean a lot of dropped packets
categorized as: before a valid route is detected. The periodic broadcasts also
• Table-driven(proactive) add a large amount of overhead into the network. New route
• Source-initiated (demand-driven)(reactive) broadcasts contain the address of the destination, the
number of hops to reach the destination, the sequence must assume that connectivity with that neighbor has been
number of the information received regarding the lost. When a mobile receives a hello message from a new
destination, as well as a new sequence number unique to the node, that new node is added to the mobile’s routing table,
broadcast [2].. and the mobile sends the new node a copy of its routing
table information. Part of the novelty of WRP stems from
2.1.2The Wireless Routing Protocol the way in which it achieves loop freedom. In WRP, routing
The Wireless Routing Protocol (WRP) described in [5] is a nodes communicate the distance and second-to-last hop
table-based protocol with the goal of maintaining routing information for each destination in the wireless networks.
information among all nodes in the network. To describe WRP belongs to the class of path-finding algorithms with an
WRP, we model a network as an undirected graph important exception. It avoids the “count-to-infinity”
represented as G.V; E. , where V is the set of nodes and E is problem [6] by forcing each node to perform consistency
the set of links (or edges) connecting the nodes. Each node checks of predecessor information reported by all its
represents a router and is a computing unit involving a neighbors. This ultimately (although not instantaneously)
processor, local memory and input and output queues with eliminates looping situations and provides faster route
unlimited capacity. In a wireless network, a node has radio convergence when a link failure event occurs.
connectivity with multiple nodes and a single physical radio
link connects a node with many other nodes. Each node in 2.2 Source-Initiated On-Demand Routing
the network is responsible for maintaining four tables:
• Distance table A different approach from table-driven routing is source-initiated
• Routing table on-demand routing. This type of routing creates routes only when
• Link-cost table desired by the source node. When a node requires a route to
• Message retransmission list (MRL) table a destination, it initiates a route discovery process within the
Each entry of the MRL of the update Message, network. This process is completed once a route is found or
a retransmission counter, an acknowledgment- required flag all possible route permutations have been examined. Once a
vector with one entry per neighbor, and a list of updates sent route has been established, it is maintained by a route
in the update message. A link is assumed to exist between maintenance procedure until either the destination becomes
two nodes only if there is radio connectivity between the two inaccessible along every path from the source or until the
nodes and they can exchange update messages reliably with route is no longer desired.
a certain probability of success The MRL records which
updates in an update message need to be retransmitted and 2.2.1 Ad Hoc On-Demand Distance Vector Routing
which neighbors should acknowledge the retransmission [5]. (AODV)
Mobiles inform each other of link changes through the use The Ad Hoc On-Demand Distance Vector routing protocol
of update messages. An update message is sent only between (AODV) is an improvement of the Destination-Sequenced
neighboring nodes and contains a list of updates (the Distance Vector routing protocol (DSDV)1. DSDV has its
destination, the distance to the destination, and the efficiency in creating smaller ad-hoc networks. Since it
predecessor of the destination), as well as a list of responses requires periodic advertisement and global dissemination of
indicating which mobiles should acknowledge (ACK) the connectivity information for correct operation, it leads to
update. Mobiles send update messages after processing recurrent system-wide broadcasts. Therefore the size of
updates from neighbors or detecting a change in a link to a DSDV ad-hoc networks is strongly limited. When using
neighbor. In the event of the loss of a link between two DSDV, every mobile node also needs to maintain a whole
nodes, the nodes send update messages to their neighbors. list of routes for each destination within the mobile network.
The neighbors then modify their distance table entries and The advantage of AODV is that it tries to reduce the number
check for new possible paths through other nodes. Nodes of required broadcasts. It creates the routes on an on-
learn of the existence of their neighbors from the receipt of demand basis, as opposed to maintain a complete list of
acknowledgments and other messages. If a node is not routes for each destination. Therefore, the authors of AODV
sending messages, it must send a hello message within a classify it as a pure on-demand route acquisition system [3].
specified time period to ensure connectivity. Otherwise, the
lack of messages from the node indicates the failure of that 2.2.1.1 Path Discovery Process
link; this may cause a false alarm. Because of the broadcast When trying to send a message to a destination node
nature of the radio channel, a node can send a single update without knowing an active route2 to it, the sending node
message to inform all its neighbors about changes in its will initiate a path discovery process. A route request
routing table; however, each such neighbor sends an ACK to message (RREQ) is broadcasted to all neighbors, which
the originator node. To ensure that connectivity with a persist to broadcast the message to their neighbors and so
neighbor still exists when there are no recent transmissions on. The forwarding process is continued until the
of routing table updates or ACKs, periodic update messages destination node is reached or until an intermediate node
without any routing table changes (null update messages) knows a route to the destination that is new enough. To
are sent to the neighbors. The time interval between two ensure loop-free and most recent route information, every
such null update messages is the HelloInterval. If a node node maintains two counters: sequence number and
fails to receive any type of message from a neighbor for a broadcast_id. The broadcast_id and the address of the
specified amount of time (e.g., three or four times the source node uniquely identify a RREQ message.
HelloInterval known as the Router Dead-Interval), the node broadcast_id is incremented for every RREQ the source
node initiates. An intermediate node can receive multiple discover a route across multiple network hops to any
copies of the same route request broadcast from various destination. DSR uses no periodic routing messages, thereby
neighbors. In this case – if a node has already received a reducing network bandwidth overhead, conserving battery
RREQ with the same source address and broadcast_id – it power and avoiding large routing updates throughout the
will discard the packet without broadcasting it furthermore. ad-hoc network. Instead DSR relies on support from the
When an intermediate node forwards the RREQ message, it MAC layer (the MAC layer should inform the routing
records the address of the neighbor from which it received protocol about link failures). The two basic modes of
the first copy of the broadcast packet. This way, the reverse operation in DSR are route discovery and route
path from all nodes back to the source is being built maintenance.
automatically. The RREQ packet contains two sequence
numbers: the source sequence number and the last
destination sequence number known to the source. The
source sequence number is used to maintain “freshness”
information about the reverse route to the source while the
destination sequence number specifies what actuality a route
to the destination must have before it is accepted by the
source. [3] When the route request broadcast reaches the
destination or an intermediate node with a fresh enough
route, the node responds by sending a unicast route reply
Figure 4. AODV Route Maintenance by using Link
packet (RREP) back to the node from which it received the
failure Notification Message
RREQ. So actually the packet is sent back reverse the path
built during broadcast forwarding. A route is considered 2.2.2.1Route Discovery
fresh enough, if the intermediate node’s route to the Route discovery allows any host in the ad hoc network to
destination node has a destination sequence number which dynamically find out a route to any other host in the ad hoc
is equal or greater than the one contained in the RREQ network, whether directly reachable within wireless
packet. As the RREP is sent back to the source, every transmission range or reachable through one or more
intermediate node along this path adds a forward route entry intermediate network hops through other hosts. A host
to its routing table. The forward route is set active for some initiating a route discovery broadcasts a route request packet
time indicated by a route timer entry. The default value is which may be received by those hosts within wireless
3000 milliseconds, as referred in the AODV RFC [4]. If the transmission range of it. The route request packet identifies
route is no longer used, it will be deleted after the specified the host, refer red to as the target of the route discovery, for
amount of time. Since the RREP packet is always sent back which the route is requested. If the route discovery is
the reverse path established by the routing request, AODV successful the initiating host receives a route reply packet
only supports symmetric links. listing a sequence of network hops through which it may
reach the target. In addition to the address of the original
initiator of the request and the target of the request, each
route request packet contains a route record, in which is
accumulated a record of the sequence of hops taken by the
route request packet as it is propagated through the ad hoc
network during this route discovery. Each route request
packet also contains a unique request id, set by the initiator
from a locally-maintained sequence number. In order to
detect each
duplicate route requests received, host in the ad hoc network
maintains a list of the h initiator address, request id i pairs
that it has recently received on any route request.
2.2.2.2 Route Maintenance

Route maintenance can be accomplished by two different
processes:
Figure.3. AODV Path Discovery Process. • Hop-by-hop acknowledgement at the data link layer
• End-to-end acknowledgements
2.2.2 Dynamic Source Routing Hop-by-hop acknowledgement at the data link layer allows
The Dynamic Source Routing (DSR) protocol presented in an early detection and retransmission of lost or corrupt
[8] is an on-demand routing protocol that is based on the packets. If the data link layer determines a fatal
concept of source routing. The Dynamic Source Routing transmission error (for example, because the maximum
(DSR) protocol is an on-demand routing protocol based on number of retransmissions is exceeded), a route error packet
source routing. In the source routing technique, a sender is being sent back to the sender of the packet. The route
determines the correct sequence of nodes through Dynamic error packet contains two parts of information: The address
Source Routing (DSR) [3][12][13] also belongs to the class of the node detecting the error and the host’s address which
of reactive protocols and allows nodes to dynamically it was trying to transmit the packet to. Whenever a node
receives a route error packet, the hop in error is removed speed distributed uniformly between 0 and some maximum
from the route cache and all routes containing this hop are speed. Upon reaching the destination, the node pauses again
truncated at that point. End-to-end acknowledgement may for pause time seconds, selects another destination, and
be used, if wireless transmission between two hosts does not proceeds there as previously described, repeating this
work equally well in both directions. As long as a route behavior for the duration of the simulation. Each simulation
exists by which the two end hosts are able to communicate, ran for 200 seconds of simulated time. We ran our
route maintenance is possible. There may be different routes simulations with movement patterns generated for 9
in both directions. In this case, replies or acknowledgements different pause times: 2, 10, 15, 25, 35, 50, 75, 85, 100
on the application or transport layer may be used to indicate seconds. A pause time of 0 seconds corresponds to
the status of the route from one host to the other. However, continuous motion, and a pause time of 200 (the length of
with end-to-end acknowledgement it is not possible to find the simulation) corresponds to no motion. Hence reducing
out the hop which has been in error. pause time increases mobility. In this way we put our
protocols in networks of varying mobility. Because the
3 .Simulation And Its Parameters performance of the protocols is very sensitive to movement
pattern, we generated scenario files with 9 different pause
3.1 Methodology times. All routing protocols were run on the same 9 scenario
The main concentration of the project was to test the ability files. We report in this paper data from simulations using a
of different routing protocols to respond on network maximum node speed of 20 meters per second (average
topology changes (for instance link breaks, node movement, speed 10 meters per second).
and so on). Furthermore the focus was set on different
network sizes, varying number of nodes and area sizes. Our 3.3 Communication Model
investigations did not include the protocol’s operation under As the purpose of our simulation was to compare the
heavy load, e.g. its operation in congestion situations. performance of each routing protocol, we select our traffic
Therefore only rather small packet sizes and one source sources to be constant bit rate (CBR) sources. When
node were selected. As referenced in many other papers, defining the parameters of the communication model, we
Our protocol evaluations are based on the simulation of 50 experimented with sending rates of 3 packets per second,
wireless nodes forming an ad hoc network, moving about networks containing maximum connection of 35, and packet
over a rectangular (1500m X 300m) flat space for 200 sizes of 512 bytes. All communication patterns were peer-to-
seconds of simulated time. We chose a rectangular space in peer, and connections were started at times uniformly
order to force the use of longer routes between nodes than distributed between 0 and 180 seconds. The 9 different
would occur in a square space with equal node density. In scenario files for maximum node movement speed (20 m/s)
order to enable direct, fair comparisons between the moving in a random waypoint model with which we
protocols, it was critical to challenge the protocols with compared the routing protocols.
identical loads and environmental conditions. Each run of
the simulator accepts as input a scenario file that describes 3.4 Performance Metrics
the exact motion of each node and the exact sequence of In order to compare routing protocols, the following
packets originated by each node, together with the exact performance metrics are considered:
time at which each change in motion or packet origination • Throughput: a dimensional parameter which gives the
is to occur. We pre-generated 9 different scenario files with portion of the channel capacity used for useful transmission
varying movement patterns and traffic loads, and then ran selects a destination at the beginning of the simulation and
all three routing protocols against each of these scenario (i.e., data packets correctly delivered to the destinations).
files. Since each protocol was challenged in an identical • Average End to End delay: the average end-to-end delay
fashion, we can directly compare the performance results of of data packets, i.e. the period between the data packet
the protocols generation time and the time when the last bit arrives at the
destination.
3.2 Mobility Model •Packet delivery ratio: the ratio among the number of
An important factor in mobile ad-hoc networks is the packets received by the TCP descends at the final
movement of nodes, which is characterized by speed, destination and the number of packets originated by the
direction and rate of change. Mobility in the “physical “application layer” sources. It is a measure of efficiency of
world” is unpredictable, often unrepeatable, and it has a the protocol
dramatic effect on the protocols developed to support node
movement. Therefore, different “synthetic” types of mobility 4. Performances Analysis
models have been proposed to simulate new protocols.
Synthetic means to realistically represent node movement, DSDV which is a table driven proactive routing protocol
but without using network traces. Nodes in the simulation completely prevails over the on demand reactive routing
move according to a model that we call the “random protocols AODV and DSR .Since DSDV proactively
waypoint” model. The movement scenario files we used for maintains the routes to all destination in its table it does not
each simulation are characterized by a pause time. Each have to initiate the route request process as frequently as in
node begins the simulation by remaining stationary for AODV and DSR while sending packets. Hence on average
pause time seconds. It then selects a random destination in DSDV clearly has less delay. Now we can easily examine
the 1500m x 300m space and moves to that destination at a
that DSR is the worst protocol in terms of delay. At high packets, with no regard to when and how frequently such
mobility and more network load (512 byte packets at 3 routes are preferred. It relies on an underlying routing table
packets/sec) insistent route caching strategy of DSR fails. In revise mechanism that involves the stable propagation of
these stressful condition links break very often leading to routing information. This is not the case, however, for on-
invalidation of routes cached .Hence in these conditions, demand routing protocols. When a node using an
picking up of staled cached routes occur leading to
utilization of additional network bandwidth and interface
queue slots even though the packet is ultimately dropped,
leading to more delay DSR performed inefficiently in our
metrics (PDR and Throughput) in these “stressful”
situations (higher mobility, more network load). The reason
of these phenomena is the aggressive use of route caching in
DSR. In our observation, such caching provides a significant
benefit up to a certain extent. With higher loads the degree
of caching is deemed too large to benefit performance.
Often, stale routes are chosen since route length (and not
any freshness criterion) is the only metric used to pick
routes from the cache when faced with multiple choices. Figure 5. Throughput of Receiving Packets
Picking stale routes causes two problems:
• Consumption of additional network bandwidth and
interface queue slots even though the packet is eventually
dropped or delayed
• Possible pollution of caches in other nodes
With high mobility, the possibilities of the caches being
stale are quite high in DSR. Eventually when a route
discovery is initiated, the large number of replies (as all
RREQs are replied) received in response is associated with
higher MAC overhead and cause increased interference to
data traffic. Hence, the cache staleness and high MAC
Figure 6. Packet Delivery Ratios
overhead mutually result in significant degradation in
performance for DSR in high mobility An efficient
desires a route to a new destination, it will have to wait until
mechanism to remove stale cached routes can improve
such a route can be discovered. On the other hand, since
performance of DSR. On other hand since in AODV only
routing information is constantly propagated and
the first arriving request packet (RREQs) is answered and
maintained in table-driven routing protocols, a route to
further no RREQs are answered therefore it leads to less no.
every other node in the ad hoc network is always available,
of replies (RREPs) .Also the error packets RERRs are are
regardless of whether or not it is needed. This feature,
broadcasted in AODV which leads to lesser MAC load as
although useful for datagram traffic, incurs substantial
compared to unicasted REERs of DSR which leads to much
signaling traffic and power consumption. Since both
MAC layer load.
bandwidth and battery power are scarce resources in mobile
computers, this becomes a serious limitation.
6. Conclusion
In this paper we provide descriptions of several routing
schemes proposed for ad hoc mobile networks. We also
provide a classification of these schemes according to the
routing strategy (i.e., table-driven and on-demand). We have
presented a comparison of these two categories of routing
protocols, highlighting their features, differences,
and characteristics .We has compared the performance of
Figure 4. Average End To End Delay DSDV, AODV and DSR We used a detailed simulation
model to demonstrate the performance characteristics of
these protocols. By simulating we can argue that if delay is
5. Comparisons
our main criteria than DSDV can be our best choice But if
The subsequent sections provide comparisons of the
reliability and throughput are our main parameters for
previously described routing algorithms. The next section
selection then AODV gives better results compare to others
compares table-driven protocols, and a further section
because its throughput and packet delivery ratio is best
compares on demand protocols.
among others. While there are many other issues that need
to be considered in analyzing the performance of ad hoc
5.1 Table-Driven vs. On-Demand Routing
networks, we believe that our work could provide intuition
As discussed former, the table-driven ad hoc routing border
for future protocol selection and analysis in ad hoc
on is similar to the connectionless approach of forwarding
networks. While we focus only on the network throughput, as a Sr. Engineer of Computern Application. Mr. Agarwal is the
reliability and the delay, it would be interesting to consider author of number of books. He has written number of books on
other metrics like power consumption, the number of hops various topics such as DBMS, Data Structure, Algoritms etc. Mr.
to route the packet, fault tolerance, minimizing the number Agarwal is also pursuing his Ph.D. in computer scince.He has
presented a number of papers in various national
of control packets etc.
conferences.Number of papers have been published in the National
Parameters On-demand Table-den
& International Journals.Mr Agarwal has also attended various
References Faculty Development Programe conducted by Infosys and TCS.
[1] J. Jubin and J. Tornow, “The DARPA Packet Radio Network Dr. Yahpal Singh is a Reader and HOD (CS)
Protocols,” Proc. IEEE, vol. 75, no. 1, 1987, pp. 21–32. in BIET, Jhansi (U.P.). He obtained Ph.D.
[2] C. E. Perkins and P. Bhagwat, “Highly Dynamic Destination- degree in Computer Science from Bundelkhand
Sequenced Distance-Vector Routing (DSDV) for Mobile University, Jhansi. He has experience of
Computers,” Comp. Commun.Rev., Oct. 1994, pp. 234–44. teaching in various courses at undergraduate
[3] L. R. Ford Jr. and D. R. Fulkerson, Flows in Networks, Princeton and postgraduate level since 1999. His areas of
Univ. Press, 1962. interest are Computer Network, OOPS, DBMS.
[4] C. Perkins, E. Belding-Royer, and S. Das, “RFC 3561: Ad hoc on- He has authored many popular books of
demand distance vector (AODV) routing,” July 2003, category: Computer Science for graduate and postgraduate level. He has
experimental. [Online]. Available: ftp://ftp.isi.edu/in- attended many national and international repute seminars and
notes/rfc3561.txt conferences. He has also authored many research papers of
[5] S. Murthy and J. J. Garcia-Luna-Aceves, “An Efficient Routing international repute.
Protocol for Wireless Networks,” ACM Mobile Networks and App.
J., Special Issue on Routing in Mobile Communication Networks, Manish Varshney received his M.Sc (C.S)
Oct. 1996, pp. 183–97. degree from Dr. B.R.A. University, Agra,
[6] A. S. Tanenbaum, Computer Networks, 3rd ed., Ch. 5, Englewood M.Tech. (IT) from Allahabad University and
Cliffs, NJ: Prentice Hall, 1996, pp. 357–58. Pursuing PhD in Computer Science. He is
[7] C. E. Perkins and E. M. Royer, “Ad-hoc On-Demand Distance working as a HOD (CS/IT) in SRMSWCET
Vector Routing,” Proc. 2nd IEEE Wksp. Mobile Comp. Sys. and Bareilly. He has been teaching various
Apps., Feb. 1999, pp. 90–100. subjects of computer science for more than
[8] D. B. Johnson and D. A. Maltz, “Dynamic Source Routing in Ad- half a decade. He is known for his skills at
HocWireless Networks,” Mobile Computing, T. Imielinski and H. bringing advanced computer topics down to
Korth, Eds., Kluwer, 1996, pp. 153–81. the novice's level. He has experience of industry as well as
[9] J. Broch, D. B. Johnson, and D. A. Maltz, “The Dynamic Source teaching various courses. He has authored various popular books
Routing Protocol for Mobile Ad Hoc Networks,” IETF Internet such as Data Structure, Database Management System, Design
draft, draft-ietfmanet-dsr-01.txt, Dec. 1998 (work in progress). and Implementation of Algorithms, Compiler Design books for the
[10] V. D. Park and M. S. Corson, “A Highly Adaptive Distributed technical students of graduation and postgraduation.He has
Routing Algorithm for Mobile Wireless Networks,” Proc. published various research papers in National and International
INFOCOM ’97, Apr. 1997. journals. He has also attended one faculty development program
[11] M. S. Corson and A. Ephremides, “A Distributed Routing organized by Oracle Mumbai on Introduction to Oracle 9i SQL and
Algorithm for Mobile Wireless Networks,” ACM/Baltzer Wireless DBA Fundamental I.
Networks J., vol. 1,no. 1, Feb. 1995, pp. 61–81.
[12] C-K. Toh, “A Novel Distributed Routing Protocol To Support Vidushi Gupta received her B.tech (C.S)
Ad-Hoc Mobile Computing,” Proc. 1996 IEEE 15th Annual Int’l. degree from Uttar Pradesh Technical
Phoenix Conf.Comp. and Commun., Mar. 1996, pp. 480–86. University, Lucknow.She is also pursuing
[13] R. Dube et al., “Signal Stability based Adaptive Routing (SSA) M.tech from Karnataka University, She is
for Ad-Hoc Mobile Networks,” IEEE Pers. Commun., Feb. 1997, working as Lecturer ( CS/IT department) in
pp. 36–45. SRMSWCET, Bareilly .She has published a
[14] C-K. Toh, “Associativity-Based Routing for Ad-Hoc Mobile research paper in an International journal. She
Networks,” Wireless Pers. Commun., vol. 4, no. 2, Mar. 1997, pp. has also attended one faculty development
1–36. program based on the “Research Methodologies”.
[15] S. Murthy and J. J. Garcia-Luna-Aceves, “Loop-Free Internet
Routing Using Hierarchical Routing Trees,” Proc. INFOCOM ’97,
Apr. 7–11, 1997.
[16] C. E. Perkins and E. M. Royer, “Ad Hoc On Demand Distance
Vector (AODV) Routing,” IETF Internet draft, draft-ietf-manet-
aodv-02.txt, Nov.1998 (work in progress).
Authors Profile
Vineet Agrawal is having total more than 15
years experience in teacing and industry.Mr.
Vineet Agarwal is presently working as a Asst.
Director of Rakspal Bahadur College of
Engineering & Technology, Bareilly. Author is
MCA & M.Tech from Birla Institute of
Technology, Mesra Ranchi. Author has worked
in Synthetic & Chemicals Ltd. For four years since 1995 to 1999
A Secure Iris Image Encryption Technique Using

