Академический Документы
Профессиональный Документы
Культура Документы
Marco Casassa Mont, Richard Brown, Simon Arnell, Neil Passingham HP Laboratories HPL-2012-206
Keyword(s):
Security Analytics; Risk Analysis; What-if Analysis; Incident Management Processes; SOC
Abstract:
This document is an example of the type of report an organisation would receive at the end of a HP Security Analytics engagement. The focus is on the analysis of the security risks and performance of the organisation's Security Incident Management Processes and related Security Operation Centre (SOC)'s activities. HP Labs carried out the underlying R&D work in collaboration with HP Enterprise Security Services (HP ESS) and involved analysis of processes, probabilistic modeling, simulation and "what-if" analysis for some of HP's key customers. The outcome of this was a set of case studies from which we have been able to create this more general anonymised report illustrating the richness of the risk assessment and "what-if" analysis that has been carried out. The lifecycle management of security is critical for organisations to protect their key assets, ensure a correct security posture and deal with emerging risks and threats. It involves various steps, usually carried out on an ongoing, regular basis, including: risk assessment; policy definition; deployment of controls within the IT infrastructure; monitoring and governance. In this context, Security Information & Events Management (SIEM) solutions play a key role. Even the best information security practices and investments in security controls cannot guarantee that intrusions accidental and criminal activities and/or other malicious acts will not happen. Controls can fail, be bypassed or become inadequate over time; new threats emerge. Managing such incidents requires detective and corrective controls to minimise adverse impacts, gather evidence, and learn from previous situations in order to improve over time. These incident management processes are usually run in the context of a SOC and/or as part of specialised Computer Security Incident Response Teams (CSIRTS), built on top of SOCs. Even with SIEM solutions in place, a potential major risk for the organisation arises due to delays introduced in assessing and handling known incidents: this may postpone the successful resolution of critical security incidents (e.g. devices exposed on the Internet, exploitation of privileged accounts, deployed malware, etc.) and allow for further exploitation. Another related risk can be introduced by sudden and/or progressive changes of the threat landscape, due to changing economic and social scenarios, new business activities or process failings within the existing IT services. This might create unexpected volumes of new events and alerts to be processed by the security team and as such, introduce additional delays. Hence, it is important for an organisation to understand the risk exposure due to their Incident Management processes, explore potential future scenarios (e.g. changes in available resources or threats landscapes or adoption of Cloud solutions) and identify suitable ways to address related issues, e.g. by introducing process changes and/or making investments in security controls.
External Posting Date: September 6, 2012 [Fulltext] Internal Posting Date: September 6, 2012 [Fulltext]
SecurityAnalytics:RiskAnalysisforanOrganisationsIncident ManagementProcesses
(1)
MarcoCasassaMont,(1)RichardBrown,(2)SimonArnell,(2)NeilPassingham Cloud&SecurityLab,HPLabs,Bristol,UK(2)HPEnterpriseSecurityServices
(1)
marco.casassamont@hp.com,richard.brown@hp.com,simon.arnell@hp.com,neil.passingham@hp.com
Abstract
This document is an example of the type of report an organisation would receive at the end of a HP Security Analytics engagement. The focus is on the analysis of the security risks and performance of the organisations Security Incident Management Processes and related Security Operation Centre (SOC)s activities. HP Labs carried outtheunderlyingR&DworkincollaborationwithHPEnterpriseSecurityServices(HPESS)andinvolvedanalysisof processes, probabilistic modeling, simulation and whatif analysis for some of HPs key customers. The outcome of this was a set of case studies from which we have been able to create this more general anonymised report illustratingtherichnessoftheriskassessmentandwhatifanalysisthathasbeencarriedout. The lifecycle management of security is critical for organisations to protect their key assets, ensure a correct security posture and deal with emerging risks and threats. It involves various steps, usually carried out on an ongoing, regular basis, including: risk assessment; policy definition; deployment of controls within the IT infrastructure; monitoring and governance. In this context, Security Information & Events Management (SIEM) solutions play a key role. Even the best information security practices and investments in security controls cannot guarantee that intrusions accidental and criminal activities and/or other malicious acts will not happen. Controls can fail, be bypassed or become inadequate over time; new threats emerge. Managing such incidents requires detective and corrective controls to minimise adverse impacts, gather evidence, and learn from previous situationsinordertoimproveovertime.Theseincidentmanagementprocessesareusuallyruninthecontextofa SOCand/oraspartofspecialisedComputerSecurityIncidentResponseTeams(CSIRTS),builtontopofSOCs. Even with SIEM solutions in place, a potential major risk for the organisation arises due to delays introduced in assessing and handling known incidents: this may postpone the successful resolution of critical security incidents (e.g. devices exposed on the Internet, exploitation of privileged accounts, deployed malware, etc.) and allow for further exploitation. Another related risk can be introduced by sudden and/or progressive changes of the threat landscape, due to changing economic and social scenarios, new business activities or process failings within the existing IT services. This might create unexpected volumes of new events and alerts to be processed by the securityteamandassuch,introduceadditionaldelays.Hence,itisimportantforanorganisationtounderstandthe risk exposure due to their Incident Management processes, explore potential future scenarios (e.g. changes in available resources or threats landscapes or adoption of Cloud solutions) and identify suitable ways to address relatedissues,e.g.byintroducingprocesschangesand/ormakinginvestmentsinsecuritycontrols. HP Security Analytics is uniquely positioned to provide the analysis of the involved risks, explore whatif scenarios andprovidedecisionsupportfordecisionmakers.ThistypeofSecurityAnalyticsassessmentsisnowavailableasa service,providedbyHPESS.
HP Security Analytics
Final Report
Disclaimer
This document is an example of the type of report a customer would receive at the end of a Security Analytics engagement [10]. The focus is on the analysis of the security risks and performance of the customers Security Events & Incident Management Processes. The report is loosely based around a real customer case study, but all contextual information, event and incident management processes, empirical data, analyses and conclusions have been modified in order to protect this customers identity and confidential information. As a result we believe this report reflects well the current situation many large organizations find themselves in with respect to the management of security events and related incidents. For this purpose, we introduce a fictitious company, ServiceBizWorld. This report focuses on two key processes in the incident management and remediation space - the process of analyzing and handling security events/alerts (to the point these events are either closed or classified as false positives or turned into incidents) and the process of managing and remediating to security incidents - which have been generalized (and represented with templates) and should be of relevance to many customers. The report shows how Security Analytics can be applied in carrying out risk assessment and what-if analyses. The analysis discussed in this report can be applied also to those companies that currently do not have explicit event and incident management processes in place but are planning to create them. In this context, Security Analytics can be of help to explore the impact of making different process choices (in terms of security risk and productivity) along with the implications of making specific technological investments. Security Analytics [10] has also been successfully applied to the areas of Identity and Access Management, Vulnerability and Threat Management and Web Infection.
Page 1 of 80
HP Security Analytics
Final Report
Page 2 of 80
HP Security Analytics
Final Report
Security Analytics
Risk Analysis: Security Event & Incident Management Processes
Executive Summary
HP Labs conducted a detailed analysis of the current ServiceBizWorlds Security Events and Incident Management Processes. It was undertaken using HPs Security Analytics modelling methodology. The analysis focused on the Event/Alert Processing & Remediation Process and the Incident Management Process. The latter is triggered when Major (IT) Security Incidents are identified by the SOC team. The analysis included the examination of the risk exposure arising from current practice and process steps measured as the time to identify and remediate security incidents. Models have been developed to simulate the arrival of events/alerts and the execution of the involved processing steps. These models took into account the complexity of events/alerts and their rate, the interactions with Third Parties (TPs), TP behaviours and non compliance issues, various SOC teams management steps. They measured, among other things, the time required to: process events/alerts; identify minor and major security incidents and remediate them; identify false positives. A first outcome of the analysis highlighted the complexity of the current processes; the models provide an explicit representation of the involved steps. Experimental results confirmed that interactions with TPs constitute a major bottleneck. Specifically they account for 60-70% of the time required to process events/alerts (i.e. between 15-50 days). This is a critical issue, as it negatively affects the processing of Major Security Incidents. Two key aspects have been identified as root causes: the lack of specific SLAs with TPs about how information should be provided to the SOC Team; the lack of information availability for TPs. In this context, three what-if scenarios were explored using the model of the Event/Alert Processing & Remediation Process (the most critical of the two analysed processes) to assess the impact on risk exposure and performance, in case of: (1) improving SLAs with TPs; (2) changing threat environments; (3) improving information availability for TPs. The main findings show that: The definition of SLAs for TPs (e.g. about response time and number of iterations with the SOC Team) can dramatically improve the overall performance and reduce the risk exposure by at least an order of magnitude; The current process might expose ServiceBizWorld to additional risks - in terms of delays to identify security incidents and saturation of TP resources/personnel - if the daily number of new events to be processed increases by about 30%; Investments in new controls not only can improve the overall TP information availability and performance (estimated to be an order of magnitude) but also help to cope with worsening threat landscapes. Specifically, the adoption of information request templates and manual scripts can ensure process resilience to a 60% increase of the daily number of new events/alerts to be processed. The introduction of automation of scripts and/or workflows enables process resilience to an increase of 90% of events/alerts.
A what-if scenario for the Incident Management Process showed similar outcomes and improvements when templates and automation are introduced to deal with incident response & remediation tasks. These findings depend on the assumptions made and the information available at the time of the analysis. They provide a starting point for conversations between ServiceBizWorld, TPs and the SOC Team, on expectations, policy & SLA changes, as well as on potential new investments, to improve the current process performances, reduce risk exposure and deal with contingency planning for worsening threat environments.
Page 3 of 80
HP Security Analytics
Final Report
Content
Executive Summary Content 1 2 3 4 5 6 7 Glossary Overview of the Report Area Selected for Investigation: Goals and Scope Overview of Security Event & Incident Management Processes Analysis of ServiceBizWorld Security Event & Incident Management Processes Security Analytics Approach Models 7.1 Model 1: Security Event/Alert Processing and Remediation 7.1.1 Modelled Process Areas
7.1.1.1 7.1.1.2 7.1.1.3 7.1.1.4 7.1.1.5 7.1.1.6 7.1.1.7 7.1.1.8 Events and Alerts Initial Assessment Phase SOC Team Interactions with TPs SOC Team Final Assessment Event Closure Security Incident TP Non Compliance Handling False Positives and Known Problems
3 4 6 7 8 10 13 15 17 17 19
19 19 20 21 22 23 23 24
7.1.2 7.1.3
7.1.3.1 7.1.3.2 7.1.3.3
25 26
26 28 32
36 38
38 38 38 39 40 40
7.2.2 7.2.3
7.2.3.1
40 41
41
Results Regarding Current State 8.1 Model 1: Events/Alerts Processing and Resolution
44 44
Page 4 of 80
HP Security Analytics
8.1.1 Summary
Final Report
48 49 50 52 52 52
57
8.2 Model 2: Incident Management Process 8.2.1 Summary 9 Results for What-If Scenarios 9.1 Model 1: Security Event/Alert Processing and Resolution 9.1.1 Impact of Changing SLAs for TPs
9.1.1.1 Summary
9.1.2
9.1.2.1
58
62
9.1.3
9.1.3.1
63
71
9.2 Model 2: Incident Management Process 9.2.1 Impact of Investments to Improve the Process Performance
9.2.1.1 Summary
72 72
75
10 Conclusion 11 References A.1 List of parameters - Model 1: Events/Alerts Processing and Resolution A.2 List of parameters - Model 2: Incident Management Process
76 77 78 80
Page 5 of 80
HP Security Analytics
Final Report
1 Glossary
CSIRTS ITIL Major Security Incident SA SIEM SIM/SEM SLA SOC TP Computer Security Incident Response Teams Information Technology Infrastructure Library Serious Security Incident with potential Major Security & Business Impacts Security Analytics Security Information and Event Management Security Information Management/Security Event Management Service Level Agreement Security Operations Centre Third Party
Page 6 of 80
HP Security Analytics
Final Report
Page 7 of 80
HP Security Analytics
Final Report
The area of Security Incident Management is wide as it covers various types of incidents, including the ones related to: IT infrastructure; people behaviours; system failures; fraud; etc. The security risk analysis discussed in this report focuses on IT security events and incidents. The following priorities have been identified by ServiceBizWorld: the need to be confident that ServiceBizWorld is ready to handle current and future security threats; the need to be able to justify investments and process changes.
