Forensics Recommended-Practice PDF
Forensics Recommended-Practice PDF
August 2008
Recommended Practice:
Creating Cyber Forensics Plans for Control Systems
August 2008
iii
EXECUTIVE SUMMARY
Cyber forensics has been in the popular mainstream for some time, and has
matured into an information-technology capability that is common among
modern information security programs. Although scalable to many information
technology domains, especially modern corporate architectures, developing a
cyber forensics program can be a challenging task when being applied to non-
traditional environments, such as control systems. Modern IT networks, through
data exchange mechanisms, data storage devices, and general computing
components provide a good foundation for creating a landscape used to support
effective cyber forensics. However, modern control systems environments are
not easily configurable to accommodate forensics programs. Nonstandard
protocols, legacy architectures that can be several decades old, and irregular or
extinct proprietary technologies can all combine to make the creation and
operation of a cyber forensics program anything but a smooth and easy process.
This document takes the traditional concepts of cyber forensics and provides
direction regarding augmentation for control systems operational environments.
The goal is to provide guidance to the reader with specifics relating to the
complexity of cyber forensics for control systems, guidance to allow
organizations to create a self-sustaining cyber forensics program for their control
systems environments, and guidance to support the maintenance and evolution of
such programs.
This document is organized into three major sections:
Section 1, Traditional Forensics and Challenges to Control Systems
Section 2, Creating a Cyber Forensics Program for Control Systems
Environments
Section 3, Activating and Sustaining a Cyber Forensics Program.
The document addresses the issues encountered in developing and
maintaining a cyber forensics plan for control systems environments. This
recommended practice supports forensic practitioners in creating a control
systems forensics plan, and assumes evidentiary data collection and preservation
using forensic best practices. The goal of this recommended practice is not to re-
invent proven methods, but to leverage them in the best possible way. As such,
the material in this recommended practice provides users with the appropriate
foundation to allow these best practices to be effective in a control systems
domain.
iv
ACKNOWLEDGEMENT
This document was developed for the U.S. Department of Homeland
Security (DHS) to provide guidance for creating a cyber forensics program for a
control systems environment. The author team consisted of subject matter
expertise from Lofty Perch, Inc. (Mark Fabro) and Idaho National Laboratory
(Eric Cornelius).
For additional information or comments, please send inquires to the DHS
Control Systems Security Program at cssp@dhs.gov.
v
CONTENTS
ABSTRACT.................................................................................................................................................iii
ACKNOWLEDGEMENT ............................................................................................................................ v
ACRONYMS.............................................................................................................................................viii
KEYWORDS................................................................................................................................................ 1
INTRODUCTION ........................................................................................................................................ 1
BACKGROUND .......................................................................................................................................... 5
4. REFERENCES ................................................................................................................................. 42
vi
FIGURES
Figure 1. Control systems forensics domain and CSSP reference architecture. ........................................... 6
Figure 2. Forensic plan components. ............................................................................................................ 7
Figure 3. Non-persistence and uniqueness in data. ..................................................................................... 11
Figure 4. Example Modern/Common components. .................................................................................... 17
Figure 5. Example Modern/Proprietary components. ................................................................................. 19
Figure 6. Example Legacy/Proprietary components................................................................................... 22
TABLES
Table 1. Modern/Common technologies and forensics compatibility. ....................................................... 18
Table 2. Modern/Proprietary technologies and forensics compatibility. .................................................... 20
Table 3. Legacy/Proprietary technologies and forensics compatibility. ..................................................... 23
Table 4. Sample of possible artifacts and relevant forensic information. ................................................... 34
Table 5. Roles matrix for incident response and forensics in control systems. .......................................... 40
vii
ACRONYMS
ANSI American National Standards Institute
CD compact disc
CPU central processing unit
CS control system
2
CS SAT Control Systems Cyber Security Self Assessment Tool
CSIM Control Systems Incident Manager
CSSP Control Systems Security Program
CSSS Control Systems Security Specialist
DCS Distributed Control Systems
DHS Department of Homeland Security
DVD Digital Video Disc
DVD-ROM DVD Read Only Memory
EWS Engineering Work Station
FW Firewall
GPS global positioning system
HMI human-machine interface
I/O input/output
IDS Intrusion Detection System
IED Intelligent Electronic Device
IPS Intrusion Prevention System
IR incident response
ISA Instrumentation, Systems, and Automation Society (f. Instrument Society of America)
IT information technology
NIST National Institute of Standards and Technology
NTP Network Time Protocol
OEM original equipment manufacturer
OS operating system
PCS Process Control Systems
PLC Programmable Logic Controller
RTU Remote Terminal Unit
RW rewritable
SCADA Supervisory Control and Data Acquisition
SLA Service/Security Level Agreement
SNMP Simple Network Message Protocol
TCB trusted computing base
viii
Recommended Practice
Improving Cyber Forensics Plans for Control Systems
KEYWORDS
Control Systems, Forensics, Event Correlation, System Recovery, Incident Logging, Incident
Auditing, Cyber Security, Control Systems Networks, Incident Response
INTRODUCTION
In the control systems domain, where the overarching security principles of confidentiality, integrity,
and availability often default to only availability and integrity (usually in that order of importance),
technologies associated with system resiliency often displace security-specific activities. Currently,
contemporary cyber security systems often need extensive aftermarket calibration to be truly effective
inside control systems domains. Additionally, the overhead associated with fine-tuning these complex
devices (in the presence of unique and proprietary solutions) often contributes to their absence. Like
security technologies, cyber security activities and capabilities also need to be fine-tuned to accommodate
for the uniqueness and nuances associated with control systems.