Bio-Chaotic Algorithm
Abdullah Sharaf Alghamdi1, Hanif Ullah2
Department of Software Engineering,
College of Computer and Information Sciences,
King Saud University, Riyadh, Kingdom of Saudi Arabia
ghamdi@ksu.edu.sa , hanif@ksu.edu.sa
Abstract: Due to dramatic enhancement in computers and masquerade like a particular person or monitor the person’s
communications and due to huge use of electronic media, private data [3].
security gains more and more importance especially in those Similarly some chaos-based cryptosystems are used to solve
organizations where information is more critical and more the privacy and security problems of biometric templates.
important. The older techniques such as conventional The secret keys are randomly generated and each session
cryptography use encryption keys, which are long bit strings and has different secret keys. Thus biometric templates are
are very hard to memorize such a long random numbers. Also it
encrypted by means of chaotic cryptographic scheme which
can be easily attacked by using the brute force attack technique.
Instead of using the traditional cryptographic techniques,
makes them more difficult to decipher under attacks [4].
Biometrics like Iris, fingerprints, voice etc. uniquely identifies a Moreover some chaotic fingerprint images encryption
person and a secure method for stream cipher, because techniques are also proposed which combines the shuttle
Biometric characteristics are ever living and unstable in nature operation and nonlinear dynamic chaos system. The
(with respect to recognition). In this paper we used the idea of proposed image encryption technique provides an efficient
bio-chaotic stream cipher which encrypts the images over the and a secure way for fingerprint images encryption and
electronic media and also used to encrypt the images to store it storage [5].
into the databases to make it more secure by using a biometric Similarly some new image encryption technique based on
key and a bio-chaotic function. It enhances the security of the hyper-chaos is also proposed, which uses an image total
images and it should not be compromised. The idea also gives shuffling matrix to shuffle the pixel positions of the plain
birth to a new kind of stream cipher named bio-chaotic stream
image and then the states combination of hyper-chaos is
cipher. The paper also describes how to generate an initial key
used to change the gray values of the shuffled image [6].
also called initial condition from a biometric string and how to
encrypt and decrypt the desired data by using the bio-chaotic In order to improve the security of the images we proposed a
function. better idea which is a new type of algorithm called Bio-
Keywords: Biometric, stream cipher, bio-chaotic algorithm Chaotic stream cipher algorithm (BCA) for image
(BCA), cryptography, key. encryption which overcomes the problems of some of the
algorithms used previously for the same purpose. In this
1. Introduction algorithm we used the iris images and extract their features
by using the L.Rosa [9] iris feature extraction code. These
Due to dramatic enhancement in computers and
features are then used to generate the initial condition for
communications and due to huge use of electronic media,
the secret key using the Hamming Distance technique,
security gains more and more importance especially the
which is then Xored to the iris extracted features to generate
security of biometric images become a hot issue. Biometric
another secret key called biometric key. This biometric key
images are mostly used for the authentication system
is then used in the chaotic function to generate the bio-
because of there ever living and unstable (with respect to
chaotic stream cipher for further encryption.
recognition) characteristics. Conventional or traditional
The rest of the paper is organized such that section 2
symmetric or asymmetric cryptography is limited only to
consists the related work of the paper. Section 3 will show
text files but it cannot be used in case of huge files like
the basic working and idea of the BCA. Section 4 presents
images and videos.
the graphical representation of the key generation process
Image encryption techniques are extensively used to
and logistic map for the algorithm. Section 5 shows some
overcome the problem of secure transmission for both
mathematical comparisons with other algorithms. Finally
images and text over the electronic media by using the
section 6 draws a conclusion.
conventional cryptographic algorithms. But the problem is
that it cannot be used in case of huge amount of data and 2. Related work
high resolution images [2].
Instead of using the traditional way of cryptography for The same work is carried out in our conference paper
image encryption we can also use biometric e.g. fingerprint, already published. The same algorithm is used for the
iris, face, voice etc for the same purpose. The main encryption of the Iris images. In this paper we elaborate the
advantage of a biometric is that it is ever living and unstable algorithm with more detail and add some new features to the
characteristics of a human being and it cannot be existing proposed system [19].
compromised. However it also suffers from some biometric
specific threats and that is the privacy risk in biometric The work that we seen relevant to our work is that of
systems. An attacker can interpret a person’s biometric data, Haojiang Gao, Yisheng Zhang, Shuyun Liang and Dequn Li
which he can use for many illegal operations such is to which proposed a new chaotic algorithm for image
encryption[2]. In this paper they presented a new nonlinear condition and to increase the security of the proposed system
chaotic algorithm (NCA) which uses power function and [10].
tangent function instead of linear function. The
experimental results demonstrated in this paper for the
image encryption algorithm based on NCA shows 3. Proposed System Bio-Chaotic Algorithm
advantages of large key space and high-level security, while (BCA)
maintaining acceptable efficiency [2].
The basic idea of the algorithm is such that we took an iris
Similarly the work done by Song Zhao, Hengjian Li, and Xu image and extract its features by using L.Rosa code [9]. L.
Yan for the security and Encryption of fingerprint images is Rosa used a code to generate a binary pattern from the given
more relevant to our work [5]. In this paper they proposed a iris image. The binary pattern is further divided into small
novel chaotic fingerprint images encryption scheme blocks of binary data to make the process simplified,
combining with shuttle operation and nonlinear dynamic because it is very difficult to encrypt the binary pattern of
chaos system. The proposed system in this paper shows that hundreds of thousands of bits at once. In our case we made
the image encryption scheme provides an efficient and each block of 128 bits to make it simpler and to encrypt each
secure way for fingerprint images encryption and storage block easily. A random block is then selected to create the
[5]. initial condition for the secret key. The random selection of
Also the work done by Muhammad Khurram Khan and the block is preferred because of the attackers, so that no one
Jiashu Zhang for implementing templates security in remote can easily understand that which block is selected for the
biometric Authentication systems seems relevant to us [4]. initial condition.
In this paper they presented a new chaos-based cryptosystem At the transmission time of the image the bits of this
to solve the privacy and security issues in remote biometric random selected block is encrypted by using Quantum
authentication over the network. Experimental results Encryption Technique [8]. Quantum encryption uses light
derived in this paper shows that the security, performance particles, also call photons instead of bits at communication
and accuracy of the presented system are encouraging for time. A photon can have one of the four orientations or
the practical implementation in real environment [4]. shapes, 450 diagonal, -450 diagonal, horizontal or vertical.
Each of these represents a bit, - and / represents a 0, while |
Similarly a new image encryption technique was introduced and \ represents a 1[8].
by Tiegang Gao and Zengqiang Chen in their paper based Fig 1 presents the block diagram of the proposed bio-chaotic
on the image total shuffling matrix to shuffle the position of algorithm. The basic steps of the algorithm are as follows.
the image pixels and then uses a hyper chaotic function to
complex the relationship between the plain image and the I. Generation of the initial condition from the randomly
cipher image. The suggested image encryption algorithm selected block taken from the binary pattern of the iris
has the advantage of large key space and high security [6]. image. The technique used to create the initial condition
is that of Hamming Distance i.e.
Moreover a coupled nonlinear chaotic map and a novel
chaos-based image encryption technique were used to
encrypt the color images by Sahar Mazloom and Amir Where n=1, 2, 3, 4…... Some other techniques can also be
Masud Eftekhari-Moghadam in their paper [10]. They used used for the same purpose like
the chaotic cryptography technique which is basically a
symmetric key cryptography with a stream cipher structure.
They used the 240 bit long secret key to generate the initial
Figure 1. Block Diagram of Bio-Chaotic Algorithm

II. This initial condition is then converted into secret key by are included in this section. The logistic maps are derived
using the LFSR method. An LFSR of length n over a on the basis of the following mathematical function.
finite field Pq consist of n stages [an-1,an-2,an-3,……..,a0]
with ai Є of Pq, and a polynomial On the basis of the above equation we generate different
logistic maps using different values. Fig.3 and 4 shows the
statistical correlation curves of the sequence. By observing
the maps carefully it’s clear that even changing in a small
III. The secret key and iris template is then Xored in parallel part of the value the whole map become different.
to generate the biometric key by using the equation, Fig.5 shows the encrypted images by using different chaotic
values. From the figure it is clear that how strong the
encryption process is that by changing even a small part of
IV. This biometric key is further Xored with different the value the image become more and more invisible.
blocks of the iris template (divided into blocks of 128 Similarly the decryption process is more sample as like the
bits/block) which encrypts the image in such a way that encryption by just Xoring the Ciphered image with the key
and we will get the original image.
no intruder or attacker can easily decrypt the image.
V. To make the algorithm stronger and more secure we
add the chaotic function to the biometric key and apply
it over the iris image which encrypts it in a more secure
way. We use the following logistic equation [4].
Where n=1, 2, 3… is the map iteration index and r is the

value taken from the algorithm. On the basis of equation 4
we generate the logistic map for different values of the
algorithm the detail of which will be given in the next
section.
3.1 Decryption Process
The decryption process of the used image is carried on by Figure 2. iris images used for experiments
the same way using the same key used for the encryption
process but in the opposite direction i.e. the ciphered image
is Xored with the biometric key to get the image back in its
original form. The receiver will first decrypt the randomly 1
selected block by using the same technique used for the

encryption process i.e. Quantum Decryption technique [8]. 0.5
After decrypting the selected block the receiver will generate
the initial condition with the same procedure used for the
encryption process and will decrypt the image. The equation 0
0 20 40 60 80 100 120
used for the decryption process is as follows. Logistic Map for Cipher and Decipher date
1
Real value b/w 0 and 1
It shows the Exclusive OR operation.
0.5
4. Experimental Analysis of the Algorithm

0
In order to evaluate and check the performance of the 0 20 40 60 80 100
Size of Biometric template
120
proposed algorithm i.e. Bio-chaotic algorithm we took iris

images from one of the renowned database CASIA (Chinese Figure 3. Logistic map when value= 0.54000000000001
Academy of sciences and institute of Automation) [11]. The
database contains a lot of iris images taken from different
people eyes. In our case we use 2 or 3 of the iris images
from this database to carry out our experimental process.
These images are shown in fig.2.
The algorithm is analyzed and tested by using different

values for x where x is any real value between 0 and 1.
Some of the logistic maps based on the experimental
analysis performed over sample and encrypted iris images
1
0.5
0
0 20 40 60 80 100 120
Logistic
Real value b/w 0 and 1
1
0.5
Figure.5. (b) Encrypted image at value= 0.7000000000001

0
0 20 40 60 80 100 120
Size of Biometric
Figure 4. Logistic map when value= 0.58000000001
Figure 5. (c) Encrypted image at value= 0.9800000000001
Figure 5. (a) Encrypted image at value= 0.580000000000
Table 1: Avalanche effect of the BCA
NO AvalanchePC Effect AvalanchePK Effect AvalancheCK Effect

for BCA for BCA For BCA
1 48.8758 % 47.9530 % 52.9465 %
Where percentPC (percent difference between plain image