The goals for the security risk analysis are: Analyse the risk exposure of ServiceBizWorld due to the current security incident management processes; Analyse the impact of introducing new SLAs for the involved stakeholders and/or deploying new controls within the incident management processes; Analyse the impact on current processes (in terms of risk exposure) of changes of the threat landscape (e.g. increase of number of new events/alerts to be processed); Analyse the impact of deploying new controls within the current processes to mitigate changes of the threat landscape.
A secondary goal of this analysis is to identify potential productivity issues related to current Incident Management Processes and explore suitable ways for improvement. The security analytics methodology adopted for this investigation has been specifically designed to answer these types of questions. The methodology is based on system environment modelling and predictive simulations, which allow the experimental exploration of risk and productivity across the set of outcomes for the current incident management and remediation processes. It also provides the ability to examine the impact of deploying different types of controls and/or process changes along with exploring the impact of a changing threat landscape. Changes to the event & incident management processes need to have the support of the various involved stakeholders and potentially continued investments in a platform that can facilitate the
Page 8 of 80
HP Security Analytics
Final Report
overall processes. This requires strategic alignment with ServiceBizWorlds overall business objectives and risk posture. The final aim of the analysis discussed in this report is to provide evidence to ServiceBizWorld and its Security & Risk Management teams to support their decision making process, specifically in identifying suitable policies and SLAs changes and/or justify future investments in process changes and controls.
Page 9 of 80
HP Security Analytics
Final Report
Page 10 of 80
HP Security Analytics
Final Report
They include: Collection of events and alerts: the initial phase, where filtered/correlated events and alerts are provided by SIEM solutions to the security team; Triage and initial analysis: it involves an initial analysis and assessment of the events and alerts, to identify potential known problems and false positives. The output of this step is either a decision to close the event/alert (in case of known issue or false positive) or to proceed with a deeper analysis; Collection of additional information from the involved stakeholders: it usually involves the collection of additional information about events/alerts, potentially from the involved stakeholders, i.e. the IT department, business areas, etc. This information is necessary to gather further evidence and data to assess the seriousness of the event/alert, identify the root causes, assess the involved risks and support informed decision making; Final Assessment: it involves making a final decision about the nature of the events and alerts. The collected evidence might still indicate that it is a false positive or it should be closed as not critical for the organisations. Alternatively, a security incident might be raised and/or escalated for remediation; Identification of Remediation steps for the Incident: it involves the identification of remediation steps to deal with the incident. This activity is usually carried out jointly with the involved parties within the organisation; Incident Remediation: it involves the actual resolution and remediation of the security incident by the involved entity, along with handling any residual risks due to failures to mitigate the security incident.
Page 11 of 80
HP Security Analytics
Final Report
It is important to notice that the management of false positives and known problems usually consume resources that could be used for the investigation of real security incidents. Ideally, it is desirable to minimise the number of false positives. This usually requires fine tuning activities between the SIEM solutions (raising events/alerts) and the initial assessment phase. The incident management processes are usually run in the context of Security Operation Centres (SOC) [5] and/or as part of specialised Computer Security Incident Response Teams (CSIRTS) [6], built on top of SOC centres. The work of the involved security team(s) can be supported by a variety of technical controls and solutions including workflow systems, incident tracking and support solutions, knowledge bases and collaborative tools, etc. However, quite often the involved processes are driven by human activities and the investigative skills of the security team members. As such, these processes might be subject to a variety of issues and potential failures, including: Delays due to lack of resources (personnel) to handle large volumes of events/alerts; Delays due to complex process steps and failures in human interactions and communications; Bottlenecks generated by some of the stakeholders to retrieve relevant information, e.g. by IT departments or business groups; Failures in assessing the severity of events/alerts and their correct categorization as incidents; Lack of tools, expertise and capabilities to remediate security incidents.
These issues and failures can expose the organisation to additional risks. A potential major risk for the organisation is due to delays introduced in assessing and handling known incidents: this might postpone the successful resolution of critical security incidents (e.g. devices exposed on the Internet, exploitation of privileged accounts, deployed malware, etc.) and allow for further exploitation. Another related risk can be introduced by sudden or progressive changes of the threat landscape, due to changing economical and social scenarios and/or new business activities. This might create unexpected volumes of new events and alerts to be processed by the security team and as such, introduce additional delays. To summarise, the space of Security Event & Incident Management is complicated and with many uncertainty points: it is subject to changing threat environments; it is defined by a variety of processes and process steps relying on people, human interactions, controls and information flows; it is subject to technical issues and failures. It is important for an organisation to understand their risk exposure due to the Incident Management processes and identify suitable ways to address related issues, e.g. by introducing process changes and/or making investments in security controls. HP Security Analytics is uniquely positioned to provide the analysis of the involved risks, explore what-if scenarios and provide decision support for decision makers. This report discusses the findings of the risk analysis carried out by HP Labs for ServiceBizWorld event & incident management processes, along with the exploration of what-if scenarios aiming at providing specific decision support and enabling further discussions between the involved stakeholders.
Page 12 of 80
HP Security Analytics
Final Report
Security
Event
&
Incident
The Security Operations Centre (SOC) team currently runs the core ServiceBizWorld processes for (IT) security event & incident management processes. ServiceBizWorld has a complex IT and business infrastructure. Part of its IT and service infrastructure is run by third parties, either in the context of outsourcing contracts and/or by using services in the Cloud. In order to assess the specific security event & incident management processes, identify potential risk exposures for ServiceBizWorld and explore relevant what-if scenarios, HP Labs requested information from key SOC and ServiceBizWorld personnel. Information and data have been provided to support this analysis along with detailed explanations of the various process steps and interactions. Various stakeholders are involved in these processes: The SOC team: it provides the key coordination, supervision and assessment capabilities for the management of events/alerts and the processing of related incidents; The Security Risk Management team: it is the ServiceBizWorld team that provides further guidance on priorities and decisions about incident management. It has a longer term, strategic view on security aspects and related policies driven not only by security requirements but also by business needs; Various third parties, including ServiceBizWorlds IT department, ServiceBizWorlds business groups and the IT departments of external companies, which provide (business and IT) services to ServiceBizWorld in the Cloud and/or in outsourcing.
The relevant SOC Event & Security Incident Management activities can be classified in two types of processes, shown in Figure 5-1:
Events/Alerts
Page 13 of 80
HP Security Analytics
Final Report
The first process, referred in this report as Events/Alerts Processing and Resolution Process, primarily aims at: providing the initial assessment of events/alerts; gathering of relevant information from the involved TPs to further analyse the events/alerts; dealing with the final assessment of events/alerts and their classification as false positives, known problems, events to be closed, minor security incidents and major security incidents. This process is fed by a variety of events/alerts, originated by notifications sent to the SOC team by various SIEM tools and TPs. The SIEM tools include, among others, the HP ArcSight [3] and RSA enVision solutions [4]. Specifically, the events/alerts collected from the SIEM system could relate to: Managed internal firewalls; Managed network devices; Domain authentication events; Malware security events; Authentication events.
Based on information provided by the SOC team, on average 35-40 events/alerts are processed on daily basis, i.e. about 12000-15000 per year. Among all these events/alerts, a small part of them involve Major Security Incidents. These are critical security incidents for which a full report has to be provided to the Risk Management team and that trigger the second process. These are also referred to as Managed Security Incident as the SOC team is directly involved in the incident resolution process. On average, 800-900 major security incidents are raised on yearly basis. The category of Major (IT) Security Incidents include: discovery of exposure of ServiceWorldBizs devices and systems to the Internet; unauthorised privilege escalation; malware security incidents; security log cleaning; misuse of authentication tokens; incidents raised from internal firewalls. The second process, referred in this report as Security Incident Management Process is triggered when Major Security Incidents are raised by the SOC team. This process involves the steps of: identifying and agreeing remediation requirements for major security incidents, related to TPs (task usually carried out jointly by the SOC team and the Risk Management team); the actual phase involving TPs remediation to the incident. Next sections provide further detail of how the information about these two processes has been interpreted and captured in Security Analytics models to carry out the investigations described in Section 3. These models will primarily focus on the interactions between the SOC team and Third Parties (TPs). The key interactions with the Risk Management team have also been modelled.
Page 14 of 80
HP Security Analytics
Final Report
Specifically, the two models have been built by using a flow chart-based graphical representation. To validate them, they have initially been used to provide an estimate of the risk exposure to and productivity for current processes. This phase was iterated over a few times with the SOC and Risk Management teams to ensure that the right process steps and the correct empirical data were captured by the. After their validation, the two models have been used to explore what-if scenarios, as requested by ServiceBizWorld. The visual representations of the models were created by using the HP proprietary Gnosis modelling toolset, with the various shapes in the model instantiated with a set of parameters. The models were then transformed by the toolset into executable code, which was used to run Monte Carlo simulations. This whole analytic chain is shown in Figure 6-1.
Page 15 of 80
HP Security Analytics
Final Report
The two models were drawn by using visual diagrams (shown in the next sections) together with detailed descriptions of the involved parameters. The complete list of parameters can be found in Appendices A.1 and A.2. The visual diagrams should be interpreted as follows: The star-shaped components represent the events that occur at regular intervals as defined with a probability distribution in a model (e.g. 7 events/alerts received per day), each starshaped component starts a separate flow that runs in parallel to other flows, within simulations; The rectangular shapes correspond to the process steps each with an internal logic, that consists of time duration (e.g. a step takes 1 day to complete) or some manipulations of internal parameters (e.g. number of iterations with TPs); The rhombus shapes are if-type decision boxes, that return Boolean values i.e. true or false based on Bernoulli trial for a certain probability (e.g. the probability that an event/alert is a false positive has a probability of 0.1); The people-shaped icons represent resources that are used by different process steps (e.g. members/people of the TP security staff); The cylinder shape stores the values of certain parameters that, for example, represent the current state (e.g. the number of iterations with TPs, the number of incidents, false positives, etc.); Finally, the process dynamics of the model is captured by the arrows connecting events to processes steps and decision boxes.
Page 16 of 80
HP Security Analytics
Final Report
7 Models
This section presents the two models built by HP Labs to carry out the risk and performance analysis for the ServiceBizWorlds Event & Incident Management processes. It also describes the set of metrics used to convey analytic results and the what-if scenarios that have been explored with the models.
Page 17 of 80
HP Security Analytics
Final Report
Page 18 of 80
HP Security Analytics
Final Report
Specifically there are 2 types of events/alerts that are processed: SOC Events and Alerts received from SOC SIEM tools: these are the SOC events and alerts received from SIEM tools such as HP ArchSight and RSA enVision. On average, 20 of these events/alerts are generated on daily basis; Notification of IT Security Events/Alerts from TPs: these are specific notifications of IT security events/alerts from third parties. On average, 15 of these events/alerts are generated on daily basis.
7.1.1.2 Initial Assessment Phase Figure 7-3 illustrates the various process steps that are carried out by the SOC team to initially assess the new event/alert:
This includes the following steps: An initial assessment of the event/alert, usually taking between 1 hour and 2 hours; A decision if the event/alert is a known security problem and in case it is, if the threat level has been changed. The available empirical data suggests that 10% of cases are known issues. In 5% of them the threat level has changed. In case it is a known problem, it is handled as such; In case it is not a known problem or the threat level changed, a decision is made if there is any need to notify the Risk Management team. This usually happens in 15% of cases. In this case, the SOC team has to prepare a briefing for the Risk Management team and agree on the
Page 19 of 80
HP Security Analytics
Final Report
relevant remediation. This usually takes between 5 hours and 1 day. An additional decision is made if this event/alert is actually an incident and its category has been established. This happens in 80% of cases. The empirical data suggests that in 10% of cases the security incident is raised as a Major Security Incident; If there is no need to brief the Risk Management team and/or the incident category has not been established, then the process continues with the analysis of the event/alert. This might require gathering information from TPs.