As a core component to incident response capabilities, cyber forensics provides for the collection,
examination, analysis, and reporting of incident data. In most cases, the proper collection and analysis of
incident data supports investigations, uncovers illegal activities, and develops better-defined security
countermeasures. Through the implementation of data exchange mechanisms, data storage devices, and
sophisticated general computing components, modern networks provide a good foundation for creating a
landscape used to support effective cyber forensics. Modern control systems environments, on the other
hand, are not as easily configured to accommodate forensics programs. Nonstandard protocols and legacy
architectures, which can be several decades old, combined with irregular or extinct proprietary
technologies can make the creation and operation of a cyber forensics program anything but a smooth and
easy process.
Building on the common elements of standardized forensics processes, such as those related to the
collection, examination, analysis, and reporting of event data, this paper will provide the reader with a
foundation to enhance and/or create a cyber forensics program for control systems environments. Due to
the diversity and disparate nature of technologies, platforms, and owner/operator deployments, this
document will provide the necessary elements to create a flexible framework, rather than those that
support specific technologies.
This document is divided into three major sections:
• Section 1, Traditional Forensics and Challenges to Control Systems
• Section 2, Creating a Cyber Forensics Program for Control Systems Environments
• Section 3, Activating and Sustaining a Cyber Forensics Program.
Section 1 addresses traditional forensics components and emphasizes the critical elements in
developing a cyber forensics capability for control systems environments. Additionally, Section 1
provides the reader with insight into the complexities and difficulties associated with building such a
capability. These core components are then mapped to specific challenges in control systems to provide
the reader with insight allowing for the development of a solution that can be customized to their specific
environment.
1
Section 2 addresses general components of the cyber forensic program and the elements that need
developing to ensure a viable and robust plan is usable by managers and users alike. In contrast to
traditional cyber forensics plans, this section also includes requirements and suggestions related to control
systems personnel, control systems operations, and business operations. This ensures the requirements
associated with the forensics program are applicable to the particular control systems environment. This
section introduces the reader to a cyber forensic framework that is tunable to any control systems
environment while still maintaining the fundamentals of an effective cyber forensic program.
Section 3 addresses activation of the cyber forensics program for control systems, and provides
insights that allow organizations to sustain their program and ensure its applicability to the architecture
for which it was developed. In addition, this section highlights techniques for integrating the forensics
capability into incident response, as well as ensuring the forensics plan is part of the decision process as it
relates to making changes in the overall information architecture.
2
AUDIENCE AND SCOPE
This document is for managers and security professionals responsible for developing, deploying, and
improving the cyber security posture of control systems domains. Although designed to be flexible
enough for system operators and engineers to read and use, this document’s intended use is by those that
are deploying cyber security incident response and/or forensics programs within control systems
environments. It is not designed to replace a sector-specific approach to creating a cyber forensics
program, but rather provide guidance as it relates to control systems specific issues. It may be found most
appropriate by either those that have experience in deploying cyber forensics programs in modern
information technology (IT) domains and who are beginning to address the issues related to deploying
cyber forensic strategies for control systems architectures, or by control systems network professionals
requiring guidance on creating a cyber forensics capability for their systems. The scope of the material is
not technically demanding and can help provide a foundation for creating or augmenting existing
information resource protection and recovery initiatives.
Although it may be required that a forensic examination of an incident on a control systems device is
intended to ascertain if criminal activity has occurred, the methodologies used for the correct collection of
digital data (although fairly ubiquitous across most domains) require modifications to accommodate for
control systems uniqueness. As such, the scope of this recommended practice will be limited to those
concepts that can impede the smooth preparation and development of a cyber forensics program for
control systems. Additionally, this document provides insight as to what the Owner/Operator community
can do to prepare proactively for a forensics investigation in support of an incident response capability.
Although this document discusses content regarding collection, handling, and analysis of control systems
data, material as it relates to the detailed methodologies involved in collection, handling, and reporting of
evidentiary data will be excluded from this document, and it is assumed that evidentiary data will be
collected and preserved using forensic best practices. The goal of this recommended practice is not to re-
invent proven methods, but to leverage them in the best possible way. As such, the material in this
document will provide users with the appropriate foundation to allow for implementing its
recommendations effectively in a control systems domain.