5. Statistical Analysis of Bio-Chaotic and ciphered image) could be found out by using the
Algorithm (BCA) equation
In this section statistical analysis and mathematical

observations like Avalanche effect, confusion and diffusion,
and entropy of the proposed algorithm are mentioned.
Where Acc is basically an Accumulator and it could be find
5.1 Avalanche Effect out by
The Avalanche effect refers to a desirable property of the

cryptographic algorithms. The Avalanche effect is evident
if, when an input is changed slightly (for example, flipping In equation 8 DiffPC means the difference between plain
a single bit) the output changes significantly (e.g., half the image and ciphered image and it is basically the Xor
output bits flip). In the case of quality block ciphers, such a operation between plain image and cipher image. Similar
small change in either key or the plain text should cause a methodology is used to find out the avalanche effect between
drastic change in the cipher text. In our case the Avalanche plain image and key and ciphered image and key. The
effect of the proposed system is determined by using the results of the above equations are tabulated in table1.
following mathematical equation [12].
The table shows that the Avalanche effect between the plain
image and ciphered image, and plain image and key is less
than 50 percent that is a more desirable value for any
algorithm. Similarly the Avalanche effect between ciphered

image and key is round about 50 percent which is slightly
bigger than the rest of the two, but again it is a desirable
value for our proposed algorithm.
5.2 Confusion and Diffusion By using the above equation we found the entropy of our
proposed system which is round about 127.3. The values
Confusion and diffusion are the two properties of the show better uncertainty and randomness of bits in the
operation of a secure cipher. Confusion refers to making the algorithm. The probability of each bit is 0.5. The entropy
relationship between the key and the cipher text as complex will be high if there is more randomness in the bits used in
and as involved as possible. Diffusion refers to the property the ciphered image. Table 2 shows the entropy of our
that redundancy in the statistics of the plain text is proposed system.
dissipated in the statistics of the cipher text [12]. Confusion
and diffusion are the same properties like Avalanche effect
which is elaborated in the previous section. The confusion Table 2: Entropy of Bio-chaotic Algorithm
and diffusion of the proposed algorithm is round about 49%,
which shows the strength of the proposed system. Bio-chaotic Algorithm Entropy(H(X))
1 64.67
5.3 Entropy
Entropy is a measure of the uncertainty or randomness

associated with a random variable. It is basically a measure 5.4 Histogram of the Images
of the average information content one is missing when one
does not know the value of the random variable [12]. Figure 6 shows the histogram for the plain and encrypted or
Entropy can be found by using the equation ciphered images used in the bio-chaotic algorithm.
8000
7000
6000
5000
4000
3000
2000
1000
0 50 100 150 200 250
Figure 6. (a) histogram of the plain image

15000
10000
5000
0 50 100 150 200 250
Figure 6. (b) histogram of the ciphered image
[2]Haojiang Gao, Yisheng Zhang , Shuyun Liang , Dequn Li, “A

6. Conclusion new Chaotic Algorithm for image Encryption”, Elsevier ,
Science Direct , Aug 2005.
This paper presents a new and novel idea for the encryption
[3] Andrew Teoh Beng Jin, David Ngo Chek Ling, Alwyn Goh, “
and decryption of the iris images. The proposed algorithm Biohashing : two factor authentication featuring fingerprint
called the Bio-Chaotic Algorithm (BCA) takes an iris image data and tokenized random number “ April 2004,”The Journal
and with the help of L.Rosa code generates the iris features Of The Pattern Recognition Society “ , Elsevier , April 2004.
or the binary bits pattern for the image. This binary bits [4] Muhammad Khurram Khan, Jiashu Zhang, “Implementing
pattern is then divided into small blocks of bits to simplify Templates Security in Remote Biometric Authentication
the process. Each block is that of 128 bits long. Then a Systems”, IEEE Conf. Proceedings on CIS’06, China, pp.
random block is selected from all these blocks to create the 1396-1400, Vol.2, 2006.
initial condition. This initial condition is then passed from [5] Song Zhao, Xu Yan,”A secure and efficient fingerprint images
encryption scheme” Proceedings of the IEEE, 2008, pp- 2803-
the LFSR to generate the secret key. A secret key of 128 bits
2808.
is generated from the result of the LFSR. This secret key is [6] Gao Tiegang, Chen Zengqiang,” A new image encryption
then used for the encryption of the iris image. A Quantum algorithm based on hyper-chaos” Elsevier, Science Direct,
encryption technique is also used to encrypt the randomly Physics Letters A, Volume 372, Issue 4, p. 394-400, 2007.
selected block, so that no one can easily attack the block [7] Muhammad Khurram Khan, Jiashu Zhang, “Improving the
used for the generation of the secret key. The same Security of ‘A Flexible Biometrics Remote User Authentication
procedure is then used at the receiver end to decrypt the iris Scheme’”, Computer Standards and Interfaces (CSI), Elsevier
image. Chaotic function is used to make the algorithm more Science UK, vol. 29, issue 1, pp. 84-87, 2007.
secure and make the process of the encryption and [8] T Morkel 1, JHP Eloff,” Encryption Techniques: A Timeline
decryption more complex. Experimental and statistical Approach”, Information and Computer Security Architecture
(ICSA) Research Group Department of Computer Science
analysis of the algorithm shows that the algorithm is
University of Pretoria, 0002, Pretoria, South Africa
stronger and more secure and can be used for the practical
implementation of the iris images encryption. [9] Iris code by Luigi ROSA, L'Aquila ITALY
(19600bits)”http://www.advancedsourcecode.com/irisphase.asp
7. Future Work [10] Mazloom Sahar, Eftekhari-Moghadam Masud Amir”, Color
image encryption based on Coupled Nonlinear Chaotic Map”,
In the future we would like to use the same technique for the the journal of Chaos, Solitons and Fractals 42 (2009) 1745–
encryption of fingerprint images. Also we would like to use 1754, ELSEVIER, 2009.
a block size more than 128 bits to make the algorithm [11] CASIA Iris Database. [Online March, 2006]
stronger and more secure. http://sinobiometrics.com.
[12] Shannon, C. E., “A Mathematical Theory of Communication,”
Bell System Technical Journal, July 1948, p.623.
References [13] Yao-Jen Chang, Wende Zhang, and Tsuhan Chen,
[1] Arroyo David, Li Chengqing, Li Shujun, Alvarez Gonzalo, “Biometrics-Based Cryptographic Key Generation” 2004 IEEE,
Halang A. Wolfgang,” Cryptanalysis of an image encryption USA.
scheme based on a new total shuffling algorithm”, Elsevier , [14] Ren Honge, Shang Zhenwei, Wang Yuanzhi , Zhang Jian ,
Science Direct, Volume 41, Issue 5, 15 September 2009, Pages “A Chaotic Algorithm of Image Encryption Based on
2613-2616
Dispersion Sampling” The eight International conference on

Electronic Measurement and Instruments” 2007 IEEE.
[15] Shenglin Yang, Ingrid M. Verbauwhede,”Secure Fuzzy Vault
Based Fingerprint Verification System”, 2004 IEEE.
[16] The MathWorksTM Accelerating the pace of engineering and
science. “www.mathworks.com” Date accessed: 2 Feb 2009
[17] Eli Biham, Louis Granboulan, Phong Q. Nguy “Impossible
Fault Analysis of RC4 and Differential Fault Analysis of RC4”
Computer Science Department, Technion – Israel Institute of
Technology, Haifa
[18] J. Daugman,”High confidence visual recognition of persons by
a test of statistical independence “, IEEE Transactions on
Pattern Analysis and Machine Intelligence vol.15, 1993,
pp.1148-61.
[19] Alghamdi S. Abdullah, Ullah Hanif, Mahmud Maqsood, Khan
K. Muhammad., "Bio-chaotic Stream Cipher-Based Iris Image
Encryption," cse, vol. 2, pp.739-744, 2009 International
Conference on Computational Science and Engineering,
Canada.
[20] J.G. Daugman, “Uncertainty Relation for Resolution in Space,
Spatial Frequency, and Orientation Optimized by Two-
Dimensional Visual Cortical Filters,” J. Optical Soc. Amer.,
vol. 2,no. 7, pp. 1,160-1,169, 1985.
Authors Profile
Dr. Abdullah Alghamdi is a full time
associate professor, SWE Department,
College of Computer and Information
Sciences, KSU. He holds a Ph.D. in the
field of Software Engineering from the
department of computer science,
Sheffield University, UK, 1997. He
obtained his M.Sc. in the field of
software development technologies
from the UK in 1993. In the academic
year 2004/5 he worked as a visiting
professor at School of IT and Engineering, University of Ottawa,
Ottawa, Canada, where he conducted intensified research in Web
Engineering as part of his Post-Doc program. He recently
published a number of papers in the field of Web engineering
methodologies and tools. Dr. Abdullah worked as a part-time
consultant with a number of governmental and private
organizations in the field of IT strategic planning and headed a
number of IT committees inside and outside KSU. Currently he is
chairing the Software Engineering Department at KSU and part
time consultant at Ministry of Defense and Aviation.
Hanif Ullah received the BIT (Hons) and

MSc. Degree in Information Technology
from Iqra University Karachi in 2004 and
Quaid-e-Azam University Islamabad,
Pakistan in 2007 respectively. In January
2008, He joined King Saud University,
Saudi Arabia as a Research Assistant and
start working on Network and Information
security related topics. Currently He is
working as a Lecturer in the Department
of Software Engineering, College of
Computer and Information Sciences, King Saud University, Saudi
Arabia.
An ASes stable solution in I-Domain