7.1.1.3 SOC Team Interactions with TPs Additional information might be required from TPs to identify the root causes of the event/alert or gather additional contextual data, audit logs, explanations, etc. Only in 20% of cases no additional information is required from TPs. In this case, the SOC team carries out their investigation (usually taking between 1 hour and 2 days). In 80% of cases, however, an interaction with TPs is required. Figure 7-4 illustrates the involved steps:
NOTE: We anticipate that this part of the process is critical, both in terms of risk exposure and performance. It introduces a bottleneck due to fact that the SOC Team usually has to go through multiple iterations with TPs in order to gather the required information: the quality and quantity of the provided information is also another potential issue. A key problem is that there are no specific policies or SLAs in place, with TPs, mandating criteria and obligations on how to effectively retrieve and provide data to the SOC Team. This part of the process usually starts with a request for information to a TP by the SOC Team, along with an informal definition of the agreed timescale for responses. In 40% of cases (on aggregated bases, over the various iterations between the TP and the SOC Team) the TP does not respond in due time (usually after an agreed response period of 1-10 days, depending
Page 20 of 80
HP Security Analytics
Final Report
on the circumstances). In this case, a decision is made by the SOC Team to escalate the issue to the Risk Management team (10% of cases) and deal with the escalation process (1-3 hours). This situation triggers another process step, involving dealing with TP Non Compliance to agreed requests. The process continues with the SOC Team asking again for the required information. If the required information is provided by the TP (60% of cases, on aggregated basis) the model captures various TP behaviours, based on observation provided by the SOC team: Usually, 5-7 iterations are required with TPs, in order to gather the right information and data; The first iteration with TPs is critical: it can take between 1 and 10 days to get some information back. This is due to various delays and overheads, for example the SOC team having to identify the right TP people to interact with, delays in getting replays, various redirections, etc. The following iterations provide a better TP response time, usually between 1 and 3 days.
After receiving some data from the TP, the SOC team carries out their investigation (duration: 1 hour2 days). This might trigger the need to brief the Risk Management team, in 20% of cases (duration: 13 hours). Time is also spent to identify the recommended level category for the event/alert (1-2 hours). The provided information might not satisfy the required quantity and quality criteria. This happens in 30% of cases: an additional request for information is made by the SOC Team. If it satisfies the criteria, an additional SOC team investigation is carried out (duration: 1 hour - 1 day). This might raise the need to brief the Risk Management team again (10% of cases) and prepare for the briefing (duration: 3 hours - 1 day). If the incident category has been established (80% of cases), than a security incident is raised. In 10% of cases the incident is a Major Security Incident. Otherwise, the SOC Team decides if additional information is still required from TPs. As previously mentioned, on average 5-7 iterations with the TP might need to take place before all the relevant information is retrieved. In case the required information has been provided, the process proceeds with the SOC Team final assessment phase. 7.1.1.4 SOC Team Final Assessment Figure 7-5 illustrates the steps involved in the SOC Team Final Assessment:
Page 21 of 80
HP Security Analytics
Final Report
This usually involves an additional analysis activity by the SOC Team (duration: 4 hours -2 days), followed by SOC team handling recommendations and agreements made with the Risk Management team, which make take between 1 and 5 days. The next process step involves the SOC Team providing recommendations about the level category for the event/alert (duration: 2-4 hours). In 20% of cases the conclusion is that the event/alert is a false positive. In this case, the procedure for handling false positives is followed. Otherwise, in 80% of cases it is a security incident. In this case, in 95% of cases a security incident is actually raised. The empirical data suggests that in 10% of cases a Major Security Incident is actually raised. In all other situations the event/alert is closed. 7.1.1.5 Event Closure Figure 7-6 shows the involved steps:
The process of closing an event/alert usually requires 1 hour 2 days. It is followed by the writing of a Report for the Risk Management Team, that usually takes between 1 hour and 1 day.
Page 22 of 80
HP Security Analytics
7.1.1.6 Security Incident
Final Report
Figure 7-7 shows the various process steps involved in case the event/alert has been classified as a (IT) Security Incident:
The actual process followed by the SOC team depends on the type of security incident: Major Security Incident (SOC teams Managed Incident): it is a critical incident that triggers the second process, captured by Model 2 (Security Incident Management). This involves raising a major security incident with the Risk Management team (duration: 1-2 hours), recording the incident (0.5-1 hour) and producing an incident report (2-5 days), before triggering the second process. On average there are 800-900 Major Security Incident per year; Minor Security Incident: it usually requires raising an internal (SOC Team) minor security incident (duration: 0.5-1 hour), recording the incident (0.5-1 hour) abd producing a minor incident report (duration: 1-5 hours), usually requiring notifying the Risk Management Team with a short email/document.
7.1.1.7 TP Non Compliance The TP Non Compliance process is usually triggered if TP does not provide the required information to the SOC Team in due time. Figure 7-8 shows the involved process steps:
Page 23 of 80
HP Security Analytics
Final Report
The SOC TEAM raises a TP Non Compliance Warning (duration: 1-2 hours), followed by reporting the TP to the Risk Management Team (duration: 1-5 hours). This might be followed by an escalation process to deal with the resolution of the problem or the acceptance of risk (this part is outside the scope of this report). 7.1.1.8 Handling False Positives and Known Problems This phase involves the handling of false positives and known problems, as shown in Figure 7-9:
Page 24 of 80
HP Security Analytics
If the event/alert is a false positive, it will be closed in 0.5-1 hour.
Final Report
If initially the event/alert was classified as a known problem, a further assessment is made to discriminate if it is a known problem with an acceptable risk (20% of cases) or if it is a false positive. After this classification, the event/alert is closed, usually in 0.5-1 hour.
7.1.2 Metrics
This model has been built to carry out the investigations discussed in Section 3, specifically explore the risk exposure and performance issues due to the SOC Teams Event/Alert Processing & Remediation Process. Metrics need to be put in place to provide measurements both for the current situation and for what-if scenarios. As such, the model has been instrumented with a set of metrics to convey outcomes in terms of risk and performance. This includes: Risk Indicators: o Time to Process Events/Alerts in order to Identify (Minor and Major) Security Incident: the longer the time to identify the nature of the event/alert, the longer the organisation is exposed to potential risks, such as exploitation of privileged accounts, spread of malware, exploitation of systems by attackers, etc; Time Spent in Iterations with TPs: this is a subset of the previous time. It specifically measure the time that has been spent in interactions with TPs, being these a bottleneck. Also in this case, the longer the time, the longer it takes to deal with the event/alert, hence further exposing ServiceBizWorld to risks; Time Spent for Event/Alerts Closures: this is the time required to close an event/alert, hence determining that there are no additional risks/issues for ServiceBizWorld; Time Spent to deal with TP Non Compliance Situations: this is the time spent to process situations of TP non compliance i.e. when TPs do not provide the requested information in due time. TP non compliance causes additional risk exposure for ServiceBizWorld due to the additional delays to process events/alerts as well as performance issues, due to the extra work carried out by the SOC Team to handle them; Various Counters: counters have been introduced to count how many times an event/alerts ends up as a (Major or Minor) Security Incident or as a Closed Event/Alert, as well as many times there has been a TP Non Compliance issue.
o o
Performance Indicators: o Time Spent in Dealing with False Positives: this is the time required to identify false positives and deal with them. The shorter it is the less time SOC Team resources are used for this type of event/alert instead of focusing on potential security issues. This time illustrates the impact on efficiency and performance; Time Spent in dealing with Known Problems: this is the time required to identify known problems and dealing with them. The same comments made before, for false positives, apply. Various Counters: counters have been introduced to count how many times an event/alerts ends up as a false negative or a known problem.
Section 8 discusses how these metrics have been used to describe the outcomes for various simulations.
Page 25 of 80
HP Security Analytics
Final Report
Ultimately, the desired outcome of this what-if analysis is to enable discussions with TPs to improve the current situation and explore trade-offs involving Performance vs. Costs vs. Security Risks. Various types of SLAs could have been explored and investigated. For the purpose of this scenario, being the focus on the SOC Team TP interactions, the following types of SLAs (and related constraints) have been explored: 1. Provision of requested information to the SOC Team, within the agreed timeframe; 2. Maximum number of allowed TP iterations, for each event/alert, to collect the requested information; 3. Quality and quantity of the information provided in TP responses; 4. Time to provide information to the SOC Team, for each iteration. It is important to notice that, in this analysis, we only considered the case where the same class of SLAs/Policies apply to all involved TPs. This might not necessarily be the case in other contexts/scenarios. Security Analytics and related models can help to capture multiple classes of SLAs and explore their implications. Figure 7-10 illustrates four cases of potential SLAs, along with the related parameters for the Security Analytics model, compared against the current situation:
Page 26 of 80
HP Security Analytics
Cases TP Responding in Time (Aggregated) CURRENT 60%
(single iteration: 92%)
Final Report
Quality/ Quantity of Response # TP Iterations Time to Provide data to the SOC Team, per Iteration
70%
5-7 (average: 6)
CASE #1
70%
(single iteration: 93.1%)
78%
4-6 (average: 5)
CASE #2
80%
(single iteration: 94.6%)
86%
3-5 (average: 4)
CASE #3
90%
(single iteration: 96.5%)
93%
2-4 (average: 3)
CASE #4
100%
(single iteration: 100%)
100%
1-2
Each case introduces an improvement in the expected TP performance, compared to the current situation. Case #4 potentially describes the best type of SLA that could be agreed with TPs i.e. the one that should provide the best possible outcomes, all other aspects being the same. For each of these cases, simulations have been carried out to determine the impact on the risk exposure. Section 9 presents the results. It is beyond the scope of this what-if scenario to indicate which specific controls and investments should be put in place, to achieve these SLAs, from a TP perspective. We understand this involves complex contractual agreements between the parties and changes in the way work is currently organised with TPs. However, in general, some of the improvements described in the SLAs can be achieved by optimising the communications between the SOC Team and TPs and potentially deploying tools and solutions to simplify and automate the involved critical process steps. This includes: Defining templates about the type of requested information, for each specific managed event and alert; Defining scripts, at the TP site, to describe how and where to collect the relevant information, to be provided to the SOC Team; Providing ways to automate these scripts, for example by using monitoring tools and SIEM solutions; Deploying workflow management solutions spanning across the SOC Team and TPs, to track the overall process of handling events/alerts and the information gathering activities.
Page 27 of 80
HP Security Analytics
7.1.3.2 Impact of Changes of the Threat Environment
Final Report
This what-if scenario aims at analysing the risk exposure implications in case of the worsening of the threat landscape. What if, the threat environment gets worst and worst? What would be the impact, in terms of risk exposure, if the number of new events/alerts to be handled by the process increases of 10%? Or 30%? Or 50%? Various situations and scenarios might introduce sudden changes of the threat landscape: Major social and economical events: this might include events such as the Olympic Games in London 2012, major economical and financial crisis, etc. that could create a surge of social engineering attacks and malware to exploit vulnerabilities and gather confidential information from ServiceBizWorlds employees; Major changes in the ServiceBizWorlds business model: this might include increasing the number of services offered online and/or in the Cloud, hence potentially increasing the attack surface and introducing new attack/exploitation opportunities for hackers.