In the interest of brevity, and assuming the general readership will have experience with some IT
security aspects, any standards for cyber security that are discussed are presented at a non-technical level
(i.e., high level). This document does not try to provide justification as to why an organization would
want to develop a forensics program, but rather it provides guidance to those that wish to augment a
proven IT forensic process for their control systems domain. However, the authors recognize that within
several critical infrastructure sectors, such as electricity (generation and transmission), cyber incident
response and forensics plans are required.a Furthermore, the authors stress that the citation of any
particular document is not to sway the reader’s opinion in favor of the practices of any one sector, but to
demonstrate the activities of that sector regarding cyber forensics.
The current landscape of terminologies associated with critical infrastructure and control systems
operations can lead to unforeseen complexity. In this document, the term “control systems” will refer to
technologies that are often defined as Supervisory Control and Data Acquisition (SCADA), Process
Control Systems (PCS), Distributed Control Systems (DCS), or any combination thereof. The authors are
not intending to discredit or avoid recognizing the often-clear differentiators in these environments. It is
intended that the nomenclature used with the recommendations herein is broad enough to be applied in
any of these domains.
a. For example NERC CIP cyber security guidelines required by the FERC Order 706 (see References)
3
To highlight some of the more important points in cyber forensics for control systems, these icons are
occasionally used.
This symbol is used to identify cautionary points that should be considered carefully
when developing a cyber forensics program for control systems.
4
BACKGROUND
While the field of cyber forensics has been in the popular mainstream for some time, cyber forensics
as applied to control systems is immature. With increasing computational capability and interoperability
coming to previously isolated networks, the necessity for cyber forensics in control systems is becoming
quite clear. In addition, the qualitative differences between corporate networks and control systems
networks often highlight why security countermeasures are not easily deployable in control systems.
Control systems have considerable requirements related to (in order of priority) availability, data
integrity, and confidentiality. As opposed to corporate domains, where the priorities reverse, cyber
security activities in control systems networks require accommodating for systems that cannot readily be
taken offline, may not be quickly modernized, and may not be able to facilitate adequate logging and
audit functions. These elements are, of course, vital to a forensic program’s success. Although all
shortcomings that may be present in control systems (from a cyber security perspective) may be
overcome by budget allocations and spending, many organizations find the cost of incorporating cyber
security functionality into control systems technology quite high. Thus, being able to create effective
security programs (such as forensics) for control systems involves reusing existing methodologies and
practices, such as those designed for corporate domains. The requirement then becomes understanding
what changes or augmentations to these proven capabilities are required to allow applicability to control
systems networks.
The term “forensics” often relates to “the post-incident collection and analysis of data obtained from
devices.” Due to the unique and uncommon nature of control systems technologies, there is often
inadequate information collected from these countermeasures following a cyber attack or incident. In
some cases, where the owner/operators are cognizant of these shortcomings in commercial security
products, tailored signatures and detection capabilities specific to control systems operations are created
and appended to the security devices. However, because the nature of control systems operations can
require deterministic and real-time data exchanges, it is often the case that these enhancements either
prove useless or inhibit the actual productivity of the systems. Additionally, these enhancements are often
deployed as a defensive activity only and are not extended to support any forensic function in the
organization.
Figure 1 is a basic control systems security reference architectureb that illustrates the concept of the
control systems forensics domain in relation to traditional IT domains. This reference architecture simply
provides a basic framework for illustrative purposes, and is not representative of any architecture other
than a notional one.
5
Figure 1. Control systems forensics domain and CSSP reference architecture.c
There are additional challenges with forensics analysis for control systems. The field devices that are
used within control systems architectures, perhaps the terminus of a cyber incident resulting in physical
consequences, often have no inherent capability for detailed logging. Furthermore, it has been found that
on devices where extensive logging is supported the feature is often disabled, or the devices lack
sufficient capacity to store enough data to allow analysts to meet forensics requirements.d Lastly, the
diversity of the control systems technologies in use today also poses significant problems.
Owner/operator staff often does not have the skill sets to collect, examine, or analyze command and
control traffic. Instead, owners of the control systems (and devices) rely upon vendor/integrator staff for
support. This in itself can precipitate delays in incident analysis and resolution, as understanding of
detailed device operations and logging capabilities are often left until it is too late or completed “after-the-
fact.” Although systems can be configured to alarm operators in a timely fashion, the interpretation of the
data and the correlation to an incident remain dependent upon the end user’s technical skill level. Indeed,
countermeasures can be instituted for any and all of these issues, but the cost associated with making such
changes to the controls systems operational domain is generally too high in both time and testing, and
requires significant amount of time and effort by the owner/operator.
c. The diagram is notional in nature and does not account for all possible architectures. As an example, it is not uncommon for
modems to be connected directly into applications servers as well as PLCs, especially in legacy environments.
d. Readers should be aware that in some sectors, such as energy/electric, specific audit requirements demand that event data is
stored at some point in the system. Although this data storage may not be fully viable to an incident, investigation, transaction,
and event data are available in some form within the information domain.
6
Overall, the challenges impacting effective forensics in control systems can be summarized as:
• Many traditional device and control systems technologies do not provide for the collection of
effective data that could be used for post-incident security analysis.e Those systems that do have such
capability are often in operation mode without such capability active.
• Current cyber-forensic methodologies are not always fully extensible to traditional control systems
architectures.