G. Mohammed Nazer1 and Dr.A.Arul Lawrence Selvakumar2
1
Asst.Pofessor & Head, Dept of MCA, IFET College of Engineering,
Villupuram, India.
kgmohammednazer@gmail.com
2
Professor & Head, Dept of CSE & IT, Kuppam College of Engineering,
Kuppam, India.
Aarul72@hotmail.com
Abstract: Routers on the Internet use an interdomain routing BGP convergence time analysis: How long it takes BGP to
protocol called the Border Gateway Protocol (BGP) to share the converge to a stable routing outcome? This is another
routing information between the Autonomous Systems (ASes). question, which is closely related to BGP. To answer this
These ASes defines local BGP policies that lead to various question, we require a formal definition of measuring the
routing anomalies like BGP divergence. In this paper, we close a convergence rate, as the Internet is asynchronous.
long-standing open question of Griffin and Wilfong, by
We analyze the BGP convergence time in particular,
showing, for any network structure, if there exists two stable
Internet-like settings. In this Gao and Rexford settings,
routing outcomes, then there is a possibility of BGP oscillations.
Our results provide the first non-trivial necessary condition for every pair of neighboring ASes can have a business
BGP safety – uniqueness of the stable routing outcome. relationship or a peering relationship, which causes natural
Another question, which is closely related to BGP, is how constraints on the ASes’ routing policies.
long it will take to converge to a stable routing outcome. We also However, our first result is negative. We show that, even
address this by analyzing a formal measure of the convergence for the restricted class of preferences, there are instances
time of BGP for the policies presented by Gao and Rexford. such that the convergence rate of BGP is linear in the size of
Even for the restricted class of preferences, we prove that (i) the the network. Specifically we show that in a network with n
convergence time is linear in the size of network (ii) BGP’s nodes, it takes n phases to converge. We also prove that the
running time cannot be more than (roughly) twice the length of lower bound is tight: BGP is always guaranteed to converge
the longest customer-provider chain in the network. in n time steps. As there are thousands of ASes in today’s
Internet, the linear bound does not signify well. However,
Keywords: BGP, Border Gateway Protocol, Interdomain routing,
one would expect BGP to converge at a much quicker rate in
network security, routing, networks, routing protocols, BGP safety.
practice as ASes’ routing policies are local in the sense that
they are not influenced by ASes that are far away. We prove
1. Introduction that the number of phases required for convergence is
bounded by approximately twice the depth of customer-
provider hierarchy.
BGP is the de facto protocol enabling interdomain
routing in the Internet. The task of Interdomain routing is to
establish routes between the administrative domains which
are called as Autonomous Systems (ASes) in the Internet. 2. A formal Model
Global routes are formed from the local decisions that are
based on the private routing policies. These routing
2.1 BGP dynamics
selections are communicated by the ASes to the neighboring
ASes. Persistent routing oscillations are formed due to the Network model and its policies: In our model, we define
lack of global coordination between the local routing a network by an AS graph G = (N, L), where N represents
policies. the set of ASes, and L represents number of physical
communication links between ASes. N consists of n source-
BGP safety – Unique stable routing outcome: The main nodes {1,…,n} and a unique destination node d. P i denotes
contribution in this paper is showing that BGP safety the set of all simple non-cyclic routes from i to d in G. Each
necessitates the existence of a unique stable solution. This is source-node i has a ranking function ≤i , that defines a strict
the result that closes the long-standing open question first order over P i (that is i has strict preferences over all routes
posed by Griffin and Wilfong [8]. To be more precise, Two from i to d). We allow ties between two routes in P i only if
stable solutions in a network implies that the network is they share the same first link (i,j). The routing policy of
unstable that lead to oscillations. To analyze the BGP each node i consists of ≤i and of i’s import policy and export
dynamics in a more simplified form, we use a more policy.
convenient structure, called state-transition graph. The
i’s import policy dictates which set of routes Im(i) ⊆ P i i
state-transition graph, not only a useful conceptual tool for
is willing to send traffic along. We assume that ø ≤i Ri for
evaluating and designing various network configurations but
any route Ri ∈ Im(i) (i prefers any route in Im(i) to not
also assist in detecting the potential routing oscillations and
getting a route at all) and that R’i ≤i ø for any route R’I ∉
how to debug them.
Im(i) (i will not send traffic at all rather than send traffic node i to a neighboring node j announcing the route
along a route not it Im(i)). that i believes it is sending traffic along. Formally, there
i’s export policy dictates which set of routes Ex(i,j) ⊆ P i i is a ji-knowledge transition from state s to state s’ if
is willing to announce to each neighbor j. there is a node i and a neighboring node j such that:
Update Messages and Activation Sequence: Basically The forwarding vectors in the two states are identical.
the BGP belongs to a family of routing protocols named The knowledge matrix in s’ is identical to the
path-vector protocols. In this model, there are two kinds of knowledge matrix in s with the exception of i’s belief
actions that an active node may carry out potentially change about j, and ks’ij = ks ij. In other words, i learns of the
the global routing state: route that j currently believes it is using.
• A node i may select a route or change its selected This definition reflects the restricted asynchrony of our
route from the routes to given destination d that it dynamic model. We can phrase this restriction equivalently
currently believes to be available. as: Update messages can be delayed in transit, but when
• A node j may send an update message to a they are delivered, a fresh update message from the same
neighboring node i, informing i of the route that j sender is delivered immediately (and thus overrides the
is currently using to destination d. The update delayed update.) Thus, the state description does not have to
message is assumed to be delivered immediately include messages in transit.
(without propagation delay), and is immediately
reflected in updated beliefs that i has about j’s Stability and Oscillations in the State-Transition Graph:
route.
The selection and update actions can occur at arbitrary A stable state is one in which the nodes forward traffic
times. In particular, note that the update messages are not along a stable solution, and have complete and accurate
required to be sent at a given time interval or whenever j’s knowledge about their neighbors’ routes. We want to prove
route changes. It is easy to show that a stable solution is the existence of potential BGP oscillations in the state
always in the form of a tree rooted in d. Further, the import transition graph. In many cases, oscillations occur only for
and export policies can be folded into the routing policies, specific timings of asynchronous events. In particular,
by modifying the preferences so that paths that are filtered starting at any given point of time, every node eventually
out have the lowest possible value. updates its route selection if its knowledge of routes has
changed, and every node eventually receives update
2.2 The State-Transition Graph messages from each neighbor that has changed a route.
Further, in a given router, there can only be a finite
In this subsection, we describe the state transition graph – number of other activations taking place between subsequent
a tool that we use to analyze the convergence of BGP on routing selections or updates. It is for this reason, we look
different instances. for oscillations that can arise through a fair activation
The state-transition graph of an instance of BGP is sequence. An infinite activation sequence σ said to be fair if
defined as follows: The graph consists of a finite number of each transition in A appears infinitely often in σ. A fair
states, each state s is represented by an n-dimensional cycle in the state-transition graph is a finite cyclic path that
does not contain a sink, such that every action in A is taken
at least once in each traversal of the cycle.
forwarding vector of routing choices rs = (rs1,…,rsn), and n ∗
2.3 Implications for the evaluation model of Griffin
We modify the dynamic evaluation model of Griffin in
n knowledge matrix Ks = {ksij}i,j. rsi specifies the identity of two ways:
the node to which node i’s traffic is being forwarded, and ksij • Update messages are not delayed, instead, arrive
specifies the (loop-free) route that node i believes that its immediately to the destinations.
neighboring node j is using. We define ksij = NULL when j • In BGP execution, it is not necessary that a node
is not a neighbor of i; any knowledge that i has about non- inform a neighboring node of every new route it
neighboring nodes’ routes is irrelevant to i’s route selection changes, rather it is enough if it announces once in a
and advertisement decisions. We assume, naturally, that while.
node i knows who it is forwarding traffic to: r si must be the
first hop in ksij. We allow two types of atomic actions that 3. Two stable solutions leads to BGP
lead to transitions from s to s’: oscillation
• Route transition – Route selection actions: Informally, a
route transition arises when a node I updates its selected In this section we prove our main result, that if there are
route by picking its favorite route from its current two stable solutions then the network is unstable in the sense
knowledge set of routes used by its neighbors. Formally, that persistent route oscillations are possible.
there is an i-route transition from state s to state s’ if Theorem: If the AS graph G contains two stable solutions,
there is a node i such that: The forwarding vector in s’ then there is a fair activation sequence under which BGP
is identical to the forwarding vector in s with the will oscillate on G. That is, two stable solutions imply that
possible exception of i. the network is unstable, in the sense that it could plausibly
• Knowledge transition – Informally, a knowledge lead to persistent route oscillations. Therefore, to achieve
transition is an update message sent from a specific
BGP stability, the network must have a unique stable the longest route in the stable solution. It is easy to construct
solution. instances with n nodes in which there are routes of length
The intuition behind our proof is as follows. In the state- Ω(n). However, these instances are unnatural; currently,
transition graph, each stable state will have a corresponding Internet routes tend to be much shorter than this. For this
“attractor region”: a subset of states (possibly just the stable reason, we focus on bounding the BGP convergence time on
state itself, or much larger) that, once reached, we can be Internet-like graphs.
certain that the system will ultimately converge to the stable Example: The graph in Figure 1 depicts a network with n
state. We can visualize the state-transition graph as a map, nodes, and a destination node d. Node 1 prefers to go
with each attractor region a different color – red, blue, etc. directly to d. Any other node i prefers the route i → i− 1 →
However, there will also be some states that do not lie in any d over the direct route i → d. All routes of length greater
one attractor region, because different evolutions from that than 2 are less desirable to any node. This set of path
state could lead to different stable states. We label these preferences is compatible with the Gao-Rexford constraints
states with a distinct color – purple, say – and show that the for the following set of customer-provider relationships: 1 is
Zero state must belong in this subset. a customer of 2, 2 is a customer of 3, etc.; and, additionally,
The key to the proof is showing that, starting from any d is a customer of every other node.
purple state, we can find a fair activation sequence that ends
at another purple state. We use the properties of route
selection and update actions to show that we can swap the
order of any two consecutive activations, perhaps repeating
one of them, and achieve the same result as the original
order. Thus, it is not possible that any given activation a
leads to a red state in the original order, but leads to a blue
state in the perturbed order. Using this, we show that we can
add each activation while staying within the purple region.
As the graph is finite, this implies the existence of a fair
cycle. If an instance of BGP results in a state- transition
graph (for a given destination) that has a fair cycle, we will In each phase, initially all update messages go through,
infer that there is a plausible sequence of route selections and then all nodes are activated. In the first phase, only node
and updates that will cause BGP to oscillate. 1 will change its routing choice and will route to d. In the
next phase, only node 2 will change its routing choice and
4. BGP’s convergence Rate will route through 1. Then node 3 will change to route
In this section, we handle the question of how long BGP through d and so on. The network will eventually converge
takes to converge to the unique stable solution. BGP is an to the routing outcome in which all odd nodes route directly
asynchronous protocol, and individual messages may be lost to d and all even nodes route the rough their counter-
or delayed arbitrarily. As we cannot assume a bound on the clockwise neighbor.
actual elapsed time of a single message, any model of We prove that this bound is tight for the class of instances
convergence “time” needs to define a unit of time that satisfy the Gao-Rexford conditions. In fact, we prove a
measurement that remains meaningful in this asynchronous slightly stronger result: The following proposal shows that
setting. Let us consider the following definition: our bound on BGP’s convergence rate is tight on the larger
Definition: A BGP phase is a period of time in which all class consisting of all instances in which the “No Dispute
nodes get at least one update message from each Wheel” condition of [3], [5] holds.
neighboring node, and all nodes are activated at least once Proposal: If “No Dispute Wheel” holds then BGP’s
after receiving updates from their neighbors. convergence rate is at most n phases.
We analyze the number of BGP phases it requires for the Proof: Let us assume that indeed the “No Dispute Wheel”
network to converge. The underlying principle in this condition holds in a network graph G with a destination
definition is that, although it is difficult for the analyst to node d. At every phase, one of the nodes of the graph
assert numerical bounds on the update frequencies at converges to a route that will not change from that point on.
different nodes, it is reasonable to expect that all nodes are The first node that converges in the first phase is the
updating at similar timescales. The definition of phases destination node d, that has the empty path, and announces
admits asynchrony, thus capturing the realistic possibility that path to its neighbors. We now show that there must
that different sequences of update activations can lead to exist a node in the network that is a direct neighbor of the
different transient behavior. At the same time, by tying the destination d and that its most preferred path is going
unit of measurement to the slowest node’s update instead of directly to d.
a fixed time unit (or the fastest update), we avoid To see that this is indeed the case, pick an arbitrary node
pathological worst-case time bounds that are only attained v, look at its most preferred path to the destination. This
if, for example, one node’s update cycle is measured in years path goes through a neighbor of d right before it reaches d.
instead of seconds or minutes. We shall denote this neighbor by v1. Now, consider the most
How many consecutive phases does it take BGP to preferred path of node v1, and the closest node to d on that
converge to a stable solution in the worst case? Routes are path that we shall denote by v2. In this manner we define
propagated through the network one hop at a time, so the the nodes vi for i = 1, 2, 3, ... At some point, nodes in the
best we can hope for is a time proportional to the length of sequence we defined must start repeating. If only one node
repeats infinitely then this node must have a direct route as Answers to these questions could provide network operators
its most preferred path, and we are done. Otherwise, the with new principles to tradeoff the desire for flexible
sequence of repeating nodes vk , vk+1 , . . . , vk+l (for some k, autonomous policies with the need for global routing
l) forms a dispute wheel: each node prefers to go through efficiency. Finally, there are practical aspects of BGP
the next one in the sequence rather than directly to d. This operations not considered in this paper such as MRAI
contradicts our assumption. Therefore, there exists a node (Minimum Route Advertisement Interval) and RFD (Route
that prefers to go directly to d over any other path. It will Flap Damping [19]), which play a significant role in BGP
choose this path to d on the second phase, send update convergence [20], [21].
messages to its neighbors, and will never again change its
path (since no path will be better).
We now continue to follow the convergence process, and References
observe that at any phase, there must exist a node v that
converges to its most preferred route given the route of the [1] K. Varadhan, R. Govindan, and D. Estrin,
nodes in the system that have already permanently “Persistent route oscillations in inter-domain
converged. This node will never again change its path routing,” Computer Networks, vol. 32, no. 1, pp. 1–
(because unless previous nodes change, it will have no better 16, March 2000.
path, and these previous nodes have also converged). To [2] T. G. Griffin and G. Wilfong, “An analysis of BGP
prove that such a node must exist, we fix the routes of all convergence properties,” in Proceedings of
permanently converged nodes, and pick an arbitrary node v1 SIGCOMM 1999.
that did not converge. We once again define the sequence of [3] T. G. Griffin, F. B. Shepherd, and G. Wilfong, “The
nodes v1, v2 , v3 . . . by defining the node vi+1 as the node that stable paths problem and interdomain routing,”
is closest to d on the most preferred path of node vi that did IEEE/ACM Transactions on Networking, vol. 10,
not permanently converge. The set of paths from which we no. 2, pp. 232–243, April 2002.
select this most preferred path, is the set of paths that are [4] L. Gao and J. Rexford, “Stable Internet routing
consistent with the nodes that have already permanently without global coordination,” IEEE/ACM
converged. Once again, this sequence of nodes must repeat, Transactions on Networking, vol. 9, no. 6, pp.
and since it cannot contain a dispute wheel, it must have 681–692, 2001.
only a single repeating node that is the closest node that did [5] L. Gao, T. G. Griffin, and J. Rexford, “Inherently
not converge on its own most preferred path. In the next safe backup routing
phase, this node’s path converges. We have thus shown that with BGP,” in 20th INFOCOM. Pistacaway: IEEE,
if the AS graph contains no dispute wheels, the convergence 2001, pp. 547–556.
time of BGP is bounded by the number of nodes in the entire [6] T. G. Griffin, A. D. Jaggard, and V. Ramachandran,
network graph. “Design principles of policy languages for path
vector protocols,” in SIGCOMM ’03: Proceedings
5. Conclusion of the 2003 conference on Applications,
technologies, architectures, and protocols for
computer communications. New York: ACM, 2003,
We studied fundamental questions related to BGP pp. 61–72.
whether it will converge to a unique stable solution and how [7] A. D. Jaggard and V. Ramachandran, “Robustness
long it will take to converge. We proved that, for any of class-based path- vector systems,” in
network, if there exists two stable routing outcome, then Proceedings of ICNP’04, IEEE Computer Society.
persistent BGP oscillations are possible. So the existence of IEEE Press, October 2004, pp. 84–93.
unique stable routing outcome is a necessary condition for [8] N. Feamster, R. Johari, and H. Balakrishnan,
the BGP safe convergence. We also analyzed the worst-case “Implications of autonomy for the expressiveness of
convergence time of BGP on instances that satisfy the policy routing,” in SIGCOMM ’05: Proceedings of
conditions mentioned by Gao-Rexford. We proved that the the 2005 conference on Applications, technologies,
convergence time on a graph with n nodes is Θ(n) in the architectures, and protocols for computer comm.
worst case, but is much smaller in networks with shallow New York, NY, USA: ACM Press, 2005.
customer-provider hierarchies. [9] Sobrinho, “An algebraic theory of dynamic network
An interesting direction for future research is proposing routing,” IEEE/ACM Transactions on
formal models for addressing these issues and assessing Networking, vol. 13, no. 5, pp. 1160–1173, 2005.
their impact on our necessary condition for BGP safety. [10]T. G. Griffin and G. Huston, “TRFC 4264: BGP
First, can we close the gap between our necessary condition wedgies,” 2005.
and known sufficient conditions for safe convergence? [11]L. Subramanian, S. Agarwal, J. Rexford, and R.
Second, can we develop a compositional theory for safe Katz, “Characterizing the internet hierarchy from
policies? If we put together two sub networks with unique multiple vantage points,” INFOCOM 2002. Twenty-
stable solutions, when does the combination also have a First Annual Joint Conference of the IEEE
unique stable solution? It would also be valuable to extend Computer and Comm.Societies. Proceedings. IEEE,
the convergence-time analysis to broader classes of vol. 2, pp. 618–627, 2002.
preferences, and to characterize the average-case (instead of [12] C. Labovitz, A. Ahuja, A. Bose, and F. Jahanian,
worst-case) convergence time following a network change. “Delayed internet routing convergence,” SIGCOMM
Comput. Commun. Rev., vol. 30, no. 4, pp. 175–

187, 2000.
[13] J. Feigenbaum, R. Sami, and S. Shenker,
“Mechanism design for policy routing.”
Distributed Computing, vol. 18, no. 4, pp. 293–305,
2006.
[14] H. Karloff, “On the convergence time of a path-
vector protocol,” in SODA ’04: Proceedings of the
fifteenth annual ACM-SIAM symposium on
Discrete algorithms. Philadelphia, PA, USA: Society
for Industrial and Applied Mathematics, 2004, pp.
605–614.
[15] T. G. Griffin and G. Wilfong, “A safe path vector
protocol,” in Proceedings of IEEE INFOCOM
2000, IEEE Communications Society. IEEE Press,
March 2000.
[16] H. Levin, M. Schapira, and A. Zohar, “Interdomain
routing and games,” in Proceedings of the 40th
ACM Symposium on Theory of Computing
(STOC), May 2008.
[17] A. Fabrikant and C. Papadimitriou, “The
complexity of game dynamics: BGP oscillations,
sink equlibria, and beyond,” in Proceedings of
SODA 2008.
[18] G. Huston, “Interconnection, peering, and
settlements,” in Internet Global Summit (INET).
The Internet Society, 1999.
[19]Z. M. Mao, R. Govindan, G. Varghese, and R. H.
Katz, “Route flap damping exacerbates internet
routing convergence,” in SIGCOMM ’02:
Proceedings of the 2002 conference on Applications,
technologies, architectures, and protocols for
computer communications. New York, NY, USA:
ACM, 2002, pp. 221–233.
[20]E. C. Jr., Z. Ge, V. Misra, and D. Towsley,
“Network resilience: Exploring cascading failures
within bgp,” in Allerton Conference on
Communication, Control and Computing, October
2002.
[21]K. Sriram, D. Montgomery, O. Borchert, O. Kim,
and D. R. Kuhn, “Study of bgp peering session
attacks and their impacts on routing performance,”
IEEE Journal on Selected Areas in
Communications, vol. 24, no. 10, pp. 1901–1915,
2006.
Biometrics Based File Transmission Using RSA

Cryptosystem
Mr.P.Balakumar1, Dr.R.Venkatesan2
1
Assistant Professor, Department of Computer Science and Engineering,
Selvam College of Technology, Namakkal, Tamilnadu, India.
P_balakumar@yahoo.com
2
Professor & Head, Department of Information and Technology,
PSG College of Technology, Coimbatore, Tamilnadu, India.
ramanvenkatesan@yahoo.com
change with time. For example, the iris-recognition does not

Abstract: Biometrics gives a lot of methods in high-secure change over a human’s lifetime and it is more secure than
applications while using natural, user-friendly and fast voice-identification. Uniqueness in human’s biometric is a
authentication. Most of the implementations of Public key scale of the variations or differences in the biometric model
Cryptosystems widely use the RSA algorithm. The RSA algorithm among the worldwide population. The high-level degree of
is one of the asymmetric algorithms in which we use two keys uniqueness produces more unique identifier. A low-level
private and public. The efforts in this paper merge the biometric
degree of uniqueness indicates a biometric pattern that is
concept with the asymmetric cryptography to offer the security
for the document sending process in the distributed network. For
document sending the sender encrypts the message using the found commonly in the general population. The iris and
receiver’s public key and for decryption the receiver uses his retina have higher levels of uniqueness than hand, voice and
private key. This system uses the fingerprints as the security- finger printing. The nature of an application helps in
providing medium. This system is developed under Graphical determining the degree of strength and uniqueness needed.
User Interface environment which is very easy to operate by the Living persons distinguish the biometric verification from
users. This system is developed using the Java language so that it forensics, which does not involve real-time recognition of a
can be executed on any platform. The design of this system living human being.
supports both the Internet and Intranet environments. Dynamic
key generation process is the main contribution of this work.
Keywords: Cryptography, Biometrics, RSA, DSS, KDC
1. Introduction
A biometrics system is a standard method for the identity

verification of a human being based on the personal or
physical identification of characteristics. The functions of
biometric systems are determining, measuring and
codification of the unique characteristics of individual
persons with one already recorded. In recent years there has
been rapid growth in the use of biometrics for user
authentication applications because biometric based
authentication provides several benefits over knowledge and
possession-based methods. General biometric systems Figure 1 General Biometric System.
consist of the four phases. They are, data collection which
includes sensing and pre-processing, signal analysing which Information sharing is a necessary part of our life. Hence,
includes feature extraction and template generation, storage, security of information from mishandling is need. A
and decision making with a matcher as shown in Fig. 1. cryptography mechanism provides a set of data
transformations called encryption and decryption to send the
A secured biometrics system does not change widely over a data in a secured manner. Encryption is applied to the
long time, but a less secure biometrics system is likely to normal message i.e. the data to be translated is used to
produce the code message (encrypted data) which is apart