In this scenario we consider the current event/alert processing situation as the baseline. Simulations are carried out with the model to explore the implications, in terms of risk exposure, of increasing the number of new events/alerts. Specifically the focus is on exploring the impact on the time required to identify security incidents, including Major Security Incidents. A desired outcome of this analysis is to enable discussion with TPs for Contingency Planning as well as exploring suitable investment trade-offs between Costs (due to resources/personnel) vs. Security Risks. In this context it is important to identify critical thresholds, after which event/alert processing delays start accumulating, hence exposing ServiceBizWorld to additional security risks. The current bottleneck introduced by TP interactions has been the specific focus of this analysis. Whilst ServiceBizWorld has direct control of the SOC Team and its incident management processes and can flexibly adapt them to changes, this is not necessarily the case for TPs, especially for the external service providers. In this context, to effectively carry out the analysis, it has been necessary to introduce the concept of Resources/Personnel into our model, related to the interactions between the SOC team and TPs. The number of TP people (potentially part of local security teams) involved in each interaction to address the SOC Team requests is limited: an increase in the workload might inevitably introduce delays. It has been proven difficult to gather the specific information about the number of involved resources (dealing with the event/alert processing activities) from the various TPs. This information is often business confidential. However, we can leverage Security Analytics and the model to make assumptions. The current model for the Event/Alert Processing & Resolution process has been validated by the SOC Team and teh Risk Management Team: it reflects the current solution. It is possible to use it, make assumptions and derive the required information about resources indirectly. Specifically, this involves: Modelling the collective impact of the various TPs; Modelling the TP allocation of resources/personnel in the context of their interactions with the SOC Team;
Page 28 of 80
HP Security Analytics
Final Report
Doing a reverse engineering with the model, to indirectly determine a reasonable number of TP resources/personnel that, once added to a new version of our model, produces the same kind of outcomes i.e. as in the current situation - in terms of risk exposure and performance; Allow for some slack in the usage of the current TP resources.
As previously mentioned, in the current modelled process, we focus on the interactions between TPs and SOC Team, see Section 7.1.1.3, to obtain the data requested by the SOC Team. The empirical data currently available suggests that the time required to get data from TPs is variable, depending if it is the first iteration with a TP or a follow-up iteration: 1st Iteration: 1-10 days; Follow-up Iterations: 1-3 days.
However, not all this time is spent by the involved TP resources in actively doing the work. A major overhead is introduced to understand who to interact with, which information to retrieve and from where, miscommunications, etc. As such, the overall time required to provide a response by the TPs has been modelled in a more accurate way, by splitting it in two parts: Overhead time: this time does not really need resource allocations. It involves delays and waiting time; Actual work: this is the time spent to carry out the information gathering work, requiring the utilisation of resources/people.
Figure 7-11 illustrates the assumption made for the current model (and validated by the SOC Team and the Risk Management team) about splitting the current available time for providing data to the SOC Team into the overhead and actual work times:
CURRENT MODEL
1-10 days Overhead: 0.5 7.5 days Actual work: 0.5 2.5 days
1-3 days Overhead: 0.45 0.55 days Actual Work: 0.5 2.5 days
Figure 7-11. What-if Analysis: Modelling Overhead and Actual Work time allocated to Resources
Of course different assumptions could have been made, by having more precise information from TPs. It is beyond the scope of this report to explore the entire space of assumptions. The main purpose of this what-if analysis is to illustrate the potential impact of changes of the landscape by using reasonable assumptions and examples. The main goal is to use the simulation outcomes as additional scientific data to enable informed follow-up discussions with TPs, in terms of Contingency Planning. The current model of the Event/Alert Processing & Remediation Process has been extended, to factor in the introduction of resources and the time breakdown, described in Figure 7-11.
Page 29 of 80
HP Security Analytics
The updated model is shown in Figure 7-12:
Final Report
Figure 7-12. What-if Analysis: Updated part of the Model to reflect Resources and Time Breakdown
Page 30 of 80
HP Security Analytics
Final Report
The only affected part of the model is the Information Gathering Phase phase, discussed in Section 7.1.1.3. Figure 7-13 shows the changes:
As previously mentioned, this model has been used to carry out various experiments, with different numbers of TP resources, to identify a reasonable number of resources that would produce results consistent to the ones observed with the current SOC event/alert processing & resolution process. Some slack in the utilisation of these TP resources has also been factored in. The outcome of these experiments is that if 250 resources (people), spread across the various TPs (along with the other assumptions mentioned before), are factored in, the updated Security Analytics model provides results comparable to the ones obtained with the initial model (Figure 7-1). More details are provided in Section 9, along with the outcomes of this what-if scenario. However, it is important to notice that, in this context, the number of identified resources should be purely considered as an internal Security Analytics mechanism to enable this type of analysis. The updated model has then been used to explore different cases, involving changes in the threat environment, in terms of an increase of the number of events/alerts received on daily basis, ranging from +10% up to +60%. Figure 7-14 shows the different scenarios that have been analysed:
Case Events/Alerts per Day
CURRENT
35
38
42
45
Page 31 of 80
HP Security Analytics
CASE #4: +40% Events/Alerts 49
Final Report
52
56
For each of these cases, simulations have been carried out to determine the impact on the time to identify a security incident (hence the impact on risk exposure) and the actual allocation of resources over time, including situations where resources saturates and delays are introduced. Section 9 presents the results. It is important to notice that in case of resource saturations, the delays are likely to increase over time. Hence results depend on the simulated time period. In the context of this what-if analysis, a period of time of 6 months has been considered. We wanted to explore the implications for the organisation of taking no actions during that entire period of time. The analysis can be tuned on a different timeframe, based on needs. 7.1.3.3 Impact of Investments to Improve Information Availability for TPs This what-if scenario is based on the previous two scenarios. Indeed, TP interactions are a bottleneck and changes in the threat landscape can further negatively impact ServiceBizWorlds exposures to risks. In case of a worsening threat landscape, TPs could inject an increased number of resources (personnel) to cope with the additional workload: however, this does not make sense for a variety of reasons, including practical and economic ones. Specifically, it would makes no sense for TPs to keep investing in additional resources (personnel) without first addressing the root causes of the delays and bottlenecks introduced in their interactions with the SOC team. Discussions with the SOC Team and the Risk Management Team highlighted that one of the key root causes of this problem is the lack of information and/or complexity in retrieving it. In other words, TPs might need to spend considerable amounts of time in understanding how and where to retrieve the requested information. This would account for most of the delays and overheads discussed in the previous sections. This is a complex problem to be addressed, as it involves legal, economical and contractual aspects. New investments might need to be made by TPs and potentially the SOC Team; contracts and SLAs might need to be modified. However, the main goal of this what-if scenario is to illustrate the implications of making these improvements and analyse the consequences, in particular in case of changes of the threat landscape (worsening threat environment). These outcomes can be used as the starting point, i.e. as additional evidence, for joint discussions, involving ServiceBizWorld, TPs and the SOC Team towards making changes. In this context, various controls have been considered as potential ways to improve the current situation:
Page 32 of 80
HP Security Analytics
Final Report
Definition of SOC Teams templates about the type of requested information, for each specific managed event and alert. Definition of scripts, at the TP site, to describe how and where to collect the relevant information, to be provided to the SOC Team; Script automation coupled with advanced monitoring and correlation tools (e.g. SIEM solutions [3], [4]); Deployment of workflow management solutions (e.g. [9]) spanning across the SOC Team and TPs, to track the overall process of handling events/alerts and the information gathering activities.
A few assumptions have been made about the impact of introducing these controls on the overall process, based on the technical capabilities and implications of the mentioned controls. Figure 7-15 illustrates three different cases that have been explored with Security Analytics simulations:
Investigation 1 Cases Additional Capabilities/Controls Parameters in the Model (Information Gathering Phase)
Overhead: 1st Iteration: [0.5 days, 7.5 days] Next Iterations: [0.45 days, 0.55 days] Actual work: [0.5 days, 2.5 days] SOC Teams templates defining the Overhead: 1st Iteration: [0.4 days, 6 days] types of required information Manual scripts, at the TP site, Next Iterations: [0.35 days, 0.4 days] describing how to retrieve data Actual work: [0.4 days, 2 days] Note: Improvement of 20% compared to current situation
CURRENT
NONE
CASE #1
CASE #2
As above. In addition: -
Improved Monitoring Solutions at TP Next Iterations: [0.225 days, 0.25 days] sites Automated execution of scripts to Actual work: [0.25 days, 1.25 days] retrieve data Note: Improvement of 50% compared to current situation
CASE #3
As above. In addition:
Overhead: 1st Iteration: [0.1 days, 1.5 days] Next Iterations: [0.09 days, 0.1 days] Actual work: [0.1 days, 0.5 days] Note: Improvement of 80% compared to current situation
It has been estimated that, in Case 1, the introduction of templates and manual scripts will provide an improvement of the overall TPs performances (see indicators in the table) of 20%, compared to the current situation. The additional automation of the scripts, coupled with monitoring solutions (Case 2) will introduce an improvement of 50%, compared to the current situation. Finally, the introduction of workflow automation (Case 3) could introduce an improvement of 80%, compared to the current situation.
Page 33 of 80
HP Security Analytics
Final Report
This what-if investigation, referred as investigation 1, still assumes that there will be 4-7 iterations with TPs, for each information request coming from the SOC Team. This is the worst case scenario, where no improvement happens in terms of reducing the number of iterations. This might be justified with the fact that the improvement in speed-up the retrieval of information is not matched by improvements in the communication with the SOC Team, hence still the need for multiple iterations. A variant of this investigation, referred as investigation 2, has factored in the situation where the number of required interactions with TPs is also affected, i.e. it progressively diminish by introducing more sophisticated controls. Figure 7-16 illustrates the updated three what-if cases that have been explored with Security Analytics simulations:
Investigation 2 - Cases Additional Capabilities/Controls Parameters in the Model (Information Gathering Phase) Number Iterations event/alert) of TP (per
CURRENT
NONE
Overhead:
1st Iteration: [0.5 days, 7.5 days] Next Iterations: [0.45 days, 0.55 days]
Actual work: [0.5 days, 2.5 days] SOC Teams templates defining the types of required information Manual scripts, at the TP site, describing how to retrieve data
CASE #1
Overhead:
1st Iteration: [0.4 days, 6 days] Next Iterations: [0.35 days, 0.4 days]
Actual work: [0.4 days, 2 days] Note: Improvement of 20% compared to current situation
CASE #2
As above. In addition: Improved Monitoring Solutions at TP sites Automated execution of scripts to retrieve data
Overhead:
1st Iteration: [0.25 days, 3.75 days] Next Iterations: [0.225 days, 0.25 days]
Actual work: [0.25 days, 1.25 days] Note: Improvement of 50% compared to current situation
CASE #3
As above. In addition: Workflow automation spanning across the SOC Team and TPs
Overhead:
1st Iteration: [0.1 days, 1.5 days] Next Iterations: [0.09 days, 0.1 days]
Actual work: [0.1 days, 0.5 days] Note: Improvement of 80% compared to current situation
For each of the above 3 cases, both for Investigations 1 and 2, the parameters defined within the model have been updated and a new version of the model generated. Each updated version of the model has been used in simulations, to explore the specific implications and impact of threat environments getting worst and worst. Figure 7-17 shows the different threat environment situations that have been analysed, against each of the 3 what-if cases (for investigations 1 and 2) discussed above:
Page 34 of 80
HP Security Analytics
Case Events/Alerts per Day
Final Report
CURRENT THREAT CASE #1: +10% Events/Alerts THREAT CASE #2: +20% Events/Alerts THREAT CASE #3: +30% Events/Alerts THREAT CASE #4: +40% Events/Alerts THREAT CASE #5: +50% Events/Alerts THREAT CASE #6: +60% Events/Alerts THREAT CASE #7: +70% Events/Alerts THREAT CASE #8: +80% Events/Alerts THREAT CASE #9: +90% Events/Alerts
35
38
42
45
49
52
56
59
63
66
Simulations have been carried out to determine the impact on the time to identify a security incident (hence the impact on risk exposure) and the actual allocation of resources over time. Section 9 presents the results.