• For the architectures that do use modern cyber-centric security procedures and technologies
(Firewalls [FW], Intrusion Detection Systems [IDS], Intrusion Prevention Systems, [IPS], etc.), the
unification of forensic data collected by these systems cannot always be effectively correlated with
device and control systems logging data.
• Post-incident analysis is often dependent on vendor involvement, and any proactive understanding of
device logging is often not required by the end user or incorporated into a defense-in-depth strategy.
To address these issues at the appropriate level, guidance for developing a control systems forensic
program is required. This guidance must be fully flexible to allow for deployment into any control
systems environment regardless of technologies used. Moreover, the guidance must provide for direction
on the integration of modern network security technologies with traditionally closed systems, the result
being a true defense-in-depth strategy for control systems architectures. This strategy will be a
combination of technical, managerial, and operational issues that provide for a structured approach that,
when the aggregate is finished, gives a flexible but strong security posture.
This document introduces some of the
more essential cyber security activities that
are important for the development of a
control systems cyber forensic program.
Figure 2 illustrates the major components
that will contribute to helping create an
organizational cyber forensics plan for
control systems domain.
e. Requirements from NERC demand that event data is stored in some fashion, usually for the purpose of event recreation. The
data may not help in a forensics investigation, but readers should be aware that such requirements do exist.
7
1. TRADITIONAL FORENSICS AND CHALLENGES
TO CONTROL SYSTEMS
To understand some of the issues associated with creating a cyber forensics program for control
systems, it is pertinent to review the core components of standard cyber forensics programs and discuss
them in terms of the irregularities associated with control systems. Generally, the definition of cyber
forensics is:
“…the application of science to the identification, collection, examination,
and analysis of data while preserving the integrity of the information and
maintaining a strict chain of custody for the data.” f
For this recommended practice document, the forensics process considered will be comprised of the
three core areas of collection, analysis, and reporting. Although cyber forensics is a mature domain, an
organization will need to tailor traditional forensics components to control systems environments to
overcome specific challenges, as described below. However, prior to understanding how these core areas
are implemented in a control systems environment, it is important to understand the differences between
traditional IT domains and control systems architectures. Having a solid understanding of some of these
major differences will help expedite the understanding of what components are necessary in a cyber
forensics plan that will be unique to control systems architectures.
To accommodate for some of the technological uniqueness within control systems, especially those
that may prevent effective collection of critical non-persistent or volatile data, in-depth reviews of
existing data analysis capabilities (followed by the reinterpretation of that capability to fit into a forensics
schema) are required. In many cases, this may simply require the inclusion of a secondary piece of
appropriate technology into the networking environment to introduce capabilities to support collection.
As with other aspects related to architecture modifications, this technology will have to meet approved
deployment (and costing) guidelines, and should not introduce any operational risk to the control systems
domain.
Methodologies regarding cyber forensics usually include the removal of impacted information
resources from the environment. Although this is usually done following complete system backups and
integrity verification of the acquired data, the nature of control systems does not always accommodate for
such procedures. In many cases, the impacted technology is simply not replaceable due to cost or lack of
available technology. Moreover, the impacted equipment may have such an integral role to business
operations that it cannot be taken offline. For the forensics investigator, these aspects alone can severely
impede effective analysis activities. While a control systems environment may not be able to come off-
line, inherent troubleshooting capabilities for production systems maintenance should provide some
useful data.
8
technologies in place that can provide some of these minimum requirements. However, some of the core
system support/control technologies (those that would be targets for an attacker) simply cannot and do not
allow for effective data collection by a cyber forensics analyst. Because the capability for embedded data
logging and security detection is vendor provided, the owner/operator cost associated with creating the
capability is often quite high. Other collection methods, such as those associated with network and host-
based detection and logging, will introduce a cost that could be high depending on the nature, size, and
criticality of the control systems network.g
For the investigator of a cyber forensics incident to understand the nature and severity of the
occurrence, he or she will need to have access to and obtain event specific information from as many
sources as possible. Record types, such as those associated with audits, authentication, authorization, and
general system activities become critical in executing a successful forensics investigation. The practice of
merging all source data from across the enterprise is a common strategy to investigators, and fusion of
these data sources is a process that has been documented at many times and in many different ways.
Effective forensics collection in any environment requires addressing several challenges such as
volatile memory, poor administrative functions, absent or inadequate logging, and general cultural
limitations. In control systems environments, there are additional challenges including:
• Automation. The key information resources in the control systems domain will be created and
deployed to handle data in such a way that the implementation of a data retention scheme is neither
cost effective nor a requirement.
• Volatility of Data. The data within the process and state information is deleted, removed, or
overwritten at a rate that makes collection on some devices unviable or impossible.h Although after-
market solutions can be architected, they are often too cost prohibitive to implement.
• Data Mingling. The total information sample that is resident in the investigated system is comprised
of both data related to incident activity (possibly malicious) and data that is unrelated to the incident.
Moreover, due to the limited memory storage of the system (see above) these data types are often
indistinguishable. Although it is not a unique problem to control systems, this can be attributed to
inadequate labeling and is in itself a function of the control systems (vendor supplied) technology.