from original data using encryption key. Decryption uses the Another significant difference concerns the matching
decryption key to convert code message to original message subsystems. A password based method always provides a
(the original data). Now, if the Encryption key and the crispy result. If the password matches, it grants access and
decryption key are same or one can be copied from the other otherwise refuses access. However the performance of a
then it is said to be symmetric cryptography. pattern detection system in general is dependent relative on
several factors such as the quality of input and enrols data
There is a drawback in symmetric cryptography. That is the along with the basic characteristics of the underlying
sender must send the same key to the receiver through algorithm. This is typically reflected in a graded overall
another secured channel. The attacker can capture it and he match “score” between the submitted biometric and a stored
could find the original secret key. This type of cryptography reference. In a biometrics-based system, they can purposely
system can be easily broken if the key used to encryption or set a threshold on the score to directly control the false
decryption is known. acceptance and false rejection rates. Inverting this, given a
good matching score the system can guarantee that the
To overcome the drawback present in the symmetric probability of signals coming from a genuine person is
cryptography we moved towards Public Key Cryptography significantly high. Such a calibrated confidence measure can
system that was found in 1976 by Whitfield Diffie and be used to tackle non-repudiation support – something that
Martin Hellman of Stanford University [22]. It uses a set of passwords cannot provide [8].
associated keys one for encryption and another one for
decryption. One key, which is known as the private key, is 3. RSA Algorithm
kept top secret by the user and another one key is public key
that is distributed to all other users. The Rivest, Shamir, Adelman (RSA) scheme is a block
cipher asymmetric cryptosystem, in which the Plaintext and
2. Security of Biometrics ciphertext are integers between 0 and n-1 for some n. A
typical size for n is 1024 bits or 309 decimal digits. In RSA
Regular biometrics can help to reduce the problems related system all the users must generate their private key
with the existing methods of user verification. The hackers KR={d,n} and kept it in secret and store their public key
will find the weak points in the existing system and attack KU={e,n} in Key Distribution Centre(KDC). The sender
the existing system accordingly. Unlike key systems, which receives the receiver’s public key from the KDC and
are able to find the message using brute-force attack, encrypts the message using the receiver’s public key. The
biometric based systems are difficult to crack. The biometric receiver uses his private key to decrypt the coded message.
systems need considerably more attempts to breakthrough. The private key is known only to the receiver himself.
Although standard encryption techniques are helpful in
many ways to avoid breach of security, there are some new 3.1 Finger Prints
types of attacks are possible. If biometric system is used as a The finger printing biometrics is an automatic digital
supervised verification tool, there may not be problems, but version of the old ink-and-paper method used for more than
in a distant unattended application, such as web oriented, e- a century for recognition, mainly by law enforcement
commerce applications, hackers may have sufficient time to agencies. Some samples of fingerprint images are shown in
make frequent attempts before being noticed or may even be the Figure.2. The Biometric device involves users placing
able to actually break the remote client [8]. their finger on a platen for the print to be read. The minutiae
are then extracted by the vendor’s algorithm, which also
2.1 Comparison to Password makes a fingerprint pattern analysis. Fingerprint template
Real benefits of biometric systems are that they are much sizes are typically 50 to 1,000 bytes.
longer in size than a password or phrase key. They vary
from hundred bytes to over a megabyte. Usually the message The Fingerprint biometrics currently has three main
content of such signals is relatively high. It is almost not application areas: Large-scale Automated Finger Imaging
possible to keep in mind a 2K password and it would take an Systems (AFIS) generally used for law enforcement
tediously long time to type in such a password anyhow purposes, fraud prevention in entitlement programs, and
(particularly with no errors). Fortunately, automated physical and computer access.
biometrics can offer the security advantages of long
passwords while still retaining the speed and simplicity of
short passwords. Still, in general smaller amount of them are
typically covered, such as dissimilarity is that there is no
“fake password” input detector equivalent to the fake
biometric.(although perhaps if the password was in some
standard dictionary it could be deemed “fake”). Additionally,
in a password or token based verification system no effort is
made to prevent replay attacks (since there is no difference
of the “signal” from one presentation to another). However,
in an automated biometric-based verification system, one can Figure 2. Sample Fingerprints
go to the extent of checking liveliness of the input signal.
3.2 Characteristics of Biometrics In voice recognition, assuming the voice and not the
pronunciation is being measured; this biometric is
Table.1 compares the seven mainstream biometrics in terms
moderately robust and distinctive. Iris scans are both highly
of a lot of properties, ranging from how robust and distinct
robust because they are not highly vulnerable to routine
[10] they are to what they can be used for (i.e., identification
changes or damages and distinctive because they are
or verification or verification alone). This table is an effort to
randomly formed. At last dynamic signature verification and
lend a hand to reader in categorizing biometrics along
keystroke dynamics are not robust or distinctive.
important dimensions. Because this industry is still
functioning to establish comprehensive standards and the
technology is varying rapidly, however, it is difficult to make 4. Problem Statement
assessments with which everyone would agree. The table
shows an assessment based on consideration with Even though the RSA algorithm uses the finger printing
technologists, vendors, and program managers. The table is biometrics system to generate the public key and private key
not proposed to be an aid to those in the market for generation there are some problems in that approach.
biometrics; rather it is a guide for the unskilled.
They are:
1. Brute-force attack: The maximum size of the public key
Table.1 Comparison of Mainstream Biometrics and private key obtained by RSA algorithm is 155
digits. It can be captured by a brute force attacker using
Identify thousands of machines and it requires three month of
Biometric versus Robust Distincti Intrusive computation. {Ref: Journal of Telecommunications and
Verify ve Information Technology. Volume 4/2002. Pages 08-
09}.
2. Increased key storage requirement: RSA key storage
Fingerprint Either Medium High Touching (private keys and public key) requires significant
amounts of memory. So, we have to store the public
Hand Verify Medium Low Touching key and private key in any equipment or in brain.{Ref:
Journal of Telecommunications and Information
Facial Either Medium Medium 12+ Technology. Volume 4/2002. Pages 41-56}.
inches 3. No Dynamic key generation: There is no dynamic key
generation in RSA algorithm. Therefore the user must
Voice Verify Medium Low Remote keep secretly his private key. There is a chance to lose
or stolen, forgotten the private key of the RSA
algorithm, hence he may lose the data.
Iris Scan Either High High 12+
inches
5. Proposed Scheme
Retinal Either High High 1–2 The architecture of the proposed scheme is shown in
inches Figure.3. The client generates the public key and sends to
the KDC. On document send process it retrieve the
Keystroke Verify Low Low Touching receiver’s public key from KDC and encode the data with
Dynamics aid of generated public key. Then it sends the encoded data
to the receiver. While viewing the document it dynamically
When comparing ways of using biometrics, half can be used generates the private key which is used to decode the
for both the identification and verification, and the encoded data.
remaining can only be used for authentication. In specific,
hand geometry has only been used for confirmation The proposed digital signature algorithm is a version of the
applications, such as physical access control and time and RSA algorithm that overcomes the problems in the RSA
attendance verification. Adding to this, voice detection system. A brute force attacker can able to hack the private
because of the need for staffing and matching using a pass- key by using every possible combination of the key (i.e.
phrase, is used for verification only. Numeric key). In our system, we use alpha numeric
(combination of alphabets and numeric) keys, hence the
There is considerable changeability in terms of robustness attacker can not able to obtained the key values easily.
and individuality. Fingerprinting is sort of robust, and, even
though it is distinctive, a small proportion of the population The second problem in the existing RSA algorithm is key
has unusable prints, always because of age, genetics, injury, storage requirement. In our proposed system we generate the
career, spotlight to chemicals, or other occupational hazards. private key dynamically. Hence there is no need for key
Hand/finger geometry is moderate on the distinctiveness storage requirement. The third problem in the existing
scale, but it is not very robust, while facial recognition is system is no dynamic key generation. Normally, by using
either highly robust or distinctive. RSA algorithm they have to generate their public key and
private key. Then they have to send the public key to the key
distribution centre and keep their private keys secretly with existing key value is replaced from the list and the new key
themselves. value is added into the list.
In our proposed algorithm we generate the public key using 6.2 Client
the finger print and send that public key to the key The client application is designed to hold the document
distribution centre. While encrypting the data the sender get transfer process and the key generation process. The client
the public key of the receiver from the key distribution centre application is divided into four modules. They are the Key
and encrypt the data with that public key. generation module, the sender module, the receiver module
and the document view module. The key generation module
To decrypt the ciphertext the receiver requires his private generates the key from the finger print data. The sender
key. At that time of decryption only the receiver will be able module is used to encode and send the document. The
to know his private key. This process is called ‘Dynamic receiver module receives the decoded documents that are
private key generation’. sent by the other clients. The document view module
Key Update maintains received documents after the decoding process the
Client user can view the document.
Pub Key
Generation 6.3 Key Generation Module
The key generation process is shown in Figure.4. This
Document module is to generate the public key by using the finger print
Send Key KDC
Retrieval data. The input data is given as an image to the system. This
value is to create the key base value that is used to generate
the public key value.
Encode Send
Capture the finger print data

Document
View Decode
Private Key
Generation
Dot matrix Conversion

Display
Figure 3. Architecture of Proposed Scheme Generate key Base
6. Key Distribution Centre

Generate public key
KDC has a very significant role in the asymmetric key
cryptosystem. It receives public key values from the clients
and stores in its locale. It is the only authoritative system to Send the key to KDC
distribute the public key values to the requesting users. The
KDC application is a server application. The KDC
application has two modules. One is key management Figure 4. Key Generation
module and another one is key distribution module. The key
management module is mainly for receiving and The public key value in KDC is stored with its client details.
maintaining the key values. On other hand the key The system supports the JPEG and the GIF image formats.
distribution module distribute the public key value based on The pixel matrix is constructed using image data. The key
the client requests. base is generated by using the image data matrix values. The
system has been implemented as a GUI based application
6.1 Key Management Process developed in Java. The main menu has three options. They
The key management module is created to perform the key are the key preparation, document list and send process. The
maintenance process. It has two main tasks. They are the key key generation windows receive the input for the finger print
receive process and key expiry management process. The image file. The generate button is used to initiate the key
key receive process is run as a separate thread. The KDC generation process. The send button is used to start the key
listen all the key value and send a key for the client request. transfer process. The key distribution centre is designed to
For the receiving process, it needs UDP socket. It does not receive and maintain all the public key values. The message
make any connection with the client application. This sending process is used to transfer a file from one client to
module maintains the entire received public key values. The other. The message file is encoded before the sending
key expiry management module keeps the validity of the key process. All the received messages are listed in the inbox.
values. KDC automatically removes the key values from the The user can select the file and perform the view process.
key list if the client application process is terminated. The The documents are decoded before the view process. The
clients can change their key value and update. So the private key value is generated at the time of the decoding
process. The decoded documents are stored in the specified Using Java language under windows platform develops the
folder simulation tool. Java supports multiple platforms, GUI
design and network operations. Using the Java language
7. Testing and Implementation develops the system. Image processing, cryptographic
operations, network transmissions and file processing are the
Testing is the important phase in the system development major are in the system. Java provides a variety of packages
process. The system is developed as a GUI based application. and classes to support all these tasks. The user interface is
The system is tested before the implementation process. The designed with GUI support. The application is designed to
system is tested with different testing methods. They are unit run under any platform. The finger print values are
test, integration test, system test, validation test and stress retrieved from image files. The image file data are extracted
test. The system is tested with different network and and converted into pixel matrix. Using the classes such as
platform environments. The system uses the image scanner Image, MediaTracker and PixelGrabber in Java the system
to capture finger print image data. The system is tested in does these processing. The Image class is used to convert an
the Intranet environment. In this system each and every image into an object. The Media Tracker and the Pixel
modules is tested separately for the unit test. For example Grabber classes are used to support the data extraction and
the RSA algorithms processes key generation, encode and pixel conversion process. These classes are available in the
decode operations are tested with the corresponding java.awt package.
modules.
Java provides a separate package JCE for the cryptography
The Client application and the Key distribution applications process. But the JCE requires the Service Providers for the
are tested separately. The integration test is performed after implementation. In Java cryptography can be implemented
all the modules are connected with the main menu. The in two ways. They are using the JCE with service providers
entire system is tested with all the operations by using a set and the other one write the code for the cryptographic
of finger print values. The stress test is conducted to test the algorithms. In this work the second method is applied. The
load management strength of the client application and the RSA algorithm is implemented by using the java.math
key distribution centre application. Connecting multiple package support. RSA requires high bit length data type
clients with the KDC tests the key distribution centre stress. support. Java provides a class Big Integer to process values
In the client application sending a large file to the other in 128 bits. All RSA key generation, encoding and decoding
client tests the strength. The validation test is performed for operations are done by using the Big Integer class.
all input values. The finger print image availability is
checked before the key base generation process. The file process and the data transmission process are
implemented with the support of java.io and java.net
The system is developed to distribute document with security package. All the files are processed using the byte stream
using the biometrics. The system is tested and the results are classes. The data transmission tasks are done using the
very good. The implementation of the system is conducted TCP/IP support classes in the java.net package. The key
as direct change over mechanism. The new system is directly distribution centre application is designed using the UDP.
installed and activated into the action for usage. The system The client application transfers the files using the Server
can be implemented in any network environment. The Socket and Socket classes. Data gram Socket and Data gram
system supports all type of file transfer operations. The Packet classes are used in KDC process.
system has developed as two applications the key
distribution centre and the client application. The key 8. Conclusion
distribution centre application is loaded into a separate
machine. The client application can be loaded into all other The System is developed to provide security for the file
client machines. transfer process in distributed environment. Document
transmission between the systems that are in the distributed
All the client application should be configured for the key environment is a usual task. The same environment is also
distribution centre IP address for the key updating and shared by a lot of members. So the system should ensure the
request process. The system now designed to get the security of the documents that are transferred. Different
fingerprints images from the image file. So the system must cryptographic techniques are used to secure the data. In the
be connected with an image scanner. The system can also be recent days, biometrics is used to recognize the users. This
connected with the fingerprint image scanner devices. The work combines the biometrics and cryptography to provide
client application and the key distribution applications can the security for the document transmission process in the
be continuously executed to maintain the connection and distributed environment. Generally passwords and smart
message receive process. All the messages are directly cards are used for the security systems.
received by the client applications. The system requires a
lesser amount of hard disk space to store the received and The system uses the biometrics technology as the security-
decoded documents. The key distribution centre should be providing medium. This system uses the fingerprints for the
connected with all the client applications. The system can be security system. Password can be hacked by trial and error
run with one or more network environments. basis. But it is not possible to break the biometrics based
security system. The system is developed as two applications.
7.1 Software Selection They are the key distribution centre application and the
client application. The KDC supplies the public key values [13] www.mit.bme.hu/events/minisy2003/papers/orvos.p
to the required clients. The client application is designed to df
handle all the data transfer and security operations. [14] http://rpmfreelancer.no
ip.com:8080/duncan21/biometrics/finger.html
The system uses a designed key base generation algorithm [15] www.cost275.gts.tsc.uvigo.es/presentations/COST2
and RSA algorithm. The system is tested with various 75_Jain.pdf
samples and clients. The performance of the system is very [16] www.research.ibm.com/ecvg/pubs/sharat-
good. The system is tested with different type of file formats. proc.pdfM. Wegmuller, J. P. von der Weid, P.
The result shows that the system supports all types of file Oberson, and N. Gisin, “High resolution fiber
format. The system stores and distributes the public key distributed measurements with coherent OFDR,” in
values for all clients in the key distribution centre. The Proc. ECOC’00, 2000, paper 11.3.4, p. 109.
system does not require any key storage process for the [17] R. E. Sorace, V. S. Reinhardt, and S. A. Vaughn,
secret key. In this work using the fingerprint values, the “High-speed digital-to-RF converter,” U.S. Patent 5
system can generate both the public key and private key. 668 842, Sept. 16, 1997.
Damages that occurred in the finger print may impact the [18] The IEEE website. (2002) [Online]. Available:
recovery of the documents. http://www.ieee.org/
[19] M. Shell. (2002) IEEE Transaction homepage on
In future the system can be implemented for all type of CTAN. [Online]. Available:
authentication process such as capillary patterns in the http://www.ctan.org/tex-
retina, hand geometry, facial characteristics, signature archive/macros/latex/contrib/supported/IEEEtran/
dynamics, voice pattern, and timing of keystrokes. Data [20] W.Diffie and M.Hellman.” New Directions in
compression technique can be used to reduce the content Cryptography”. IEEE Transaction on Information
size, process time and transmission time. In future the Theory.IT-22(1978).472-492.
system will include noise detection and filtering facility for
the input process.
Authors Profile
References
Mr.P.Balakumar received the B.E. and
[1] Bruce Schneier, “Applied Cryptography Protocols, M.E. degrees in Computer
Algorithms” 2nd Edition, Wiley publication. Science and Engineering from PSG
[2] Naughton.P and H.Schildt, “Java 2: The Complete College of Technology, Coimbatore, in
Reference” , McGraw-Hill,1999 1997 and Anna University, Chennai in 2004
[3] William Stallings, “Cryptography and Network respectively. During 1999-2001, he worked
Security Principles and practice”, 2nd Edition, as Lecturer in PSG College of Technology in
Prentice Hall, Upper Saddle River. Coimbatore. Later during 2003-2008, he worked as Lecturer &
[4] Anil Jain, Lin Hong, Sharath Pankanti, and Ruud Assistant Professor in AMS Engineering College, Namakkal. He
now with Selvam College of Technology, Namakkal, Tamilnadu,
Bolle, “ An Identity Authentication System Using
India as Assistant Professor in Department of Computer Science
Fingerprints” Department of Computer Science, and Engineering.
Michigan State University East Lansing.
[5] James L. Wayman, “ Biometrics Identification” , Dr.R.Venkatesan was born in Tamilnadu,
Communications of the ACM, February 2000. India, in 1958. He received his B.E (Hons)
[6] Katrin Franke, Javier Ruiz-del-Solar, Mario, “ Soft- degree from Madras University in 1980. He
Biometrics: Soft-Computing for Biometric- completed his Masters degree in Industrial
Applications” Dept of Pattern Recognition, Engineering from Madras University in
Fraunhofer IPK, Berlin, Germany. 1982. He obtained his second Masters
degree MS in Computer and Information
[7] Nalini K. Ratha, Jonathan H. Connell, and Ruud
Science from University of Michigan, USA
M. Bolle J, “ An Analysis of Minutiae Matching in 1999. He was awarded with PhD from Anna University, Chennai
Strength” Watson Research Center. in 2007. He is currently Professor and Head in the
[8] Rowley. T, “ Silicon Fingerprint Readers: A solid Department of Information Technology PSG College of
state approach to biometrics” , Proc. of the Card Technology, Coimbatore, India. His research interests are in
Tech / Secure Tech, Orlando, Florida, May 97. Simulation and Modeling, Software Engineering, Algorithm
[9] Schneier.B, “ The uses and abuses of biometrics” . Design, Software Process Management.
Communications of the ACM, August 1999.
[10] Schneir.B, “ Security pitfalls in cryptography” ,
Proc. of Card Tech /Secure Tech, Washington D.C.,
April 98.
[11] Wong C K and Lam S S, “Digital Signatures for
flows and multicasts”, IEEE/ACM Transaction and
Networking”, August 1999.
[12] www.rand.org/publications/MR/MR1237/MR1237.
appa.pdf
Artifact Extraction and Removal from EEG Data