Page 35 of 80
HP Security Analytics
Final Report
Page 36 of 80
HP Security Analytics
Final Report
Page 37 of 80
HP Security Analytics
Final Report
7.2.1.2 Initial Assessment Phase Figure 7-20 shows the process steps involved in the incident assessment phase for Major Security Incidents:
This phase involves an exchange of available information between the SOC Team and the Risk Management team (duration: 1 hour -1 day) along with a joint analysis of the incident (duration: 1 hour 1 day). These steps are followed by the SOC Team writing a summary report (duration: 1 hour4 hours). This step might be followed by the need to collect additional information from the Risk Management Team, in terms of priorities, risk mitigation practices and directions. 7.2.1.3 Information Gathering from Risk Management Team Figure 7-21 describes the process steps involved in this phase:
Page 38 of 80
HP Security Analytics
Final Report
In 70% of cases, additional information about the security incident needs to be collected from the Risk Management Team. In this case a request for additional information is made to the Risk Management Team (duration: 1-4 hours). The requested information is usually provided in a timeframe of 1 hour -8 hours. This step is followed by an additional joint analysis involving the SOC Team and the Risk Management Team (duration: 1 hour 1 day). This step is usually followed by an agreement, between the SOC Team and the Risk Management team, about the remediation requirements, to handle the security incident. This step usually requires 12 hours. 7.2.1.4 TP Handling of Incident Response Figure 7-22 describes the process steps involved in this phase:
This is the phase where a TP is in charge of handling their response to the security incident. The SOC Team provides the TP with recommendations on how to handle the incident (duration: 1 hour 2 days). This step is potentially followed by a set of interactions between the SOC Team and the TP: The TP provides a remediation document to the SOC Team (duration: 5 days -40 days); The SOC Team carries out an analysis of the outcomes and the proposed mitigations (duration: 4 hours -2 days);
Page 39 of 80
HP Security Analytics
Final Report
The SOC Team might decide that the TP response is inadequate. This happens in 45% of cases. In this situation, an additional request for information is made by the SOC Team to the TP (duration: 1-4 hours). In general, the empirical data suggests that there might be between 1 and 3 iterations before a satisfactory response is provided by the TP.
In case the TP response is satisfactory, a decision is made by the SOC Team about how to proceed. In 90% of cases the TP will remediate the incident. However in 10% of cases, this might not be possible (e.g. it might not be obvious how to proceed, or the TP might not have the means/controls/tools to do it). In this case the SOC Team will raise the risk and deal with related risk management steps. 7.2.1.5 TP Incident Remediation Figure 7-23 describes the process steps involved in this phase:
An incident report is written (duration: 2-3 days), followed by the actual remediation of the Security Incident by TP (duration: 5 - 40 days) 7.2.1.6 Risk Management Figure 7-24 describes the process steps involved in this phase:
An incident report is written (duration: 2-5 days) followed by risk management steps (beyond the scope of this modelling activity).
7.2.2 Metrics
Also this model has been built to carry out the investigations discussed in Section 3, specifically to explore the risk exposure and performance issues due to the Incident Management Process. The model has been instrumented with a set of metrics to convey outcomes in terms of risks. This includes the following Risk Indicators:
Page 40 of 80
HP Security Analytics
Final Report
Time to Remediate an Incident: the longer the time to remediate the incident, the longer the organisation is exposed to the potential risks, such as exploitation of privileged accounts, spread of malware, exploitation of devices accessible via the Internet, etc; Time Spent in Iterations with TPs: this is a subset of the previous time. It specifically measures the time that has been spent in interactions with TPs. Also in this case, the longer the time, the longer it takes to deal with the security incident, hence further exposing ServiceBizWorld to risks; Time Spent to Deal with Risk Management: this is the time to process the incident and determine that it cannot be remediated, hence dealing with risk management; Various Counters: for completeness, counters have been introduced to count how many times an incident ends up to be remediated or in a Risk Management situation. However, their value is obvious, given the simple nature of this model, based on the known frequency of the security incidents (800-900 major security incidents per year) and the fact that 90% of them will be remediated.
Section 8 discusses how these metrics have been used to describe the outcomes of the simulations.
The second issue is related the TP Incident Remediation phase. As mentioned above, it might require a long period of time for TPs to remediate to incidents (up to 40 days).
Page 41 of 80
HP Security Analytics
Final Report
Also in this case various kinds of investments and solutions could be considered to potentially improve the performance. For the purpose of this what-if analysis, the following types of solutions have been considered, each of them with an increasing level of impact and automation: Manual templates and scripts to carry out remediation activities for known, recurrent security incidents; Automated templates, knowledge-base and scripts to carry out remediation activities; Change management tools and solutions to track and automate the overall remediation activities (such as [9]).
The combined deployment of these controls and solutions at the TP sides is likely to affect the following aspects: Number of required iterations with TPs to obtain a satisfactory response in terms of remediation plans/documents; Time required by TPs to provide the remediation plans/documents; Time required by TPs to remediate the security incidents.
Figure 7-25 shows the cases and assumptions that have been explored in this what-if analysis involving the deployment of combinations of the above controls:
Cases Additional Capabilities/Controls Parameters in the Model
CURRENT
NONE
TP Handling of Incidence Response Time to Produce Remediation Documents: [5 days, 40 days] Num TP Iterations: (1,3)
CASE #1
TP Handling of Incidence Response Time to Produce Remediation Documents: [4 days, 32 days] Num TP Iterations: (1,2)
CASE #2
TP Handling of Incidence Response Time to Produce Remediation Documents: [2.5 days, 20 days] Num TP Iterations: (1,2)
Page 42 of 80
HP Security Analytics
and scripts TP Incident Remediation Time to TP Incident Remediation: [2.5 days, 20 days]
Final Report
CASE #3
TP Incident Remediation
Tools
and
Simulations have been carried out to determine the impact on the time to remediate an incident, the time to decide to do risk management and the time spent in TP interactions. These metrics give an indication both of the performance of the process and the potential risk exposure for ServiceBizWorls. Section 9 presents the results.
Page 43 of 80
HP Security Analytics
Final Report
Frequency
Days
Figure 8-1. Distribution of Time to Identify Security Incidents
The graph shows that there is lot of variability in the time required to identify security incidents, in the [2 days to 50-60 days] range. Most occurrences are in the [5-40] day range. Just a small percentage requires a period of time less than 9-10 days, reflecting the case where no TP interactions are required. This graph illustrates the potential impact on ServiceBizWorlds risk exposure due to the delays introduced to identify a security incident and act on it. The very long involved time might give opportunities to attackers for further exploitations. This is particularly important for Major Security Incidents, as the same time distribution applies, see Figure 8-2:
Page 44 of 80
HP Security Analytics
Final Report
Days
It is important to notice that to the time shown in Figure 8.2, additional time has to be added before a Major Security Incident is actually remediated, as discussed in the next section. This further highlights the potential risk exposure for the company. The SOC Team needs to process about 12000 events/alerts per year, of which 800-900 are classified as Major Security Incidents. As there is currently no specific process differentiation, their handling is affected by the processing workload introduced by the other events/alerts. This might be an area subject to further exploration and analysis. A critical aspect, confirmed by this analysis, is the impact that TP Interactions have on the overall results, as shown in Figure 8-3:
Frequency 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75
Figure 8-3. Distribution of Time Spent in TP Iterations
Days
Page 45 of 80
HP Security Analytics
Final Report
This time distribution has a wide spread, with about 80% of cases ranging between 12 and 50 days. This creates a bottleneck that affects the entire process and has a negative impact not only on the overall performance but also on ServiceBizWorlds risk exposure. Figure 8-4 shows the time to deal with TP non compliance, i.e. the time to track and handle TP failures to provide the required information in due time to the SOC Team:
Frequency
Days
Given the nature of the current process, TP non compliance issues can be raised by the SOC Team at any time during the execution of the process itself. Once again, the graph shows the potential large delays in identifying situations that affect risk exposure and performance. This is a direct consequence of the potential large period of time spent in the process to deal with TP Iterations. Figure 8-5 shows the distribution of the time to process events/alerts that ends up in their closure:
0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75
Figure 8-5. Distribution of Time to Process the Closure of Events/Alerts
Page 46 of 80
HP Security Analytics
Final Report
This graph clearly shows that events/alerts can be closed in 2-10 days, if the SOC Team has the information required for their processing and no TP interaction is required. This situation however covers about 25% of the cases. If the SOC Team has to interact with TPs, the required time further increases, up to 50-60 days. Figure 8-6 shows the distribution of the time required by the SOC Team to process false positives:
Frequency
64
12
16
20
24
28
32
36
40
44
48
52
56
60
68
72
The graph shows that, in about 40% of cases, false positives are identified in the early stages of the process and dealt within 1-2 days. This is an encouraging result in terms of performance, in the sense that the time required for their processing is minimal, and the SOC Team resources can be used for other tasks. However, about 40% of false positives are handled in the [25,50] days range, reflecting the fact that 20% of the processed events/alerts are classified as false positives only after TP interactions. Figure 8-7 shows the distribution of the time to handle known problems:
Frequency
Hours
Page 47 of 80
More
Days
HP Security Analytics
Final Report
This time is in the range of [5,50] hours (i.e. max 2 days) reflecting the fact that all known problems are identified in the early stages of the process and efficiently managed by the SOC Team. Finally, Figure 8-8 provides a breakdown of the types of outcomes of the process, based on various counters (metrics) introduced in the model:
counterMinorSecurityIncident
counterMajorSecurityIncident counterEvent/AlertClosure counterClosureFalsePositive counterClosureKnownProblem
This figure shows that about 65% of the events/alerts are eventually classified as minor security incidents. TP Non compliance affects almost 35% of the processed events/alerts, reflecting the empirical data introduced in the model. False positives account for 20-25% of the overall events/alerts.
8.1.1 Summary
The analysis of the current situation highlights the fact that TP interactions have a major impact on ServiceBizWorlds risk exposures, along with potential performance implications. This issue is well known by the SOC Team and the Risk Management Team: this analysis confirms it by quantifying the implications and adding scientific evidence. The analysis shows that TP interactions usually account for 60-70% of the involved processing time. The delay introduced by these interactions is reflected in all metrics analysed in the simulations. To conclude, the impact in terms of additional exposure to risks is evident, due to the introduced delays: this is particular important in the context of Major Security Incidents. Section 9 illustrates, by means of what-if analysis, how the current situation could be improved by introducing specific SLAs for TPs, along with making investments in specific controls/solutions.
Page 48 of 80
HP Security Analytics
Final Report
The result shows a huge variability in the time required to remediate a major security incident, ranging from a few days up to 5 months. The outcomes directly reflect the empirical data available. The main issues are around the potential delays introduced by TP iterations to deal with the incident response and the actual remediation time required by TPs, that might take up to 40 days. This reflects the complex nature of dealing with this type of incidents as well as specific issues within TP organisations to timely react and remediate them. The implications, in terms of risk exposure for ServiceWorldBiz, are evident: the potential critical vulnerabilities and issues that are the root cause of the security incident might be exploited by attackers before they are remediated. Figure 8-10 shows the distribution of the time to Risk Management, i.e. the case when the incident cannot be remediated and the involved risk has to be handled:
0 7 14 21 28 35 42 49 56 63 70 77 84 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 196
Figure 8-9. Distribution of Time to Remediate a Major Security Incident
Days
Page 49 of 80
HP Security Analytics
Final Report
Frequency
0.025 0.02 0.015 0.01 0.005 0
The graph shows that it might take up to 4 months to conclude that the major security incident cannot be remediated and the risk has to be managed. This is again a reflection of the time spent in TP iterations. Finally, Figure 8-11 shows the distribution of time that is actually spent in TP iterations to handle the incident response:
Frequency
0.025 0.02 0.015 0.01 0.005 0
0 7 14 21 28 35 42 49 56 63 70 77 84 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 196
Figure 8-10. Distribution of Time to Risk Management
Days
The result confirms that it might take up to 4 months to interact with TPs, to handle the incident response. This influences the other process outcomes.
8.2.1 Summary
The analysis of the current situation confirms the fact that, also in this process, TP interactions have a major impact on ServiceBizWorlds risk exposure, along with potential performance implications. These results reinforce the findings of the analysis carried out in the first process. Noticeably, it might take up to 7 months to fully remediate a major security incident, if we consider the time required to identify the security incident and the actual time required to mitigate it.