Research has shown that the most valuable asset to an attacker may be those that influence the control
or behavior of the system endpoints (including field devices).i Thus, it becomes important to consider the
capabilities of control systems information resources in terms of data retention, and how that data can
support an investigation. If the control technology is simple or older in nature (i.e., legacy or no longer
supported), they often have no inherent capability for detailed logging. Without after-market
modifications or costly system enhancements, these factors can make the collection of incident-specific
information very difficult if not impossible.
g. Initiatives are in place to for federal, state, local asset owners, and regulators to obtain a common control systems security
understanding using these procurement guidelines to help ensure that security is integrated into control systems.
http://www.msisac.org/scada/documents/4march08scadaprocure.pdf.
h. Within some safety systems, high-speed data recorders have been used for many years. This type of recording activity will
maintain the data that is often overwritten within system components and can be used for analysis and event recreation.
i. The Use of Attack Trees in Assessing Vulnerabilities in SCADA Systems, http://www.threatmind.net/papers/SCADA-Attack-
Trees-IISW.pdf.
9
Data volatility and non-persistence can also contribute to the complexities associated with harvesting
meaningful data from an information resource. As listed in order of decreasing volatility, the volatile
components arej:
• Registers, cache
• Routing table, arp cache, process table, kernel statistics, memory
• Temporary file systems
• Disk
• Remote logging and monitoring data that is relevant to the system in question
• Physical configuration, network topology
• Archival media.
Looking at the levels of volatility, the non-persistent data associated with both registers and caches
are at risk. Within the control systems environment, both registers and cache play a significant role as it
pertains to both network and field devices alike, but the rate at which information is pushed to and taken
from these devices does not always permit for useful incident-specific information to be collected.
However, it is important to know that as the non-persistence of data that could be used by an investigator
increases, (as does the reduction of possible data mingling) the overall uniqueness of the data decreases
(see Figure 3).k
Non-persistent data may become persistent in some form. Some data that can be
considered volatile due to its lifetime on a device (i.e., overwritten quickly) may become
persistent in some other data store in the system. Some regulatory requirements, such as
those in the energy sector, mandate the recording of all transactions and instructions
leading to some device activity for maintaining, but not at the device itself.
j. RFC 3227, http://www.faqs.org/rfcs/rfc3227.html Example order of decreasing volatility for a typical system.
k. Non-persistence and volatility are important in forensics, as they each impact how accessible incident data is and how the
investigator can access it.
10
Figure 3. Non-persistence and uniqueness in data.
The diversity of the technologies used in modern control systems environments also pose significant
challenges, as the ability to understand device or operational log data is often a vendor-only skill. This
fact can force the efficiency of any post-incident response to be proportional to the level of vendor
support, or until local investigators become appropriately learned in the technology. This in itself can
precipitate delays in incident analysis and resolution, as detailed understanding by the end user regarding
device operations and logging capabilities are often left until it is too late or are completed “after-the-
fact.” Establishing a vendor relationship and service level agreements that are capable of ensuring
effective and timely forensics support is a task that is beyond the scope of this report. However, the reader
is advised that such relationships with the vendor, and thus the creation of incident or forensics programs
that are tailored to a specific installation, are critical to creating and maintaining an effective cyber
security posture.
In some incident response plans, the standard organizational practice is to replace the impacted
components from the architecture or take the resources offline. Such activities in many control systems
environments are quite unrealistic, as taking equipment offline is either infeasible, too cost intensive, will
have an unacceptable impact on critical infrastructure, or is simply prohibited from a business operations
perspective. Because of the concerns relating to the possibility of equipment becoming unavailable may
create adverse impacts on the production system, organizations have a challenge in choosing how to
correct problems created by cyber incidents. Moreover, cost issues associated with control systems
technology equipment can prevent the swapping in and out of impacted resources, and organizations often
do not have backup devices available in quantity nor the time required to wait for the delivery of
replacement technology.
To account for these and other nuances associated with control systems, the introduction of proactive
monitoring and analysis capabilities are recommended. These can be considered countermeasure support
functions, and can include (but are not limited to):
• Inclusion of real-time forensics tools for active analysis
11
• Embedded forensics analysis tools within the critical operational environment
• Security information management and collection systems.
Without fully understanding the extent to which forensics tools can influence control
systems operations, due care must be practiced in trying to use standard forensic
methods on control systems technologies. If it is not fully understood from evaluation in
a test environment how the tools can impact the production system, then deployment in
the production environment should be avoided.
A core component of any forensics capability is to ensure that the information that has been collected
from the environment is correctly assimilated and reviewed. Being able to have only one or two data
sources often limits the investigator’s overall effectiveness at carrying out data analysis, so understanding
how the incident may have impacted numerous resources within the domain becomes critical.
Currently, it is commonplace to find control systems architectures that are devoid of any logging or
multi-source security information collection capability, as well as finding an architecture that has not
utilized any add-on or aftermarket logging capability for the core control systems information resources.