Using ICA
Ashish Sasankar1, Dr. N.G Bawane2 ,Sonali Bodkhe3 ,M.N.Bawane4
1
G.H.Raisoni College of Engineering
CRPF Gate 3 , Digdoh hills Hingna road Nagpur (India)
Ashish_sasankar@yahoo.com
2
narenbawane@rediffmail.com
3
sonalibodkhe@raisoni.com
4
Govt Polytechnic,Nagpur(India)
madhuribawane@rediffmail.com
In this paper we will focus on extracting and removing

Abstract: The Independent Component Analysis (ICA) has
emerged as a novel and promising new tool for performing artifacts from recorded EEG data using Independent
artifact corrections on EEG data. In this paper, ICA is used to Component Analysis.
perform artifact correction on three types of artifacts namely,
frontal (eye), Occipital (rear-head), and muscle. EEG is 2. EEG Database and Preprocessing
composed of electrical potentials arising from several sources.
Each source (including separate neural clusters, blink artifact or The BIOMED team and department of Clinical and
pulse artifact) forms a unique topography onto the scalp – ‘scalp Experimental Neurology, both at Katholieke Universiteit
map‘. Scalp map may be 2-D or 3-D.These maps are mixed Leuven (KUL) (Belgium) have given public access to two of
according to the principle of linear superposition. Independent their long-term EEG recordings from patients suffering from
component analysis (ICA) attempts to reverse the superposition Mesial Temporal Lobe Epilepsy. Patient was 35 years-old
by separating the EEG into mutually independent scalp maps, or male. The data was collected from 21 scalp electrodes placed
components.
according to the international 10-20 System with addition
Keywords: EEG, Independent Component Analysis(ICA), BCI electrodes T1 and T2 on the temporal region. The sampling
frequency was 250 Hz and an average reference montage
1. Introduction was used. The electrocardiogram (ECG) for each patient was
also simultaneously acquired and is available in channel 22
EEG stands for Electroencephalogram. It senses electrical of each recording
impulses within the brain through electrodes placed on the Under this system, the EEG electrodes are placed on the
scalp and records them. It is a recording of brain activity, scalp at 10 and 20 percent of a measured distance. For
which is the result of the activity of billions of neurons in the example, if a circumference measurement around the skull
brain. EEG can help diagnose conditions such as seizure was approximately 55 cm, a base length of 10% or 5.5 cm
disorders, strokes, brain tumors, head trauma, and other and 20% or 11.0 cm would be used to determine electrode
physiological problems. The pattern of EEG activity changes locations around the skull. The skull may be different from
with the level of a person's arousal. A relaxed person has patient to patient but the percentage relationships remain the
many slow EEG waves whereas an excited person has many same. Figure 1 shows a typical 10–20 electrode placement
fast waves. A standardized system of electrode placement is looking down on the skull. Each site has a letter and a
the international 10-20 system. A common problem with number or another letter to identify the hemisphere
EEG data is contamination from muscle activity on the location.The letters Fp, F, T, C, P, and O stand for Front
scalp. It is desirable to remove such artifacts to get a better polar, Frontal, Temporal, Central, Parietal and Occipital
picture of the internal workings of the brain. In latest respectively. Even numbers (2, 4, 6, 8) refer to the right
publication[20] a new method is proposed for EEG signal hemisphere whereas odd numbers (1, 3, 5, 7) refer to the
classification in BCI systems by using nonlinear ICA left hemisphere. The z refers to an electrode placed on the
algorithm. An ICA based EEG feature extraction and midline. The smaller the number, the closer the position is to
modeling approach for person authentication is presented in the midline.
[21]. Efficient use of modern DSP and Soft computing tools
in the area of medical diagnosis is recently covered in [22].
3) Muscle artifacts – dispersed throughout the brain .
3.2 Assumptions for the ICA model
The following assumptions ensure that the ICA model

estimates the independent components meaningfully.
Actually the first assumption is the only true requirement
which ICA demands. The other assumptions ensure that the
estimated independent components are unique.
Figure 1. Typical electrode placements under the (1) The latent variables (or independent components) are
International 10 –20 system statistically independent and the mixing is linear.
(2) There is no more than one gaussian signal among the
The original EEG in f1.EDF from second 7500 to second latent variables and the latent variables have cumulative
8100 (f1_750to810.set, 12 Mbytes). This EEG frame density function not much different from a logistic sigmoid .
contains a seizure[18][19]. (3) The number of observed signals, m, is greater than or
equal to the number of latent variables, n (i.e. m _ n).
3. Independent Component Analysis(ICA) If n > m, we come to a special category of Independent
Component Analysis called ICA with over-complete bases .
Independent Component Analysis (ICA) is one of a group of In such a case the mixed signals do not have enough
algorithms to achieve blind separation of sources [Jutten & information to separate the independent components. There
Herault 1991]. ICA finds an unmixing matrix which linearly have been attempts to solve this particular problem but no
decomposes the multichannel EEG data into a sum of rigorous proofs exist as of yet . If m > n then there is
maximally temporally independent and spatially fixed redundancy in the mixed signals. The ICA model works
components. These Independent Components (ICs) account ideally when n = m.
for artifacts, stimulus and response locked events and (4) The mixing matrix is of full column rank, which means
spontaneous EEG activity. One of the standard applications that the rows of the mixing matrix are linearly independent.
of ICA to EEG includes artifact detection and removal . If the mixing matrix is not of full rank then the mixed
Selected components responsible for artifacts are set to zero signals will be linear multiples of one another.
and all other ICs can be projected back onto the scalp (5) The propagation delay of the mixing medium is
yielding EEG in true polarity and amplitudes. Related negligible.
approaches to magneto encephalographic signals can be
found . Some simple neural network algorithms cane blindly 3.3 The ICA model applied to EEG Data
separate mixtures, of independent sources. On maximizing
the joint entropy(y), of the output of neural processor In case of EEG signals we have m-scalp electrodes picking
minimizes the mutual information among the output up correlated brain signals where we would like to know
components, yi = g(ui), where g(ui) is an invertible bounded what effectively independent brain sources produced these
nonlinearity and u=Wx, a version of the original sources. signals. The ICA model appears well suited for this scenario
ICA is suitable for performing blind source separation on because it satisfies most of the model assumptions
EEG data because: (1) it is possible that EEG data recorded considered in section 4. Start with assuming that EEG data
at multiple scalp sensors are linear sums of temporally can be modeled as a collection of statistically independent
independent components arising from spatially fixed, brain signals. Assumption (5) is valid since volume
distinct brain or extra-brain networks, and, (2) EEG data by conduction in the brain is effectively instantaneous and
volume conduction does not involve significant time delays. assumption (2) is plausible . In this paper, it will attempt to
In EEG analysis, the rows of the input matrix x are the EEG separate the m-observed EEG signals into n-statistically
signals recorded at different electrodes, while the columns independent components (thus satisfying assumption (3) and
are measurements recorded at different time points. (4)). However, it is questionable to assume whether EEG
data recorded from m-electrodes is made up of exactly n-
3.1 Types of artifacts statistically independent components since it ultimately
cannot know the exact number of independent components
Severe contamination of EEG activity by artifacts such as embedded in the EEG data. Nonetheless, this assumption is
eye movements, blinks, head movements, muscle, and line usually enough to identify and separate artifacts that are
noise create a problem for proper EEG interpretation and concentrated in certain areas of the brain such as eye,
analysis. The three types of muscle artifacts studied in this temporal, and occipital artifacts . The ICA model tends to
paper are: have a more difficult time in separating artifacts that are
1) Eye artifacts – project mainly to the frontal side more spaced out over the scalp such as muscle artifacts.
2) Rear head artifacts – project mainly to the occipital
Side
While steps< Maxsteps
For i=1:EndofBlock
u=weightsxdata+w0
y=1/(1+exp(-u))
weights=weights+Irate*(I-2yu)*weights
Weights
=> Yes wts_blowup=1
maxWts? noChange=1
No
wts_blowup End for loop

No
? ?
Yes Yes
Yes
oldwtchange=weights-old_wts
step++
~wts_blowup angledelta=0
Yes
? delta=oldwtchange
change=oldwtchange 2
No
step=0;change=nochange;
wts_blowup=0
block=1;
Irate=Irate*lowerIrate%restart with
lower Irate
weights=identity matrix;
old wts=weights;
Yes
2
3
1
3 2
Irate=nochange display("matrix may not be

No
? invariable")
Yes
rnk=rank(data)
rank<chans Display('data has rank<rnk>,Channels

Yes
? are not independent')
No
RETURN
Display("lower learning rate to <Irate> and starting again")
angledelta=cos-1   delta * olddelta  

 
  change * nochange  
Step>2 ? Yes
No
display('setp<step>-
Irate<Irate>,wchange<change>')
oldwts=weights;
Irate=IratexlowerIrate;
angledelta>60? Yes olddelta=delta;
oldchange=change;
No
olddelta=delta;
step==1? Yes
oldchange=chnage;
step>2 & laststep=step;

Yes
change<nochange? step=maxsteps;
No
change>maxWts? Yes Irate=Irate*lowerrate2
No
End W=weights RETURN

Yes
while?
Figure 2. The Bell & Sejnowski Infomax Algorithm

flowchart.
valued random variables in place of the scalar valued

3.4 THE ICA ALGORITHM functions.
Now here we present a brief derivation of the Bell Sejnowski Consider a scalar-valued function x with a gaussian fx(x)
Information Maximization algorithm. It has been consider a that passes through a transformation function y = g(x) to
simple case of a one-input one-output system to derive the produce the output with fy(y) (Figure 3). This is analogous
ICA algorithm. The general multi-input multi-output system to our matrix operation:
is similarly derived with n-dimensional matrices of vector-
Y = WX partial of H(y) with respect to w. The second term in eq.(3)

For our work with EEG data we will take the transformation does not depend on w and therefore can be ignored. The
function y to be the logistic change in slope, ∆ w, necessary for maximum change in
sigmoid function defined as: entropy is then:
1
y = g ( x) = ∂H ( y) ∂  ∂y 
1 + e −u ∆ wα =
∂ w  ∂ x 
E  ln (4)
u = wx + w0 ∂w
where w = slope of y (also called the weight)

w0 = bias weight to align the high density parts of the input
with y (Refer Figure 2.) We now come to an important step. We would like to
compute the derivative, but we cannot compute the
expectation. Hence, we make the stochastic gradient
approximation:
 ∂y  ∂y
E  ln  ≈ ln
 ∂x  ∂x
to get rid of the expectation [4],The equation then

simplifies to:
−1
∂H(y) ∂  ∂y   ∂y ∂  ∂y
∆wα = ln  =    
∂w dw ∂x   ∂x ∂w ∂x
(5)
Figure 3. Transformation of the fx(x), of x when x is mixed
with a sigmoid mixing function
The above equation is the general form of the weight change
An increase in the joint entropy of the output, H(y),means a rule for any transformation function y. For the logistic
decrease in its mutual information. The entropy of the output sigmoid function eq.(1), the terms in eq .(5) are evaluated
is maximized when we align the high density parts of pdf of as:
x with the high sloping parts of the function g(x) (hence the
need for the biasing weight w0). The function g(x) is ∂y
monotonically increasing (i.e. has a unique inverse) and thus = wy (1 − y) (6)
the output fy(y) can be written as a function of the input fx(x) ∂x
as:
( y) =
f (x) ∂  ∂y 
  = y(1 − y)(1 + wx (1 − 2 y))
x
f
∂y
y
(7)
∂w  ∂x 
∂x
(1) Substituting the above equations into eq.(5) gives the weight
The entropy of the output is given by, update rule for y = logistic sigmoid function:
∞
H ( y) = −E{ln f y ( y)} = − ∫ f y ( y) ln f y ( y)dy (2) ∆ w α w − 1 + (1 − 2 y ) x (8)
−∞
Similarly, the bias weight update, w0, can be evaluated as:

Substituting (1) into (2) gives,
 ∂y  ∆ w 0α 1 − 2 y
H ( y ) = E  ln  − E {ln f x ( x )} (3) (9)
 ∂x 
Following similar steps we can derive the learning rules for
We now would like to maximize H(y) of eq.(3) for statistical multivariate data for a sigmoid function:
independence. Looking at the right hand side we see that the
function x is fixed and the only variable we can change is y.
Or more preciously, the slope, w, of y. Hence we take the ∆Wα W T [ ] −1
+ (1 − 2 y) xT (10)
∆ w 0α 1 − 2 y (11)
4. MatLab Implementation
Equations (10) and (11) give the learning rules for updating
the weights to perform ICA. Implementing them directly
into Matlab will involve performing the inverse function,
which is computationally very intensive. We therefore
modify eq.(10) by multiplying it by WTW (this does not
change anything since W is orthogonal):
∂H ( y) T
∆Wα W W
∂W
[ ]
⇒ ∆Wα ( W T
−1
− (1 − 2 y) xT )W TW Figure 4. EEG data from Data Set (f1.set)
⇒ ∆Wα ( I + (1 − 2 y) xTW TW (12)
5.1 Independent Components:
⇒ ∆Wα ( I + (1 − 2 y)u T )W, Execution of the data is processed through Matlab function of
EEG toolbox[19]. The resulting independent components are
shown in figure(5).
Where u=xW
The bias weight update rule remains the same:
∆w0α (1 − 2 y) (13)
The proportionality constant in eq.(12) and (13) is called the

learning rate (lrate).
In summary, the following two weight update rules are used
to perform ICA in Matlab:
[
Wnew = Wold + Irate ( I + (1 − 2 y)u T )W ]
(14)
W0 new = W0 old + Irate (1 − 2 y ) (15) Figure 5. Independent Components of Dataset(f1.set)
5.2 Topographical Projections

Where The topographical projections of independent components
Irate= Learning Rate ; are shown in figure (6).
W=weight matrix ;
W0=bias weight ;
I=Identity matrix;
y=logistic sigmoid ;
u=W x data+w0 ;
5. Result Discussion
Data Set f1.set considered in the paper contains 600 seconds
of data with sampling frequency Fs=250 Hz. There are 21
channels of data. The data was collected from electrodes
placed on the scalp at standard locations using the
international 10-20 system[4]. The EEG data is plotted using
function implemented in Matlab[18] and is depicted in
figure(1).
This data contains a seizure onset around 300 onwards on T3- Figure 6. Independent Components with their respective
T5 channel with the appearance of rhythmic waves. Occipital topographical projection of Dataset(f1.set)
artifacts on O1and O2 .Eye blink artifact are on Fp1 and Fp2
and Muscle Artifact are on all channels.
5.3 Corrected EEG Data Monitoring in Epilepsy,volume 37. Elsevier,

Selected the Right ,Left and Frontal Artifact were removed Amsterdam, eeg supplement no. 37 edition, 1985.
from EEG data using ICA technique. The resulting artifact [5] H. G. Wieser. Monitoring of seizures. In M. R.
corrected EEG data is shown in figure 7. A Comparison with Timberland and E. H. Reynolds, editors, What is
the original EEG data (figure 4) clearly shows that the Epilepsy? The Clinical and Scientific Basis of Epilepsy,
identified muscle artifacts have been greatly reduced pages 67–81. Churchill Livingstone, London, 1986.
[6] Anthony J Bell & Terrance J Sejnowski, An