0 7 14 21 28 35 42 49 56 63 70 77 84 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 196
Figure 8-11. Distribution of Time spent in TP Iterations
Days
Page 50 of 80
HP Security Analytics
Final Report
There are potential margins for improvement involving investments in controls/solutions and streamlining/optimising the activities carried out by TPs. A few potential options are explored in Section 9, by means of a what-if analysis.
Page 51 of 80
HP Security Analytics
Final Report
9.1 Model 1: Security Event/Alert Processing and Resolution 9.1.1 Impact of Changing SLAs for TPs
This what-if analysis aims at exploring the impact of changing SLAs for TPs, based on the set of cases described in Section 7. The goal is to explore the improvement that this could bring by reducing the bottleneck with TPs. Figure 9-1 summarises the four cases that have been explored:
Cases TP Responding in Time (Aggregated) CURRENT 60%
(single iteration: 92%)
# TP Iterations
70%
5-7 (average: 6)
CASE #1
70%
(single iteration: 93.1%)
78%
4-6 (average: 5)
CASE #2
80%
(single iteration: 94.6%)
86%
3-5 (average: 4)
CASE #3
90%
(single iteration: 96.5%)
93%
2-4 (average: 3)
CASE #4
100%
(single iteration: 100%)
100%
1-2
For each case, comparative graphs (for each of the agreed metrics) have been created, to illustrate the impact of the SLA changes against the current situation. Figure 9-2 shows the impact that different SLAs have on the time to identify security incidents:
Page 52 of 80
HP Security Analytics
Final Report
0.1 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01
Frequency
Days
Figure 9-2. Comparative Analysis of Time to identify Security Incidents based on different SLAs
The graph shows a major improvement in the overall time required to identify a security incident depending on the SLA case. In the best case scenario (Case 4), the required time is almost 1/3 of the current one, with an average of 6-15 days compared to the current 20-45 days. Further improvements could be made in other parts of the process, including the initial and final assessment phases carried out by the SOC Team. Being able to reduce this timeframe is important as it further reduces ServiceBizWorlds risk exposure to security incident, in particular for Major Security Incidents (e.g., exposures of companys systems to attacks, exploitation of privileged accounts, etc.). The sooner the root cause of the incident is identified, the sooner the remediation process can start and the risk be mitigated. Figure 9-3 shows the specific impact that different SLAs have on the time to identify major security incidents: Frequency
Days
Figure 9-3. Comparative Analysis of Time to identify Major Security Incidents based on different SLAs
The graphs show similar patterns and improvements to the ones shown in Figure 9-2. The graph highlights that it takes at least 6-9 days to identify Major Security Incidents, independently from the SLA improvements. This is due to the fact that, independently from the TP interactions, time is spent
Page 53 of 80
HP Security Analytics
Final Report
by the SOC Team to carry out various analysis steps: this aspect could be potentially improved by further streamlining the SOC Team activities. Furthermore, the overall time is also affected by the need to produce Incident Reports and recording the incident: it currently takes an additional 1 day. Figure 9-4 shows the specific impact that different SLAs have on the time spent in TP iterations: Frequency
0.25
0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75
Figure 9-4. Comparative Analysis of Time Spent in TP Iterations based on different SLAs
Days
The graphs shows that this time can be drastically reduced, at least for Cases 3 and 4 (1/3 and 1/5 respectively of the current time). The reduction of this time positively affects the overall process as currently TP interactions are the major bottleneck. Figure 9-5 shows the specific impact that different SLAs have on the time to report TP non compliance:
Page 54 of 80
HP Security Analytics
Final Report
Frequency
0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0
The graph shows improvements, nevertheless the time to report TP non compliance is still affected by the fact that violations can happen at any time in the interaction process with TPs. Apart from the ideal Case 4, where the SLA assures that there should be no violations, Case 3 highlights that the reporting of TP non compliance situations can still happen in a range of [2,30] days, reflecting the empirical data in the model. Further improvements can be introduced by reducing the deadlines by which the TPs are requested to provide the information to the SOC Team (after which the SOC Team escalates the violations to the Risk Management Team). Figure 9-6 shows the specific impact that different SLAs have on the time to close events/alerts:
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75
Days
Figure 9-5. Comparative Analysis of Time to Report TP Non Compliance based on different SLAs
Page 55 of 80
HP Security Analytics
Final Report
Frequency
Days
This graph shows a similar improvement trend to the ones shown before, based on better and better SLAs. In the best case, events/alerts can be closed in 3-18 days, against the current 24-60 days. Figure 9-7 shows the specific impact that different SLAs have on the time to close false positives: Frequency
0.25
0.05
0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75
Figure 9-7. Comparative Analysis of Time to Close False Positives based on different SLAs
Days
This figure shows the specific impact of the SLA changes on handling false positives. SLAs changes have no impact if the false positives are discovered in the early stages of the process (10% of cases),
Page 56 of 80
HP Security Analytics
Final Report
and/or if no TP interactions are involved. In this case the closure time still ranges between 1 and 3 days. However, the major impact is when the false positives are discovered after TP interactions. In the best case, the time to handle them is in the [3,18] day range, a major improvement compared to the current [24,50] days. In this context, SLA improvements have a major productivity impact, as less time is spent to process false positives and resources can be utilise to handle other types of events/alerts and incidents. Finally, Figure 9-8 shows the impact that SLAs have on different counters used as metrics in the model:
9000 8000 7000 6000 5000 4000 3000 2000 1000 0 Num False Num Known Num Positive Problem Event/Alert Closure Num Major Security Incident Num Minor Num TP Non Security Compliance Incident CURRENT CASE1 CASE2 CASE3 CASE4
As expected, only the number of TP Non Compliance is directly affected by SLA changes. All other events (and related counters) are unaffected. 9.1.1.1 Summary The analysis of this what-if scenario shows that changes of SLAs for TPs can have a major impact in reducing risks for ServiceBizWorld and potentially improving productivity. This is particularly true in case of: handling security incidents where the time to identify them can be reduced from the current [20, 45] days to [6,15] days; handling false positives in [3,18] days, compared to the current [24,60] days. It is beyond the scope of this analysis to suggest specific controls and contractual agreements with TPs in order to achieve the SLAs highlighted in the various cases.
Page 57 of 80
HP Security Analytics
Final Report
CURRENT
35
38
42
45
49
52
56
As already mentioned in Section 7, it is important to notice that in case of resource saturations, the delays are likely to increase over time. Hence results depend on the simulated time period. In the context of this what-if analysis, a period of time of 6 months has been considered. We wanted to explore the implications for the organisation of taking no actions during that entire period of time. This note applies to all the what-if scenarios discussed in this section, involving allocation of resources/personnel. The analysis can be tuned on a different timeframe, depending on needs and requirements. Based on the assumptions discussed in Section 7, Figure 9-10 shows the current situation in terms of time distribution to identify security incidents and the (predicted) allocation of TP resources, over time (in days):
Page 58 of 80
HP Security Analytics
Final Report
Allocated TP Resources
50
100
150
200
Figure 9-10. Current Situation Time to identify Security Incidents and TP Resource Allocation
Figure 9-11 shows the predicted impact of changes of the threat environment, for Case 1 (+10% events): Time to Identify Security Incidents
0.1 0.08 0.06 0.04 0.02 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 120% 100% 80% 60% 40% 20% 0% 0 50 100 150 200
Allocated TP Resources
Figure 9-11. Case 1 Time to identify Security Incidents and TP Resource Allocation
Figure 9-12 shows the predicted impact of changes of the threat environment, for Case 2 (+20% events):
Time to Identify Security Incidents
0.1 0.08 0.06 0.04 0.02 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 120% 100% 80% 60% 40% 20% 0% 0 50 100 150 200
Allocated TP Resources
Figure 9-12. Case 2 Time to identify Security Incidents and TP Resource Allocation
Figure 9-13 shows the predicted impact of changes of the threat environment, for Case 3 (+30% events):
Page 59 of 80
HP Security Analytics
Final Report
Allocated TP Resources
50
100
150
200
Figure 9-13. Case 3 Time to identify Security Incidents and TP Resource Allocation
Figure 9-14 shows the predicted impact of changes of the threat environment, for Case 4 (+40% events): Time to Identify Security Incidents
0.1 0.08 0.06 0.04 0.02 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 120% 100% 80% 60% 40% 20% 0% 0 50 100 150 200
Allocated TP Resources
Figure 9-14. Case 4 Time to identify Security Incidents and TP Resource Allocation
Figure 9-15 shows the predicted impact of changes of the threat environment, for Case 5 (+50% events): Time to Identify Security Incidents
0.12 0.1 0.08 0.06 0.04 0.02 0 0 5 10 15 20 25 30 35 40 45 50 55 60 65 70 75 120% 100% 80% 60% 40% 20% 0% 0 50 100 150 200
Allocated TP Resources
Figure 9-15. Case 5 Time to identify Security Incidents and TP Resource Allocation
Figure 9-16 shows the predicted impact of changes of the threat environment, for Case 6 (+60% events):
Page 60 of 80
HP Security Analytics
Final Report
Allocated TP Resources
50
100
150
200
Figure 9-16. Case 6 Time to identify Security Incidents and TP Resource Allocation
Figures 9-11, 9-16 show the increasing negative impact that the worsening of the threat landscape has of the overall process, in particular with regards to the time to identify security incidents. These simulations highlight that the process copes well in case of an increase of events/alerts up to 20%: in the simulated timeframe (6 months) the time to identify incidents is not affected and the allocated TP resources. These resources can deal with the increased workload without major delays, despite the fact that they increasingly saturate. Some initial delays are already noticeable with an increase of 30% of the events/alerts. To summarise, Figure 9-17 provides a comparative analysis of the time to identify security incidents, across all Cases explored in this what-if analysis: Time to Identify Security Incidents
0.12
0.1 Current 0.08 "+10%" "+20%" 0.06 "+30%" "+40%" "+50%" "+60%"
0.04
0.02
0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75
Figure 9-17. Comparative Analysis Time to identify Security Incidents for various Cases
Days
An increase of 30% of alerts/events or more causes the saturation of the allocated TP resources: TPs will increasingly have problems to cope with the workload. This is reflected in the figure, comparing the various distributions of time to identify security incidents. More and more time is required to deal with events/incidents. Case 5 and 6 explicitly show the impact of the accumulation of delays, due to
Page 61 of 80
HP Security Analytics
Final Report
the resource saturation (in the considered timeframe): the required time to process incidents can exceed 75 days. It is important to notice that these outcomes reflect the assumptions initially made in Section 7- in the situation where no changes are made in the 6 months timeframe. These results are potentially sensitive to the parameters used in the model. However, these results provide a qualitative analysis of the negative impact that changes in the threat environment have on the overall process and the consequences for ServiceBizWorld risk exposure (all other conditions being the same): the more the threat environment get worst, the more time is required to process events/alerts, hence ServiceBizWorld is exposed to additional risks. Finally, Figure 9-18 provides a summary of the potential additional risk exposures for ServiceBizWorld, by means of a comparative view for the different threat environment cases. A green box indicates that the risk exposure situation is comparable to the current one (or improved). An orange box indicates that the risk exposure situation is potentially getting worst (moving towards TP resource saturation and/or increase of the time to deal with security incidents). A red box indicates a potential critical risk exposure for ServiceBizWorld, involving major saturation of TP resources and time delays:
Current Case 1 (+10%) Case 2 (+20%) Case 3 (+30%) Case 4 (+40%) Case 5 (+50%) Case 6 (+60%)
Figure 9-18. Comparative Analysis Risk Exposure due to Changes of the Threat Environment
This notation is also used in the next what-if scenario. 9.1.2.1 Summary The analysis of this what-if scenario illustrates (in a qualitative way) the potential implications for the Event/Alert Processing & Resolution process in case of the worsening of the threat landscape. The bottleneck introduced by TP interactions is critical. Simulations, based on assumptions made in terms of available TP resources, illustrates that the situation can rapidly get worse and worse if the number of new events/alerts to be processed (on daily bases) increases more than 20%. This has a negative impact on ServiceBizWorlds risk exposure, due to the increased time to identify security incidents, in particular Major Security Incidents. This analysis provides additional evidence to ServiceBizWorld and the SOC Team to enable contingency planning discussions with TPs.