Within control systems architectures where firewalls and intrusion detection or other security
technologies have been deployed, centralized data collection from the security resources is often not a
l. This is assumed for more modern environments. There is no question that there exists a significant number of proprietary
systems that do not use Windows or UNIX operating systems as a base operating system.
12
part of the overarching cyber security strategy. This is not to say that administrators and architects are at
fault, but rather to emphasize the fact that in order to create an effective cyber forensics capability for
control systems there are some proactive strategies involving logging and data collection that need to be
developed.m
Defense-in-depth strategies can greatly improve the cyber security posture (and
forensic applicability) of a control systems environment while supporting the need for
business operations.
The forensics investigator will need to draw on all types of records as well as the interaction of the
record holder with other core devices in the network. However, current information technologies that
have the capability to maintain these types of audit records may not necessarily be in a data exchange
relationship with the devices in the control systems domain that experience a cyber incident. As such, this
can limit the scope to which the forensics investigator can extend his or her investigation.
The complexity of many control systems environments, along with the uniqueness that is often
associated with proprietary installations, drives the requirement for vendor interaction before, during, and
following a cyber security incident. Historically, installations that have experienced a cyber incident will
often opt to communicate directly with the vendor in an attempt to get support, insight, or some direct
interaction that will allow for a deeper understanding concerning the incident. Moreover, because many of
the control systems environments around the world are involved in critical infrastructure systems, the first
priority during and following any incident is usually the resumption of services. In many cases, this can
lead to the replacement of key devices or the overwriting of operational data, some of which could have
been critical in analyzing the root cause of the incident.
Documentation is paramount to the success of forensics investigations in the control systems
environment. Asset owners should take preemptive steps to identify and document any changes made to
operating systems, hardware configurations, device drivers, or any other elements whose modified
behavior may differ from its original equipment manufacturer’s (OEM’s) defaults. Furthermore, it is
recommended that asset owners communicate with their vendors in order to identify and document
similar changes made by the vendor. This information should be provided to the forensics investigator
prior to any forensics activity. The investigator shall note the modifications and account for them
accordingly. The investigator shall thoroughly document and explain his/her actions with respect to
modified components and maintain compliance with proven forensics best practices.
13
2. CREATING A CYBER FORENSICS PROGRAM
FOR CONTROL SYSTEMS ENVIRONMENTS
Within any networked environment, cyber forensics should be a capability that supports the
protection of key information assets, while facilitating the collection of evidence assisting in the
understanding of cause, mitigation strategies, resiliency programs, or even law enforcement activities. By
definition, a cyber forensics program is designed to support a cyber incident response program, and it is
the components within the incident response program that should be appropriately integrated into the
fabric of the forensics capability. Before, during, and following a cyber incident, and after returning the
control system to its operational state (if possible), it is the cyber forensics process that intends to
discover what happened, how it happened, who did it, and what can prevent it from happening in the
future. In the event that any activity related to the incident was illegal, the cyber forensics program will
possibly provide evidence that is admissible in court.
As discussed above, the issues that can impede a successful forensic investigation include vendor
technology, volatility of key data, and lack of extensive system-specific understanding. Although these
factors are not unique to control systems per se, they combine to make the process of creating a control
systems forensics program complicated at best. An effective cyber forensics program for control systems
environments is contingent on a number of different variables. From a macroscopic level, a forensics
policy that indicates process and procedures for forensics analysis on control systems components is
paramount. In addition to this policy, specific information about existing services within the operational
environment is also required.
With effective preparation mandatory for any successful response program, understanding the
environment is critical. A clear understanding of the architecture in question, combined with clear
documentation highlighting unique modifications, will empower the investigator, and most likely
expedite analysis. For control systems environments, the development of a concise operational picture
should be a straightforward task provided the critical information regarding device applications, any
control systems specific security technologies, primary operator interfaces, and device control logs are
accessible.
A cyber forensics program for control systems should be flexible enough to accommodate for the
appropriate response based upon the significance or extent of a given incident. To accommodate for this
requirement, the cyber forensics plan should include (but not be limited to) these phasesn:
1. Identifying system environment and uniqueness
2. Defining environment specific requirements
3. Creating capabilities for identifying, collecting, and preserving data evidence and artifacts
n. Other phases, such as creating capabilities for examining data and reporting results can be closely mapped to standard forensic
practices, with special emphasis paid to ensuring subject matter expertise (control systems cyber security) are available to provide
support. See Section 3.1 for discussion on forensics activity supporting roles.
14
align with business operations, and appropriately pass information to law enforcement agencies, if
required.o
An organization looking to deploy a forensics capability for their control systems environment needs
to be able to understand fully a wide range of consequences that could be associated with a cyber event.