information-Maximisation approach to blind
separation and blind deconvolution, Neural
Computation, 7,6, 1004-1034 (1995).
[7] Dominic Chan, Blind Signal Separation, PhD
Dissertation, University of Cambridge Jean-Francois
Cardoso Blind Signal Separation: Statistical principles
Proceedings of the IEEE, vol 9, no 10. p 2009-2025 Oct
1998.
[8] Scott Makeig, Independent Component analysis of
Electroencephalographic Data Advances in Neural
Information Processing Systems 8’’. MIT Press,
Cambridge MA 1996 .
[9] Foldvary N, Klem G, Hammel J, Bingaman W, Najm I,
Lüders H.(2001) The localizing value of ictal EEG in
focal epilepsy. HXURORJ\ 57:2022-2028.
[10] Klass DW.(1995) The continuing challenge of artifacts
Figure 7. Corrected EEG Data of Dataset (f1.set) in the EEG. 35: 239–269.
[11] B. Boashash, M. Mesbah, and P. Colditz, "Time
Frequency detection of EEG abnormalities," chapter
6. Conclusions 15, Article 15.5. pp. 663-669, Elsevier 2003.
[12] R.G. Andrzejak, G.Widman, K. Lehnertz, C. Rieke, P.
It is clear that Independent Component Analysis is well suited David, C.E. Elger, "The epileptic process as nonlinear
to perform artifact correction on EEG data. The topographical deterministic dynamics in a stochastic environment: an
views provided the first clues as to which components might evaluation on mesial temporal lobe epilepsy," Epilepsy
be artifacts. These plots together with the time plots of the Res. Vol. 44, pp. 129-140, 2001.
independent components were used to identify the eye and
[13] H.D.I. Abarbanel, R. Brown, and M.B. Kennel,
occipital artifacts. One of the unique properties of ICA is that
"Lyapunov exponents in chaotic systems: Their
it can eliminate the artifacts alone without disturbing the
importance and their evaluation using bserved data,"
surrounding EEG activity. An alternate approach for artifact
extraction could be simply subtracting the frontal, temporal, International Journal of Modern Physics, vol. 5(9),
and occipital readings from the EEG data. But this would pp.1347-1375, 1991
lead to considerable loss in collected information. [14] A. Wolf, J.B. Swift, H.L. Swinney, and J.A. Vastano,
The muscle artifacts appearing on all channels for Dataset "Determining Lyapunov exponents from a time series,"
(f1.set) after a seizure onset could not be removed or reduced Physica D, vol. 16(3), pp. 285–317, 1985.
significantly. One reason could be that these artifacts are not [15] P. Grassberger, T. Schrieber, "Nonlinear time sequence
concentrated in any one region alone and hence ICA analysis," Int. J. Bifurcat. Chaos 1 (3), pp.512-547,
algorithm cannot interpolated them as originating from any 1991.
single electrode. ICA technique will be useful in various BCI [16] N. Kannathala,b, M. L. Choob, U.R. Acharyab, P.K.
application such as mental task detection, detecting various Sadasivana "Entropies for detection of epilepsy in
brain disorder such as epilepsy and Image preprocessing etc . EEG," Computer Methods and Programs in
References Biomedicine 2005.
[1] M. Ungureanu, C. Bigan, R. Strungaru, V. Lazarescu, [17] A. Subasi, "Epileptic seizure detection using dynamic
Independent Component Analysis Applied in Wavelet network," Expert Systems with Applications,
Biomedical Signal Processing, Measurement Science vol. 29, pp 343–355, 2005
Review, Vol. 4, Section 2, 2004. [18] De Clercq,W,Vergult, A.,Vanrumste B.,Van
[2] Arnaud Delorme, Scott Makeig, EEGLAB: An open Paesschen,W.,and Van Huffel, S.’Canonical
source toolbox for analysis of single-trial EEG dynamics Correlation analysis applied to remove muscle
including independent component analysis, Journal of artifacts from the electroencephalogram’, IEEE
Neuroscience Methods, 134, pp. 9-21, 2004. T.Biomed.Eng.2006;53:2583-2587.
[3] In J. Laidlaw, A. Richens, and D. Chadwick, editors, [19] ergult,A,Delercq,Q.,Palmini,A.,Vanrumaste,B.,Dupont
ATextbook of Epilepsy, pages 1–22. Churchill P.,Van Huffel,S.,Van Paesschen,W.’Improving the
Livingstone, London, 4th edition, 1993. Interpretation of Ictal Scalp EEG:BSS-CCA Algorithm
[4] J. Gotman. Seizure recognition and analysis. In J. for Muscle Artifact Removal, Epilepsia
Gotman, J.R. Ives, and P. Gloor, editors, Long-term 2007;48(5):950-958.
[20] Farid Oveisi,”EEG signal classification using nonlinear

independent component analysis”,icassp,pp.361-
364,2009 IEEE International Conference on Acoustics,
Speech and Signal Processing,2009.
[21] Chen He Wang ,J.”An Independent Component
Analysis(ICA) Based Approach for EEG Person
Authentication” , Proc. Of Bioinformatics and
Biomedical Engineering ,IEEE explore , pp 1-2,2009.
[22] Prabhakar Khandait,Narendra Bawane,Shyam Limaye,
“Efficient ECG Signal Analysis using Wavelet
Technique for Arrhythmia Detechtion: An ANFIS
Approach”, Proc. Of SPIE, Vol 7546,pp 75461G1-
G6,2010.
Author Profile
Ashish B. Sasankar received the MCA and M.phil

degrees in Computer Science from HVPM Amravati
University in 1996 and 2008,respectively. He is currently
pursuing M.Tech in Computer Science from GHRCE,RTM
Nagpur University (India).
Analysis of Three Phase Four Wire Inverter for UPS

Fed Unbalanced Star Connected Load
R.Senthil Kumar1, Dr. Jovitha Jerome2 and S.NithyaBhama3
1
Department of Electrical and Electronics Engineering,
Bannari Amman Institute of Technology Anna University, Tamil Nadu India
ramsenthil2@gmail.com
2
Department of Control and Instrumentation Engineering,
PSG College of Technology, Anna University, Tamil Nadu India
jjovitha@yahoo.com
3
Department of Electrical and Electronics Engineering,
Bannari Amman Institute of Technology Anna University, Tamil Nadu India
nithyabhamaeee@gmail.com
a connection to the neutral terminal should be provided by

Abstract: A Three phase inverter with a neutral connection i.e., adding an extra wire to the inverter.
three phase four wire inverter is proposed. The uninterruptible
The load neutral terminal can be connected to the inverter
power supply (UPS) system is fed by three phase four wire
inverter and the load neutral point voltage is low to meet the using two topologies:
requirement of the system. The four leg inverters effectively • Three phase four-wire, in which the neutral point is
provide the neutral connection in three phase four wire system. connected directly to the midpoint of the supply by
They are used in many applications to handle the neutral current means of a capacitor divider.
caused by the unbalanced and non-linear load. The unbalanced • Three phase four-leg, employing an additional inverter leg
load becomes non-linear, where the neutral of the loads are that permits to modify the neutral point voltage.
accessible. The four leg inverter produces the three output
voltages independently with one additional leg.
The first topology is certainly simplest one, but the three-
The main feature of a three phase inverter, with an additional phase inverter turns into three independent single-phase
neutral leg, is its ability to deal with load unbalance in a system. inverters. As consequence, zero-sequence harmonics are
The goal of the three phase four leg inverter is to maintain the generated; moreover, especially when the load is unbalanced
desired sinusoidal output voltage waveform for all loading or non-linear, a high voltage ripple over supply capacitors is
conditions and transients. The neutral connection is present to produced by neutral currents. A further limitation is
handle the ground current due to unbalanced loads. The
represented by the maximum voltage value that the
feasibility of the proposed modulation technique is verified by
MATLAB/SIMULINK. amplitude of each phase fundamental harmonic can reach.
The second topology requires additional power switches
Keywords: Four wire inverter, Rectifier, THD, UPS. and a more complex control strategy, but it offers different
advantages, such as an increased maximum output voltage
1. Introduction value, a reduction of neutral currents and the possibility of
neutral point voltage control [5-7].
The primary function of an UPS is to maintain a constant
The block diagram for the four wire inverter for online
voltage and constant frequency supply for critical loads,
UPS as shown in Figure 1.
irrespective of variations in the input source or load
condition [2]. The way of providing a neutral connection for
three phase four wire systems using a four leg inverter
topology by tying the neutral point to the mid point of the
fourth neutral leg. The three phase four inverter has more
control flexibility, because two additional power switches
3
doubles the number of inverter output states from 8(=2 ) to
4
16(=2 ).This allows to improve the output waveform
quality.
In the medium or low power UPS; an output transformer
Figure 1.Block diagram for four wire inverter
is used to mitigate the neutral to earth voltage. In the high
power UPS; it is to eliminate the output transformer so that
The main components of the UPS are rectifier, battery, four
load is fed by the inverter directly, so the neutral of earth
wire inverter, four wire inverter and load. When the main
voltage is emerged. The currents flowing on each phase are
supply is present, the rectifier provides power to an inverter
generally not balanced so, that a transformer is not required,
as well as battery. The battery is charged. The inverter is on
and feeds power to the load through UPS switch. The UPS
switch is always on and connects load to inverter output.

When the UPS fails, then load is connected directly to the
mains directly through main switch. When the supply is not
available, then battery bank supplies power to an inverter.
Thus an inverter is always on and it takes power from
rectifier or battery.
The three phase four wire inverter is suitable for use in
high power UPS for its advantage of feeding unbalanced
load and the higher dc voltage utilization [3]. As the load is
fed by three phase three wire inverter is shown in Figure 2. Figure 5.Simplified diagram for four wire inverter
The switch in the inverter legs R,Y,B,N denoted as
In this paper, the load neutral point voltage for the three
Sk(SR,SY,SB,SN) corresponds to each vector Vk,for S=1 upper
phase four leg inverter is proposed and it is shown in Figure
switch in the inverter wire is conducting and for S=0,the
3.
lower switch is conducting. The vector V (1011) represents
switching state is shown in Figure 5.[8].
The equivalent circuits for states (1011) and (1010) are
represented in Figure 6(a) and Figure 6 (b) respectively.
Figure 2.Three phase three wire inverter
Figure 6(a).For switching state SRSYSBSN 1011

VRN =VBN=0 and VYN=-2Vd
Figure 3.Three phase four wire inverter

Figure 6(b).For switching state SRSY SBSN 1010
VRN =VBN=2Vd and VYN=0
2. Three Phase Four Wire Inverter
The three phase four wire inverter obtained by replacing the The comparison of a three phase 3 wire and 4 wire voltage
three wire switching network with a four wire switching source inverter as shown in table 1.
network is shown in Figure 4.
Table 1: Comparison of three phase 3 wire and 4 wire
inverter
THREE PHASE THREE PHASE

S.NO PARAMETER THREE WIRE FOUR
LOAD WIRE LOAD
1. Number of 6 8
Figure 4.Four wire Switching Network required power

switches
The simplified diagram of four leg inverter circuit feeding
Three Three
four wire load is shown in Figure 5.The neutral inductor Ln
can reduce switching frequency ripple. 2. Equivalent independent dependent
topology single phase half single phase full
bridge. bridge.
3. Number of the 6(no zero vectors) 16(14 active + 2

output vectors zero vectors)
9 1001 2/3 Vd 0 -2/3 Vd

4. Maximum 0.5Vd 0.577Vd
achievable peak 10 1010 1/3Vd 2/3 Vd
-1/ 3 Vd
value of line to
neutral voltage. 11 1011 1/3Vd -1/3Vd
-1/ 3 Vd
12 1100 1/3Vd 2/3 Vd
There are 16 switching states which are listed in table 2; it 1/ 3 Vd
can be taken from the graphical representation of switching
13 1101 1/3Vd -1/3Vd
vectors in Figure 7. 1/ 3 Vd
There are 14 non-zero voltage vectors and two zero
vectors(1111), (0000).The three phase variables Kr,Ky and 14 1110 0 0 Vd
Kb can be transferred as orthogonal coordinates kα,kβ,kγ
using eq (1). Any three phase sinusoidal set of quantities 15 1111 0 0 0
can be transformed to an orthogonal reference.
For given switching states of the inverter, the voltage
vector components can be calculated as,
 kα 
  cos θ cos (θ − 2π 3) cos (θ − 4π 3)  k r 
 k β  = 2 3  sin θ sin (θ − 2π 3) sin (θ − 4π 3 ) k y  (1)
    
k 
 1 2 1 2 1 2   kb 
 γ 
Where θ is the angle of orthogonal set α-β-0 with respect
to arbitrary reference. If α-β-0 axes are stationary and the α-
axis is aligned with the- axis, then θ=0 at all times. Thus, we
get
 kα   1 −1 2 − 1 2 kr 
   − 3 2   k y 
(2)
 kβ  = 2 3 0 3 2
k 
 γ  1 2 1 2 1 2   k b 
The above matrix can be rewritten as Figure 7.Switching vectors for three phase four wire inverter
Vα = 1 3 .V d (2 S R − S Y − S B ) (3)
3. Circuit Description of Four Wire voltage
V β = 1 3 .V d (S Y − S B ) (4)
Source Inverter
V γ = − 1 3 .V d (3 S N − (S R + S Y + S B )) (5)
The three phase four wire voltage source inverter, commonly
used for three phase voltage generation is shown in Figure 8.
Table 2: Switching combination and output voltages for It consists of eight switches Srp-Sxn and filter of inductor LR-
3 phase 4-wire inverter LX and capacitors CR-CB.The LC filter filters out the
switching harmonics. The voltage source inverter able to
NO. SR,SY,SB,SN Vα Vβ Vγ
generate balanced and high quality AC output voltage,
shown in Figure 8.
0 0000 0 0 0
1 0001 0 0 -Vd
2 0010 -1/3Vd 1/3Vd

-1/ 3 Vd
3 0011 -1/3Vd -2/3 Vd
-1/ 3 Vd
4 0100 -1/3Vd -1/3Vd
1/ 3 Vd
5 0101 -1/3Vd -2/3 Vd
1/ 3 Vd
6 0110 -2/3 Vd 0 2/3 Vd Figure 8.Three phase output voltages
In the three phase output voltage waveform shown in Figure
7 0111 -2/3 Vd 0 -1/3Vd
8, one line cycle is divided into six regions. In region 0˚–60˚,
8 1000 2/3 Vd 0 1/3Vd 120˚–180˚ and 240˚–300˚, the voltage waveforms in Figure
8 have similar pattern, i.e., one-phase voltage is always

lower than the other two [2].
The modulation method for four wire inverter are
1) The switch Sin (i = r, y, b) for the phase with the lowest
voltage is always turned ON and the corresponding Sip
for this phase is always turned OFF.
2) The switches Sin and Sip for the other two phases are
driven complementarily.
3) The switches Sxn and Sxp for the neutral phase are driven
complementarily. (a)
The main circuit diagram in Figure 3 is equivalent to
Figure 9(a) in 0˚–60˚region, which can be further organized
into Figure 9(b). The same equivalent circuit is also
applicable to 120˚–180˚ and 240˚–300˚ regions. The
switching of the inverter is shown in table 3.
(b)
Figure 10.Equivalent circuit for four wired VSI for 60˚–120˚

For further analysis, following assumptions are made.
(a)
1) LR = LY = LB = LX = L.
2) CR = CY = CB = C.
3) Switching frequency is much higher than fundamental
frequency.
Table 3: Switching logics for proposed controller
SWITCHES
(b) S1 S2 S3 S4 S5 S6 N1 N2
DEGREES
Figure 9.Equivalent circuit for four wired VSI for 0˚–60˚

0˚-
ON OFF OFF OFF OFF ON ON OFF
In region 60˚–120˚, 180˚–240˚and 300˚–360˚, the voltage 60˚
waveforms in Figure 8 have another pattern, i.e., one phase
voltage is always higher than the other two [2]. 60˚-
The modulation method for four wire inverter are ON ON OFF OFF OFF OFF OFF ON
120˚
1) The switch Sip (i = r, y, b) for the phase with the
highest voltage is always turned ON and the
corresponding Sin for this phase is always turned OFF. 120˚-
OFF ON ON OFF OFF OFF ON OFF
180˚
2) The switches Sip and Sin for the other two phases are
driven complementarily.
3) The switches Sxn and Sxp for the neutral phase are 180˚-
OFF OFF ON ON OFF OFF OFF ON
driven complementarily. 240˚
With this Figure 3 is equivalent to Figure 10 (a) in 60˚–
120˚ region, which can be further organized into Figure 240˚-
10(b).The same equivalent circuit is also applicable to 180˚– OFF OFF OFF ON ON OFF ON OFF
300˚
240˚and 300˚–360˚ regions. The switching of the inverter is
shown in table 3.
300˚-
OFF OFF OFF OFF ON ON OFF ON
360˚
4. Simulation Results
The Figure 11 shows the three phase AC rectifier and its
output.
Figure 14.Input voltage for three phase AC source
The Figure 14 is the three phase input source voltage for the
UPS.
Figure 11.Simulation circuit for rectifier.
Figure 12.Simulation result for rectifier
The above rectified output voltage in Figure 12 obtained

across the capacitor.
The Figure 13 is the three phase four wire inverter for Figure 15.Simulation result for four wire inverter
online UPS is proposed.
The Figure 15 is the simulation result for four wire inverter
for three phase each output is phase shifted by 120˚.
The Figure 16 shows the DC source input voltage for four

wire inverter.
Figure 13.Simulation circuit for three phase four wire

inverter
From the simulation analysis of Figure 13

(i) The wire N provides a lower impedance loop for
unbalanced current and triplen harmonics, so the
imbalance of output is dramatically reduced.
Figure 16.Simulation circuit for DC source four wire
(ii) The neutral inductance Ln can reduce the current that
inverter
flows through the Switching components of wire N.
Three line voltages VRY, VYB and VBR are step waves, with
step height Vdc/2 and Vdc. The three line voltages are
mutually phase shifted by 120˚ as shown in Figure 17.
Figure 19.THD level for three phase four wire inverter
5. Conclusion
The three phase four wire UPS has been proposed in this
paper. The fourth wire makes the inverter have the ability of
Figure 17.Simulation result for three phase four wire handling unbalancing loads. The inductor in fourth wire
inverter reduces the current through the switching components. The
Table 4: Simulation result parameters
inverter control has the advantages of both lower switching
to fundamental frequency ratio and outstanding ability to
carry unbalanced loads.
Parameter Values
References
Voltage for each phase 100V
[1] Fanghua Zhang, and Yangguang Yan “Selective
Frequency 50HZ Harmonic Elimination PWM Control Scheme on a
Three-Phase Four-Leg Voltage Source Inverter” IEEE
DC input voltage 200V Trans. Power Electronics, vol. 24, no. 7, July 2009.
[2] Lihua Li and Keyue Ma Smedley, “A New Analog
Inductance(L) 1mH Controller for Three-Phase Four-Wire Voltage
Generation Inverters” IEEE Trans. Power Electronics,
Capacitance(C) 1000µF
vol. 24, no. 7, July 2009.
Neutral inductance(Ln) 1mH [3] Liu Zeng, Liu Jinjun and Li Jin “Modeling, Analysis
and Mitigation of Load Neutral Point Voltage for
Rated resistive load 100Ω Three-phase Four-leg Inverter” IPEMC2009.
The neutral voltage waveform for four wire inverter as [4] Salvador Ceballos, Josep Pou, Jordi Zaragoza, José L.
shown in Figure 18. Martín, Eider Robles, Igor Gabiola, and Pedro Ibanez,
“Efficient Modulation Technique for a Four-Leg Fault-
Tolerant Neutral-Point-Clamped Inverter” IEEE Trans.
Industrial Electronics, vol. 55, no. 3, March 2008.
[5] Armando Bellini and Stefano Bifaretti “Modulation
Techniques for Three-Phase Four-Leg Inverters”
Proceedings of the 6th WSEAS International
Conference on Power Systems, Lisbon, Portugal,
September 22-24, 2006.
[6] Bellini and S. Bifaretti “A Simple Control Technique
for three phase four leg inverters”. SPEEDAM 2006.
[7] Richard Zhang, V. Himamshu Prasad, Dushan
Boroyevich and Fred C.Le “Three-Dimensional space
Vector Modulation for Four –Leg Voltage-Source
Figure 18.Simulation result for neutral voltage Converters” IEEE Trans.Power Electronics, vol.17,
no.3, May 2002.
The Figure 19 shows the THD level for three Phase four [8] Salem M. Ali Marian and P. Kazmierkowski “PWM
wire system. The harmonic distortion is reduced and its Voltage and Current Control of Four-Leg VSI” 1998
THD level is 3.92%. IEEE.
Authors Profile
SenthilKumar.R was born in
Tamilnadu, India, on November 2, 1966.
He received the B.E degree in Electrical
and Electronics Engineering from
Madurai Kamaraj University, in 1989.
He received his M.E (Power systems)
from Annamalai University, in 1991. He
has 15 yrs of teaching experience.
Currently he is working as Asst.
Professor in EEE department, Bannari
Amman Institute of Technology Sathyamanglam. Currently he is
doing research in the field of power
converters for UPS Applications.
Dr.Jovitha Jerome was born in