Page 62 of 80
HP Security Analytics
Final Report
CURRENT
NONE
CASE #1
CASE #2
As above. In addition: -
Improved Monitoring Solutions at TP Next Iterations: [0.225 days, 0.25 days] sites Automated execution of scripts to Actual work: [0.25 days, 1.25 days] retrieve data Note: Improvement of 50% compared to current situation
CASE #3
As above. In addition:
Overhead: 1st Iteration: [0.1 days, 1.5 days] Next Iterations: [0.09 days, 0.1 days] Actual work: [0.1 days, 0.5 days] Note: Improvement of 80% compared to current situation
This first investigation assumes that the number of involved iterations with TPs is the same as in the current situation. As discussed in Section 7, for each of the 3 investment cases shown in Figure 9-19, models have been built and used in simulations to predict the outcomes (i.e. time to identify a security incident and resource allocation) for the different threat environment scenarios shown in the following Figure 9-20:
Page 63 of 80
HP Security Analytics
Case Events/Alerts per Day
Final Report
CURRENT THREAT CASE #1: +10% Events/Alerts THREAT CASE #2: +20% Events/Alerts THREAT CASE #3: +30% Events/Alerts THREAT CASE #4: +40% Events/Alerts THREAT CASE #5: +50% Events/Alerts THREAT CASE #6: +60% Events/Alerts THREAT CASE #7: +70% Events/Alerts THREAT CASE #8: +80% Events/Alerts THREAT CASE #9: +90% Events/Alerts
35
38
42
45
49
52
56
59
63
66
Related to the Investigation 1- Case 1 (Figure 9-19: definition of templates and manual scripts, with an overall performance improvement of 20%), Figure 9-21 compares the impact that different threat scenarios have on the time to identify security incidents:
Page 64 of 80
HP Security Analytics
Final Report
0.1 0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0
Days
Figure 9-21. What-if Analysis Case 1: Comparative View based on Changes of the Threat Landscape
This figure shows that, based on simulations, the controls and solutions defined in Case 1 introduce a major improvement, compared to the situation discussed in the previous what-if scenario. Basically, there are no major delays in the time to identify security incidents in case the number of events/alerts increases up to 50% (on daily basis). However, after that threshold, the situation quickly deteriorates, due to the saturation of the TP resources. This is a noticeable improvement, compared to the current process and controls, where the situation can rapidly get worst if the number of new events/alerts to be processed increases more than 20% on daily basis (see the what-if analysis in the previous section). Related to the Investigation 1- Case 2 (Figure 9-19: use of monitoring tools and script automation, with an overall performance improvement of 50%), Figure 9-22 compares the impact that different threat scenarios have on the time to identify security incidents: Case 2 - Time to Identify Security Incidents
Current "+10%" "+20%" "+30%" "+40%" "+50%" "%60" "+70%" "+80%" "+90%" 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75
Days
Figure 9-22. What-if Analysis Case 2: Comparative View based on Changes of the Threat Landscape
Page 65 of 80
HP Security Analytics
Final Report
This figure shows that, based on simulations, the controls and solutions defined in Case 2 introduce an even bigger improvement. There are no noticeable changes of the time to identify security incident in case the number of events/alerts increases up to 90% (on daily basis). The results also show a reduction of the overall time to identify a security incident, consistently with the assumptions that have been made. Finally, related to the Investigation 1 - Case 3 (Figure 9-19: use of automated workflows, with an overall performance improvement of 80%), Figure 9-23 compares the impact that different threat scenarios have on the time to identify security incidents: Case 3 - Time to Identify Security Incidents
Current "+10%" "+20%" "+30%" "+40%" "+50%" "+60%" "+70%" "+80%" "+90%" 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75
Days
Figure 9-23. What-if Analysis Case 3: Comparative View based on Changes of the Threat Landscape
This figure confirms and consolidates the previous trend. There are no noticeable changes of the time to identify security incident in case the number of events/alerts increases up to 90% (on daily basis). The results also show a significant reduction of the overall time to identify a security incident, consistently with the assumptions that have been made. Finally, Figure 9-24 provides a summary of the potential additional risk exposures for ServiceWorldBiz, by means of a comparative view of the different threat environment cases. A green box indicates that the risk exposure situation is comparable to the current one (or improved). An orange box indicates that the risk exposure situation is potentially getting worst (moving towards TP resource saturation and/or increase of the time to deal with security incidents). A red box indicates a potential critical risk exposure for the company, involving major saturation of TP resources and time delays:
Page 66 of 80
HP Security Analytics
CURRENT Threat Environment
Final Report
+20% +30% +40% +50% +60% +70% +80% +90%
+10%
CURRENT
Case 1
Case 2
Case 3
Figure 9-24. Comparative Analysis Risk Exposure due to Changes of the Threat Environment
It is important to notice, that in this first investigation, it was assumed that the new controls and solutions do not affect the overall number of iterations with a TP required to gather the information (requested by the SOC Team). In other words, this number of TP iterations (i.e. 4-7, with an average of 5 iterations for each processed event/alert) is exactly the same as in the current process. The additional investments primarily reduce the time spent by TP resources to deal with information requests from the SOC team. The next investigation, i.e. Investigation 2, considers a more optimistic scenario where the improvement of the information availability for TPs has also a positive impact on the number of TP iterations, by progressively reducing it, for each analysed case. Figure 9-25 shows the cases that have been analysed, in the second investigation, as discussed in Section 7:
Investigation 2 - Cases Additional Capabilities/Controls Parameters in the Model (Information Gathering Phase) Number of TP Iterations (per event/alert)
CURRENT
NONE
Overhead:
1st Iteration: [0.5 days, 7.5 days] Next Iterations: [0.45 days, 0.55 days]
Actual work: [0.5 days, 2.5 days] SOC Teams templates defining the types of required information Manual scripts, at the TP site, describing how to retrieve data
CASE #1
Overhead:
1st Iteration: [0.4 days, 6 days] Next Iterations: [0.35 days, 0.4 days]
Actual work: [0.4 days, 2 days] Note: Improvement of 20% compared to current situation
CASE #2
As above. In addition: Improved Monitoring Solutions at TP sites Automated execution of scripts to retrieve data
Overhead:
1st Iteration: [0.25 days, 3.75 days] Next Iterations: [0.225 days, 0.25 days]
Actual work: [0.25 days, 1.25 days] Note: Improvement of 50% compared to current situation
Page 67 of 80
HP Security Analytics
CASE #3 As above. In addition: Workflow automation spanning across the SOC Team and TPs Overhead: 1st Iteration: [0.1 days, 1.5 days] Next Iterations: [0.09 days, 0.1 days] Actual work: [0.1 days, 0.5 days] Note: Improvement of 80% compared to current situation
Final Report
1-2 iterations (50% chances 1 or 2)
Again, as discussed in Section 7, for each of the 3 investment cases shown in Figure 9-25, models have been built and used in simulations to predict the outcomes (i.e. time to identify a security incident and resource allocation) for the different threat scenarios shown in Figure 9-20. Related to the Investigation 2 - Case 1 (Figure 9-25: definition of templates and manual scripts, with an overall performance improvement of 20% and 4-6 TP iterations per event/alert), Figure 9-26 compares the impact that different threat scenarios have on the time to identify security incidents: Case 1 - Time to Identify Security Incidents
0.09 0.08 CURRENT 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75 "+10%" "+20%" "+30%" "+40%" "+50%" "+60%" "+70%" "+80%" "+90%"
Days
Figure 9-26. What-if Analysis Case 1: Comparative View based on Changes of the Threat Landscape
The results show a neat improvement in terms of the overall performance, across the various threat environment scenarios, compared to the results shown in Figure 9-21, in particular for those cases involving an increase of the number of alerts/events greater than 60%. This is due to the reduced number of interactions with TPs. However simulations show that the TP resources still saturates for increases of number of events/alerts greater than 80% and delays are introduced in the process. Related to the Investigation 2 - Case 2 (Figure 9-25: use of monitoring tools and script automation, with an overall performance improvement of 50% and 2-3 TP iterations per event/alert), Figure 9-27 compares the impact that different threat scenarios have on the time to identify security incidents:
Page 68 of 80
HP Security Analytics
Case 2 - Time to Identify Security Analytics
0.09 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0
Final Report
CURRENT "+10%" "+20%" "+30%" "+40%" "+50%" "+60%" "+70%" "+80%" "+90%"
These results are quite positive. If the deployed controls effectively reduce the number of TP iterations as assumed, no noticeable delays and TP resource saturations are observed, even in the worst case threat environment considered in this what-if scenario (90% increase of events/alerts). The number of TP iterations indeed decreases the overall time to identify security incidents of about 20-30%, compared to the times shown in Figure 9-27. Related to the Investigation 2 - Case 3 (Figure 9-25: use of automated workflows, with an overall performance improvement of 70% and 1-2 TP iterations per event/alert), Figure 9-28 compares the impact that different threat scenarios have on the time to identify security incidents:
0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75
Days
Figure 9-27. What-if Analysis Case 2: Comparative View based on Changes of the Threat Landscape
Page 69 of 80
HP Security Analytics
Final Report
0.1 0.09
CURRENT 0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 0 3 6 9 12 15 18 21 24 27 30 33 36 39 42 45 48 51 54 57 60 63 66 69 72 75 "+10%" "+20%" "+30%" "+40%" "+50%" "+60%" "+70%" "+80%" "+90%"
Days
Figure 9-28. What-if Analysis Case 3: Comparative View based on Changes of the Threat Landscape
Also these results confirm the trends shown in Figure 9-23. If the deployed controls effectively reduce the number of TP iterations as assumed, no noticeable delays and TP resource saturations are observed, even in the worst case threat environment considered in this what-if scenario (90% increase of events/alerts). The number of TP iterations further decreases the overall time to identify security incidents of about 40%, compared to the times shown in Figure 9-23. Finally, Figure 9-29 provides a summary of the potential additional risk exposures for ServiceBizWorld, by means of a comparative view of the different threat environment cases. A green box indicates that the risk exposure situation is comparable to the current one (or improved). An orange box indicates that the risk exposure situation is potentially getting worst (moving towards TP resource saturation and/or increase of the time to deal with security incidents). A red box indicates a potential critical risk exposure for the company, involving major saturation of TP resources and time delays:
Page 70 of 80
HP Security Analytics
CURRENT Threat Environment +10% +20% +30% +40% +50% +60% +70%
Final Report
+80% +90%
CURRENT
Case 1
Case 2
Case 3
Figure 9-29. Comparative Analysis Risk Exposure due to Changes of the Threat Environment
We can notice a sensible improvement in the context of Case 1, where the situation is now critical only when the increase of events/alerts is greater than 80%. Simulations highlight the saturation of TP resources only in that situation. In all other cases (in the simulated timeframe) there are no resource saturation and delays. 9.1.3.1 Summary The analysis of this what-if scenario illustrates (in a qualitative way) that by tackling the issues related to TP information availability, it is possible to get major improvements in the overall performance of the process, and contain the ServiceBizWorlds risk exposure to situations where the threat landscape deteriorates. Based on the assumption made, the simple deployment of templates with just manual scripts improves the situation: it ensures that the overall process can cope with an increase of 60% of the current number of events/alerts, without degradations and negative impacts of ServiceBizWorlds risk exposure: the increase can be of 80% if the controls also sensibly reduce the number of TP iterations. The deployment of templates and automated scripts or automated workflows, ensures that the overall process can cope with an increase of 90% (or more) of the current number of events/alerts, without degradations and negative impacts of the companys risk exposure. Again, the main goal of this analysis is to provide additional evidence to ServiceBizWorld and the SOC Team to enable contingency planning discussions with TPs.