By using risk models that are functions of threat, vulnerability, and consequence, an organization can
have a much more granular understanding of how a cyber incident can impact certain components and
control systems operations.p The consequences associated with cyber incidents in a control systems
environment can vary and can include:
• Loss of localized or remote control over the process
• Loss of production
• Compromise of safety
• Catastrophic cascading failures that affect critical infrastructure and can extend to peer sites and other
critical infrastructure sectors
• Environmental damage
• Injury or loss of human life
To help better categorize the consequences, an initial phase in the development of a cyber forensics
plan for control systems should be a review of both the theoretical and practical consequences of incidents
on the system. For most medium-to-large scale enterprises (and even some smaller ones), scenario-based
reviews often provide insight into the associated criticality of the information resources and devices in the
control systems domain. Using any traditional model for forensics, organizations can quickly come to
understand what information resources can support forensics methodologies and those that cannot.
Understanding consequence is usually within the province of an organization’s risk management
team, insurance underwriters, or some combination of these and other corporate resources. Recently, as a
product of the Department of Homeland Security (DHS) Control Systems Security Program (CSSP), a
specialized toolkit designed to help an organization understand their consequence associated with their
control systems was created. The Control Systems Cyber Security Self Assessment Toolq (CS2SAT)
provides owners and operators a non-intrusive mechanism to understand operational risk better as it
relates to the cyber security of all components within control systems architectures. The tool, which also
allows users to learn more information about various cyber security standards associated with control
systems, provides for the calculation of security assurance levels that can be tied to the most common
consequences in which organizations are familiar. This includes physical damage, human injury and loss
of life, environmental impact, and other qualitative or quantitative aspects related to an incident that
impacts control systems operations. In addition, the tool provides the users with the functionality to create
effective reproductions of the control systems architecture to understand the relationships between the
control systems domain, the corporate domain, peer domains, and the security relationships in amongst
core operational devices. Using tools such as this, along with other viable consequence analysis
methodologies, can provide an organization with a much-needed proactive review of how a cyber incident
can manifest in a control systems domain. The information learned in this process can help organizations
have a better cognitive view of the extensiveness of incidents, as well as understand what relationships
o. Factors that can impact system environment and overall uniqueness can be very extensive, and can include mergers,
acquisitions, corporate security policies, and regulatory/compliance issues.
p. Department of Homeland Security, “Risk=ThreatxVulnerabilityxConsequence from the Nation Infrastructure Protection Plan,”
http://www.dhs.gov/xprevprot/programs/editorial_0827.shtm.
q. CS2SAT, http://www.us-cert.gov/control_systems/pdf/CS2SAT_Trifold_Web_version_Final_062308.pdf.
15
among key devices need to be examined with regards to formulating a response associated with a cyber
forensics investigation.
It is important for organizations to understand the extensiveness of the connections in amongst their
peers’ sites and partners. By looking to understand the uniqueness associated with the system,
organizations developing the components for a tactical forensics plan must also take into consideration
other variables such as:
• Connections to and from partner or peer locations
• Access to mechanisms used by the vendor to support control systems technology
• Access to mechanisms used by contractors to support control systems environments
• Actual relationships between cyber resources and real-world operations
• The pervasiveness of effective cyber security policies governing control systems operations
• Information exchange mechanisms within the supply chain
It is particularly interesting to note that this last point related to supply chain
operations is often neglected or misunderstood. Supply chain operation networks,
unfortunately, often include direct connections into the primary users of the supplied
materials, but due to the trusted nature of the connections, the incident pathway can be
overlooked.
By assessing the uniqueness of the control systems environment and the components within that
environment, organizations can categorize their technologies and create a cohesive plan that contains
flexibility for network elements that may lack logging or audit capabilities. To offset any inadequacies
regarding logging and audit capability, this method also supports the granular understanding and
calculation of the appropriate levels of detail that each information resource in the network is generating.
In lieu of devices that are unable to store information adequately, specifics relevant to information
generation and the role of the device in the network can be very helpful.
In the interest of simplicity, it may serve the organization very well to categorize their technologies
into one of three areas: Modern/Common, Modern/Proprietary, and Legacy/Proprietary.
The technological differences and characteristics between devices in these three categories can be
quite extensive, thus the difference in approaches to each by a forensic examiner will need to be adjusted
accordingly. The concepts introduced in the following sections are provided to successfully augment
proven methods, and are not intended to recreate any existing forensics investigation methodologies. The
concept of categorizing technologies can greatly expedite investigators arriving at conclusions in both the
collection and analysis phases. The range of currently deployed control systems technologies encompass
those that are very new and deemed forensics friendly, to those that are very old and rely almost entirely
on knowledge held by a small number of people. The next section explores the differences in these
technologies, as well as forensics responses to each that are likely to be successful.
16
or other aftermarket modifications that allows the technology to remain current as it relates to compliance
with standard practices (i.e., OSI interoperability). Figure 4 illustrates some of the technologies in the
control systems environment most likely to be of the Modern/Common type.
This category of technologies inside the control systems domain will be those that would be most
susceptible to modern cyber threats and vulnerabilities, while at the same time being mature enough to
allow some contemporary forensic methods to be successfully performed on them. Most common
technologies that fall into this category include Microsoft Windows operating systems, those systems
using the UNIX platform, or another vendor specific solution that has functionality that can be
investigated using standard forensics methodologies.
Table 1 below highlights several of the major features associated with these types of technologies
relevant to creating a cyber forensics program.