Tamilnadu, India, on June 2, 1957. She
received the B.E. degree in Electrical
and Electronics Engineering and M.E.
degree in Power Systems from College
of Engineering, Guindy, Chennai. She
did her DEng in Power Systems.
Presently she is working as Professor
and Head in Instrumentation and Control Engineering Department
of PSG College of Technology, Coimbatore.
NithyaBhama.S was born on September

4, 1987. She received her B.E Degree in
Electrical and Electronics Engineering
from Erode Sengunthar Engineering
College, Thudupathi, Anna University.
Currently she is pursuing M.E in Power
Electronics and Drives at Bannari Amman
Institute of Technology, affiliated to Anna
University.
Self Encrypting Data Streams for Digital Signals

A.Chandra Sekhar* Ch.Suneetha1 G.Naga Lakshmi2
*Professor,Department of Engineering Mathematics, GITAM University,Visakhapatnam,India
acs@gitam.edu
1,2
Asst.Professor,Department of Engineering Mathematics, GITAM University,Visakhapatnam,India
Abstract: Cryptography plays a vital role in implementing

φ = a 11x 12 + a 12 x 1 x 2 + ....... + a 1n x 1x n
+ a 21 x 2 x 1 + a 22 x 2 + ....... + a 2 n x 2 x n
electronic security systems. Cryptography can be used for 2
signing electronic documents digitally, digital rights
management, banking and for controlling access key documents. +…………………..
Public key cryptography is secure if and only if extracting secret +…………………..
key from public information is intractable. Hence design of …………………...
easily replicable cryptographic primitives is essential. In this
paper a new technique for self encrypting data streams is + a n1x n x1 + a n 2 x n x 2 + ....... + a nn x n 2
proposed basing on the matrices and quadratic forms. This In the quadratic form there are n square terms x12,
algorithm uses a quadric form as private key. The key matrix is x22……xn2 and xC2 product terms x1x2, x2x3…..,xn-1xn
also sent along with the cipher text in form of a string, which is so there are
further encrypted. The elements of the key matrix will be n (n − 1) n (n + 1)
different for different data streams. This consists of data n + nC 2 = n + = terms. For our discussion
encryption at two levels and the cipher text so obtained becomes 2 2
quite impossible to break or to interrupt. we confine to a quadratic forms of three variables.
φ = ax 2 + by 2 + cz 2 + 2hxy + 2fyz + 2gzx.
Keywords: Quadratic forms, cryptography, key matrix We know very well that a homogenous second
1. Introduction degree equation
φ = ax 2 + by 2 + cz 2 + 2hxy + 2fyz + 2gzx = 0
Public key cryptography was introduced by Diffie and
Helman[3][4] In public key cryptosystems, the decryption represents a pair if planes if and only if.
keys must be kept secret. A decryption key is therefore called abc + 2fgh – af2 –bg2 – ch2 = 0
a “secret key” or a “private key”. Some important public key that is
cryptosystems are RSA cryptosystem, Rabin encryption, El a h g
Gamal encryption. But the known Public key cryptosystems h b f =0.
are not efficient as many symmetric cryptosystems. Therefore g f c
in practice, hybrid cryptosystems. i.e., combinations of public
key systems and symmetric systems are used. For example, From this we can say that any quadratic form can be written
Alice if wants to send a message m in encrypted form to Bob. in the form of a matrix as X’ A X where
She generates a session key for an efficient symmetric x  a h g 
cryptosystem. Then she encrypts the message m using the X =  y  A = h b f  .
session key and the symmetric system obtaining the cipher  z  g f c 
text C. This encryption is fast because an efficient
symmetric cryptosystem has been used. Alice also encrypts Here A is a symmetric matrix. So, a symmetric matrix can be
the session key with Bob’s public key. Since the session key recognized as representative of a quadratic form.
is small, this encryption is also fast. Bob decrypts the session A is know as the discriminate of a quadratic form
key using his private key. Then he decrypts the cipher text C If A ≠ 0 then we say the quadratic form is non singular, in
with the session key and obtains the original message M. In
this paper we deal with non singular quadratic forms.
this paper we applied this method to the matrices and
Let x1,x2-----xn and y1,y2------yn be two sets of variables
quadratic forms.
consider the following system of linear equations.
y1 = b11x1 + b12x2 +----- b1nxn
2. Quadratic Form y2 = b21x1 + b22x2 +----- b2nxn
A homogeneous Polynomial of the second degree in n -------------------------------------------
variables i.e., the expression -------------------------------------------
n n yn = bn1x1 + bn2x2 +----- bnnxn
φ= ∑∑a
i=1 j=1
ij x i x j
Expressing each yi in terms of x1, x2 ------xn.
This system of linear equations can be written as matrix
where aij ∈ R and aij = aji is called a quadratic form in n
equations Y = B X where
variables x1, x2,…,xn over a real field.
Thus
 y1  4 The result N is adjusted to mod(27) to obtain E.

Y =  y 2  5 The key matrix K is obtained by the
adjustment factor for each entry
 y n 
6 The key matrix K and E are mixed to get the
 b11 b12 − − − b1n  cipher
b  x1 
b 22 − − − b 2 n  7 For decryption , from the cipher K and E are
B=  21
and X =  x 2  separated to get N
− − − − − − − − − − − − 
  x n  8 N is multiplied with the inverse of the matrix Q
 b n1 b x 2 − − − b nn  and adjusted to mod(27) to get the required
If B is non singular, than we can write X = B-1Y. Which plain text M.
shows that these is one–to–one correspondence between the
pairs of vectors X and Y. In this context we say that the 4.1 Example
system Y = BX determines a linear transformation and 4.1.1 Encryption
where B is called the matrix of linear transformation. If Alice wants to send a message GOOD LUCK to Bob she
2.1 Linear transformation of a Quadratic form converts message to a matrix
Let ϕ = X ' AX be a real Quadratic form in n variables. Since M = [6 14 14; 3 26 11; 20 2 10]
A is a symmetric matrix ϕ can be reduced by a non singular She takes a non-singular Quadratic form
2x2 + 3y2 + 4z2 + 2xy + 4yz + 4zx
linear transformation X = PY over the real field[1][2][3][4]. Q = [2 1 2;1 3 2;2 2 4]
Now X’ A X = (PY)’A (PY) N = M * Q=[54 76 96;54 103 102; 62 46 84]
= Y’(P’ A P)Y N is reduced to mod 27 resulting
= Y’BY [where B = P’ A P] E = [0 22 5; 0 22 21; 8 19 3 ]
We note that B’= (P’ A P)’ And key matrix K= [2 2 3;2 3 3; 2 1 3]
= P’ A (P)’= P’ A P = B The cipher text
Clearly Y’ B Y is also a Quadratic form over the real field. C = [0 2 22 2 5 3 0 2 22 3 21 3 8 2 19 1 3 3 ]
Thus if a Quadratic form X' AX is subjected to linear Which is equivalent to ACWCFDACWDVDICTBCC
transformation X = PY then the transformation is also a
Quadratic form Y’B Y where B = P’A P 4.1.2 Decryption
3. Proposed Method The message ACWCFDACWDVDICTBCC is converted into
the equivalent code to obtain the cipher text C =[0 2 22 2
5 3 0 2 22 3 21 3 8 2 19 1 3 3 ]
Alice wants to send a message M to Bob. She convents the The Key matrix K and E are separated as
message with equivalent numbers according to the following K=[2 2 3;2 3 3; 2 1 3]
table[7][8][9]. E== [0 22 5; 0 22 21; 8 19 3 ]
A B C D E F G H I J K L M By multiplying the key with 27 and on adding the
0 1 2 3 4 5 6 7 8 9 10 11 12 cipher we get
N O P Q R S T U V W X Y Z
13 14 15 16 17 18 19 2 21 22 23 24 25 N = [54 76 96;54 103 102; 62 46 84]
0 By multiplying N with inverse of Q we get the original
Null or Space
26 message as [6 14 14 32 6 11 20 2 10] which is GOOD
She chooses a non singular Quadratic form which is LUCK.
otherwise known as a private key in such a way that, the
message matrix M and the corresponding matrix of the 5. Conclusions
Quadratic form Q are conformable for multiplication. She In the new algorithm technique for self encrypting
multiplies M with Q resulting a matrix N. Then the elements data streams is proposed basing on the matrices and
of N are reduced to mod 27 which gives the encrypted quadratic forms. This algorithm uses a quadric form as
message matrix E. This encrypted matrix E is sent to Bob in private key. The key matrix is also sent along with the cipher
public channel. Along with this she also sends a session key text in form of a string, which is further encrypted. This
matrix K, whose elements are the integer parts of the consists of data encryption at two levels and the cipher text
elements of matrix N with respects to mod27. To maintain so obtained becomes quite impossible to break or to
absolute secrecy this key matrix K is sent in form of a string interrupt.Thus the computational overhead is very low. It is
which is further reduced to binary numbers. Bob with the almost impossible to extract the original information in the
help of the cipher text and secret key matrix K and the proposed method even if the algorithm is known.
private key Q and gets the original message back.
4. Algorithm
1 The non singular quadratic form is chosen.
2 The matrix Q of the quadratic is obtained.
3 The plain text is converted into its equivalent
message matrix M which is multiplied with Q
to obtain N.
References
Mrs.Naga Lakshmi is working as Assistant
Professor in the Department of Engineering
[1] K.R.Sudha, A.Chandra Sekhar and Prasad Mathematics . She is pursuing her MPhil in
Reddy.P.V.G.D “Cryptography protection of digital number theory and cryptography under the
signals using some Recurrence relations” IJCSNS guidance of Dr.A.Chandra Sekhar
International Journal of Computer Science and Network
Security, VOL.7 No.5, May 2007 pp 203-207
[2] A.P. STAKHOV, ”THE ‘‘GOLDEN’’ MATRICES AND A NEW

KIND OF CRYPTOGRAPHY”, C HAOS, SOLTIONS AND
FRACTALS 32 ( (2007) PP1138–1146
[3]. A.P. Stakhov. “The golden section and modern harmony
mathematics. Applications of Fibonacci numbers,”
7,Kluwer Academic Publishers; (1998). pp393–99.
[4]. A.P. Stakhov. “The golden section in the measurement
theory”. Compute Math Appl; 17(1989):pp613–638.
[5]. Whitfield Diffie And Martin E. Hellman, New
Directions in Cryptography” IEEE Transactions on
Information Theory, Vol. -22, No. 6, November 1976
,pp 644-654
[6]. Whitfield Diffie and Martin E. Hellman “Privacy and
Authentication: An Introduction to Cryptography”
PROCEEDINGS OF THE IEEE, VOL. 67, NO. 3,
MARCH 1979,pp397-427
[7]. C. E. SHANNON Communication Theory of Secrecy
Systems The material in this paper appeared in a
confidential report “A Mathematical Theory of
Cryptography” dated Sept.1, 1946, which has now been
declassified.
[8]. E. Shannon, A Mathematical Theory of
Communication, Bell System Technical Journal 27
(1948) 379–423, 623–656.
[9]. A. Chandra Sekhar , ,K.R.Sudha and Prasad
Reddy.P.V.G.D “Data Encryption Technique Using
Random Number Generator” Granular Computing,
2007. GRC 2007. IEEE International Conference, on 2-
4 Nov. 2007 Page(s):573-576
Authors Profile
Dr.A .Chandra Sekhar received his PhD
degree in number theory from JNT University
and MSc., degree with specialization in
algebraic number theory from Andhra
University . He Secured the prestigious
K.NAGABHUSHANAM Memorial Award in
M.Sc., for obtaining University First rank. He
did his MPhil from Andhra University in 2000.He was with Gayatri
degree college during 1991to 1995 and later joined GITAM
Engineering college in 1995. Presently he is working as Professor
and Head of the Department of Engineering Mathematics at
GITAM Engineering college, Visakhapatnam, INDIA.
Mrs.Ch.Suneetha is presently working as

Assistant Professor in the Department of
Engineering Mathematics . She is pursuing her
PhD in number theory and cryptography under
the guidance of Dr.A.Chandra Sekhar

Vol 2 No 4

Загружено:

Сведения о документе

Оригинальное название

Авторское право

Доступные форматы

Поделиться этим документом

Поделиться или встроить документ

Параметры публикации

Этот документ был вам полезен?

Это неприемлемый материал?

Авторское право:

Доступные форматы

Vol 2 No 4

Загружено:

Авторское право:

Доступные форматы

(IJCNS) International Journal of Computer and Network Security, 1

Vol. 2, No. 4, April 2010

RD-Optimisation analysis for H.264/AVC scalable

If B* (λ) is the solution to the constrained problem (4) then:

Where the solution to (4-11), M*, would satisfy,

If this is true for all B in S, it is true for a subset of B in S

Where Po (i, j) denotes a pixel from the original image

In this part of the simulation, basic parameters such as

Initially, spatial dimensioning is represented by

Figure 2. Sequential scalable coding (Foreman)

In this case, several spatial resolution or bitrates are taken

Figure 5. evaluation of SVC and SLC

Table 1: comparison of different video streams

Sequences Scalable Video Coding Single layer coding

Optimized Unoptimized Optimized Unoptimized

Garden 45.0008 38.5645 42.5682 40.2658

Foreman 34.5545 35.2564 32.5654 30.1254

Markov Based Mathematical Model of Blood Flow

During the time of pregnancy it can be found that the

exclusive states of the system. The state of the system at t =0

There are two stages:

Stage 1: Placenta attached to the mother side, say node I,

Stage II: Umbilicus, the entering point to fetus, say node II

Let us P0(t) the quantity of the good blood at node 1

After ∆t the blood at node I and node II is given by

P1(t+∆t) = P1(t) (1- u1 ∆t) + P1(t) (1- u2 ∆t) + P0(t) gv ∆t

From equation 1 and 2

Figure 2. Fetal Circulatory System in nodes representation

[4]. Here tissue is represented by capacitor in parallel with

Figure 3. Problems related to node Ia and their causes

Figure 4. Problems to fetus due to Intrauterine Growth

This will increase further if the contractions are more [1].

fetus extremely asphyxia. The fetus may die because of less

Figure 11. Deoxygenated blood flow curve, when both

Figure 8. Tissue’s Cell Membrane & its Electrical

Figure 9. Output at gv 0.5 with blood flow time of 10

When equation 10 is simulated at various levels of arteries

[1] C. Hernandez Sande, G. Rodriguez-Izquierdo, and M.

Redevelopment of Graphical User Interface for

Victoria University, School of Engineering and Science

course designer to create groups and manage group

Figure 1. System topology [Martin]

Examples of the IPMS major menus are shown in Figure 2

Figure 2. Student Menu [4]

Figure 3. Admin Menu [4]

Login Logout PHPMyAdmi Update_Web_Conten Team_Report

Add_New Delete Details

Check_Login External Update Team _Document User_Report Project_Report

Figure 4. Top-down Module Design for IPMS [15]

Back to the default home page

Contain general web pages

Contain web pages which

Contain the assigned

Figure 6. General Information submenus

Figure 5. Newly developed GUI for Admin Menu [15]

Q1. Registration/Signup process

The detailed results of the usability test are shown in Figure

Contain general web pages

Contain web pages which

(a) RCM3001 submenu

Figure 9. Student Menu [15]

(b) Assessment submenu Figure 10. My Team submenus

Q1. Re gistration/Signup proce ss