Page 71 of 80
HP Security Analytics
Final Report
9.2 Model 2: Incident Management Process 9.2.1 Impact of Investments to Improve the Process Performance
This what-if analysis aims at investigating the impact of making investments in controls and solutions, based on the set of cases described in Section 7. The goal is to explore the improvement in the process performance and related risk reductions, by both reducing the bottleneck introduced by TPs and the involved remediation time. Figure 9-30 summarises the three cases that have been explored (against the current situation):
Cases Additional Capabilities/Controls Parameters in the Model
CURRENT
NONE
TP Handling of Incidence Response Time to Produce Remediation Documents: [5 days, 40 days] Num TP Iterations: (1,3)
CASE #1
TP Handling of Incidence Response Time to Produce Remediation Documents: [4 days, 32 days] Num TP Iterations: (1,2)
CASE #2
TP Handling of Incidence Response Time to Produce Remediation Documents: [2.5 days, 20 days]
TP Incident Remediation -
Num TP Iterations: (1,2) Automated templates, Knowledge base TP Incident Remediation and scripts Time to TP Incident Remediation: [2.5 days, 20 days]
CASE #3
TP Handling of Incidence Response Time to Produce Remediation Documents: [1 day, 8 days] Num TP Iterations: (1)
TP Incident Remediation
Page 72 of 80
HP Security Analytics
Change Management Solutions Tools and TP Incident Remediation Time to TP Incident Remediation: [1 day, 8 days]
Final Report
For each case, comparative graphs have been created, to illustrate the impact of the additional controls and solution against the current situation. Specifically, simulations have been carried out to explore the implications in terms of time to remediate a major security incident and time spent in TP Iterations. Figure 9-31 shows the impact that the new controls have on the time to remediate a major security incident, for all the above three cases: Time to Remediate Major Security Incidents
0.14
0.12
0.04
0.02
Based on the assumptions made, in case 3 (involving automated solutions and controls) the time to remediate a major security incident is in the range of [7,21] days, compared to the [21,150 days that might take in the current situation. The improvements in terms of performance and reduction of risk exposure for ServiceBizWorld are evident. Interestingly, it takes at least 7 days to remediate an incident: this is due to other process steps (including writing reports) that are not affected/improved by these controls. Figure 9-32 shows the specific impact that the new controls have on the time spent in TP iterations, for all the above three cases:
0 7 14 21 28 35 42 49 56 63 70 77 84 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 196
Figure 9-31. Comparative Analysis: Time to Remediate Major Security Incidents
Days
Page 73 of 80
HP Security Analytics
Final Report
The results directly reflect the assumptions made in this what-if scenario. The improvement is evident. The deployed controls increasingly reduce the overall time spent in TP interactions, from the current [5,120] days to [1,8] days, in Case 3. Finally, Figure 9-33 shows the specific impact that the new controls have on the time spent to arrive to the decision of dealing with risk management, for all the above three cases:
0 7 14 21 28 35 42 49 56 63 70 77 84 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 196
Figure 9-32. Comparative Analysis: Time Spent in TP Iterations
Days
0 7 14 21 28 35 42 49 56 63 70 77 84 91 98 105 112 119 126 133 140 147 154 161 168 175 182 189 196
Figure 9-33. Comparative Analysis: Time to Risk Management
Days
Page 74 of 80
HP Security Analytics
Final Report
The results show the neat improvement of the time required to deal with the risk management steps in the process: this is a direct consequence of the improvement of the time required to interact with TPs. Case 3 shows that this improvement is of an order of magnitude compared to the current situation. 9.2.1.1 Summary The results shown in Figures 9-31, 9-32 and 9-33 are intuitive and predictable, given the simplicity of this process and the direct impact that the various parameters have on the outcomes. The assumptions made in this what-if scenario are coarse grained, as it was not possible to gather more details about the actual implication and impact that the control and solutions have on the internal TP processes and their operational activities. However, these analytic results can be used as a starting point to engage with TPs, in order to explore options for performance improvement as well as to further reduce ServiceBizWorlds risk exposures related to major security incidents. The Security Analytics model, used in this what-if analysis, can be further refined to accommodate additional details, information and constraints when available from TPs.
Page 75 of 80
HP Security Analytics
Final Report
10 Conclusion
The findings of the study suggest that there is major room for improvement for risk reduction and increase of performance for the analysed processes. The analysis confirms that TP interactions are a major bottleneck both in the Event/Alert Processing & Resolution Process and for the Security Incident Management Process. Two key aspects have been identified as root causes: the lack of SLAs with TPs about how information should be provided to teh SOC Team; the lack of information availability for TPs. The what-if scenarios explored using the model of the Event/Alert Processing & Remediation Process show that: The definition of SLAs for TPs (e.g. about response time and number of iterations with the SOC team) can dramatically improve the overall performance and reduce the risk exposure by at least an order of magnitude; The current process might expose ServiceBizWorld to additional risks - in terms of delays to identify security incidents and saturation of TP resources/personnel - if the daily number of new events to be processed increases by about 30%; Investments in new controls not only can improve the overall TP information availability and performance (estimated to be an order of magnitude) but also help to cope with worsening threat landscapes. Specifically, the adoption of information request templates and manual scripts can ensure process resilience to a 60% increase of the daily number of new events/alerts to be processed. The introduction of automation of scripts and/or workflows enables process resilience to an increase of 90% (or more) of events/alerts.
Similarly, what-if scenarios for the Security Incident Management Process shows that major improvements can be obtained (for the overall time spent in TP interactions and for the time required by TPs to remediate to incidents) when templates and incident management support tools are deployed. These findings depend on the assumptions made and the information available at the analysis time. However they provide a starting point for conversations between ServiceBizWorld, TPs and the SOC Team, on expectations, policy & SLA changes, as well as on potential new investments, to improve the current process performances, reduce risk exposure and deal with contingency planning for worsening threat environments.
Page 76 of 80
HP Security Analytics
Final Report
11 References
[1] [2] ITIL, Incident and Incident Management, http://www.itil-officialsite.com/, 2011 Homeland Security, Build Security In, Incident Management, https://buildsecurityin.uscert.gov/bsi/articles/best-practices/incident.html, 2011 [3] HP ArcSight, SIEM solution, http://www.arcsight.com/, 2011 [4] RSA enVision, SIEM solution, http://www.rsa.com/node.aspx?id=3170, 2011 [5] Renaud Bidou, Security Operation Center (SOC) Concepts & Implementation, http://www.iv2technologies.com/SOCConceptAndImplementation.pdf [6] CERT, Computer Security Incident Response Team (CSIRT), http://www.cert.org/csirts/csirt_faq.html, 2008 [7] MetricStream, Collaborative Incident Analysis and Remediaiton Solutions, http://www.metricstream.com/solutions/it_incident_management.htm, 2011 [8] WolfPack, Incident Response Plan & Management, http://www.wolfpacsolutions.com/solutions/incident_response_plan, 2011 [9] BMC Remedy, Change Management, http://www.bmc.com/products/offering/itil-changeconfiguration-release-management-software.html [10] HP Security Analytics, http://h20195.www2.hp.com/V2/GetPDF.aspx/4AA3-2046EEW.pdf, 2011
Page 77 of 80
HP Security Analytics
Final Report
Type of Parameter
Value
Triage: Initial Assessment Phase 3 4 5 6 7 8 SOC Team: Initial Analysis of the Event/Alert Known security problem? --Threat Level Changed? Immediate need to Brief Risk Management Team? -- Prepare Briefing, Brief Team, Agree Remediation -- Incident Category Established? Activity Duration Probability/Likelihood Probability/Likelihood Probability/Likelihood Activity Duration Probability/Likelihood Probability of Major Security Incident: 10% 9 10 Input required from TP? SOC Investigation (case where TP Input not required) Probability/Likelihood Activity Duration between 1 hour and 2 days 80% between 5 hours and 1 day 80% between 1 hour and 2 days 10% 5% 15%
Information Gathering Phase 11 12 13 Request Input from TP Agree Timescale for Response TP responds within defined timeframe? (aggregated over the set of iterations with TP) Activity Duration Activity Duration Probability/Likelihood between 0.5 and 2 hours between 1 and 5 hours 60%
14 15 16
-- Failure to respond within the defined timeframe -- Escalate to Risk Management Team? -- Escalation Process
between 1 days and 10 days 10% between 1 hour and 3 hours First Iteration: between 1 and 10 days Follow up iterations: between 1 and 3 days between 1 hour and 2 days 20% between 1 and 3 hours between 1 and 2 hours 70% between 1 hour and 1 day 10% between 3 hours and 1 day
17 18 19 20 21 22 23 24 25
Data provided to SOC Team SOC Team Investigation Brief Risk Management Team? -- Risk Management Team Briefing Recommended Level Category Input of acceptable Quality and Quantity? SOC Team Investigation Brief Risk Management Team? -- Risk Management Team Briefing
Activity Duration Activity Duration Probability/Likelihood Activity Duration Activity Duration Probability/Likelihood Activity Duration Probability/Likelihood Activity Duration
Page 78 of 80
HP Security Analytics
26 -- Incident Category Established? Probability/Likelihood
Final Report
80% Probability of Major Security Incident: 10%
27
Probability/Likelihood
SOC Team Final Assessment 28 29 30 31 32 33 SOC Team Final Analysis SOC Team Handling Recommendations Recommended level Category False Positive? Security Incident? -- Raise Security Incident? Activity Duration Activity Duration Activity Duration Probability/Likelihood Probability/Likelihood Probability/Likelihood Probability of Major Security Incident: 10% between 4 hours and 2 days between 1 and 5 days between 2 and 4 hours 20% 80% 95%
Event Closure 34 35 Event Closure Report to Risk Management Team Activity Duration Activity Duration between 1 and 2 hours between 1 hour and 1 day
Security Incident see various probabilities for Major Security Incidents, above between 1 and 2 hours between 0.5 and 1 hour between 2 and 5 days between 0.5 and 1 hour between 0.5 and 1 hour between 1 and 5 hours
36 37 38 39 40 41 42
Major Security Incident? Raising Major Security Incident Recording of Major Incident Produce Incident Report Raising Minor Security Incident Recording of Minor Incident Produce Minor Incident Report
Probability/Likelihood Activity Duration Activity Duration Activity Duration Activity Duration Activity Duration Activity Duration
TP Non Compliance 43 44 Raise TP non Compliance Reporting to Risk Management TEam Activity Duration Activity Duration between 1 and 2 hours between 1 and 5 hours
Handling False Positives and Known Problems 45 46 Known Security Problem with Acceptable Risk? Alert Closure Probability/Likelihood Activity Duration between 0.5 and 1 hours 20%
Page 79 of 80
HP Security Analytics
Final Report
A.2
Parameter Name
Major Incident Notification
Frequency
2 3 4
Exchange of available information between SOC Team and Risk Management Team Analysis of the Incident Write Summary Report
between 1 hour and 1 day between 1 hour and 1 day between 1 and 4 hours
Information Gathering from Risk Management Team Additional Information Required from Risk Management Team? -- Request Additional information -- Wait for Information -- Additional Joint Analysis Agreement on the need for Remediation Requirements
5 6 7 8 9
Probability/Likelihood Activity Duration Activity Duration Activity Duration Activity Duration between 1 and 4 hours between 1 and 8 hours between 1 hour and 1 day
70%
TP Handling of Incident Response 10 11 12 13 14 15 Providing Recommendations to TP TP Production of Remediation Documents SOC Team: Analysis of Outcomes and Mitigations Satisfactory Response from TP? - Make additional requests to TP Raise Risk? Activity Duration Activity Duration Activity Duration Probability/Likelihood Activity Duration Probability/Likelihood between 1 and 4 hours 10% between 1 hour and 2 days between 5 days and 40 days between 4 hours and 2 days 45% NOTE: about 1-3 TP iterations
TP Incident Remediation 16 17 Incident Report TP Remediation Activity Duration Activity Duration between 2 days and 3 days between 5 days and 40 days
Risk Management 18 Incident Report Activity Duration between 2 days and 5 days
Page 80 of 80
Copyright 2012 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice. The only warranties for HP products and services are set forth in the express warranty statements accompanying such products and services. Nothing herein should be construed as constituting an additional warranty. HP shall not be liable for technical or editorial errors or omissions contained herein.