17
Table 1. Modern/Common technologies and forensics compatibility.r s
Section 2.3 discusses specifics relating to identification and collection of evidence from the system.
Within this category of Modern/Common technologies, technologies are found that meet the criteria
to be categorized as Modern/Common, but they do not have any inherent data collection capabilities (such
as local logging or audit) that could be leveraged by standard forensics methodologies. Simple examples
of this could be the Programmable Logic Controller (PLC), Remote Terminal Unit (RTU), Intelligent
Electronic Device (IED), or some other field-level device that is empowered to communicate through
modern communications mechanisms. These devices often have available services embedded in them to
enhance administration, some of which include Web services, Simple Network Message Protocol
(SNMP), and diagnostic tools. These devices, categorized as modern and common, often lack the
capability to provide an investigator with any data that could be useful following an incident.
r. Some functions, such as those supporting safety operations, have extensive event logging functions built in. As this paper
focuses on cyber incidents, this table relates to the ability of system components to provide effective forensic data that can be
utilized during standard forensic investigations.
s. In this and following tables, the term “Forensics Compliant” refers to whether or not the technology is easily investigated by
contemporary forensics technologies. “Reference Materials Available” refers to the availability of open source information
regarding methods associated at diagnosing failures and incidents.
18
most appropriate response strategy, given the circumstances of the incident. The strategy should take into
consideration both technical and business factors, and should be approved by management.
In summary, the deployment of a flexible forensics framework needs to consider both live and dead
(powered on versus powered off) systems analysis. On devices categorized as both Modern and Common,
contemporary forensics techniques are most likely to yield favorable results. Wherever possible, the
forensic examiner should employ live systems analysis to avoid disrupting the operational process, or
where critical volatile information risks of being lost. As a corollary, the examiner should employ
traditional dead systems analysis whenever a backup is easily available in order to obtain a greater
amount of information.
19
The reader should note that there is some overlap between the Modern/Common and
Modern/Proprietary classifications. This is because there may (and probably will) be proprietary
components residing within a Common technology (Windows, UNIX), which modify its behavior or
otherwise change its default operating system (OS) functionality.
When categorizing the control system technology to be both modern but proprietary, the complexity
related to executing a successful forensics investigation is increased. In an ideal world, most of the
contemporary deployments for control systems would be of the previous Modern/Common type, thus
allowing for effective incident handling and forensics techniques to be applied. In the previous category,
it introduced complexity when the modern field device technology or operating systems did not have any
inherent capability to support logging or audits. This of course will force the investigator to initiate
secondary or tertiary activities to collect as much information as possible from a live system (see Section
2.3 on collection). Table 2 illustrates some of the more important aspects associated with
Modern/Proprietary systems.
Table 2. Modern/Proprietary technologies and forensics compatibility.
When the modern control system contains a high level of unique and proprietary technology, the
investigator or team of investigators will require significant in-depth information concerning the vendor
solution. In such circumstances, forensics investigation (not to mention incident response) can be
impeded. Although the investigator may be able to use contemporary imaging and collection methods on
some of the supporting technology and the environment, without a clear and concise understanding of
how the proprietary technology is working in the operational environment, the investigator risks the
possibility of misinterpreting the data. It could be assumed that, over time, the investigator may come to
learn the internals of the proprietary technology, but it makes sense that the time required to do so would
push the timelines associated with the investigation into the realm of the impractical.t When the data has
been stored and is available for event recreation, the investigator does not need to know any internals.
However, for a cyber incident, the data retained for event recreation may be insufficient for forensic
purposes.
t. There may also be evidentiary admissibility issues in such discovery without appropriate vendor support.
20
When categorizing the devices in question as both Modern/Proprietary, the forensic investigator is
encouraged to perform contemporary dead (offline) analysis wherever possible. Performing a dead
analysis will allow the investigator to try many different techniques on an image of the system without
risking a compromise of the data’s integrity. If the system in question cannot be taken offline, the
investigator is encouraged to perform a contemporary live analysis. However, the investigator is
cautioned to be fully aware of the potential ramifications these techniques can have on production
systems. Unless the investigator has adequate knowledge of the proprietary technology in question, it is
highly recommended to only perform a passive analysis. Furthermore, interaction with the vendor is
encouraged when examining proprietary technologies.
In summary, as the organization develops a forensics methodology using consequence-based analysis,
the Modern/Proprietary type of environment is at risk due to possible excessively slow responses in
correcting the impact on network components. This type of environment is also plagued with the
problems associated with restoration, the replacement of technology, and most likely a concurrent need to
have continuous operations. Issues associated with collecting evidence from these types of environments
are discussed in detail in Section 2.3.
21
Figure 6. Example Legacy/Proprietary components.u
The assumption that the vendor is available to support these types of technology environments could
often be false. In many instances, where the systems are beyond 20 years old, not only are the technology
and the command-and-control protocols no longer supported, but the physical elements, such as field
devices, are no longer being manufactured or are no longer available on the open market. Thus,
developing a cyber forensics program that will assist an organization to understand what information can
be harvested from these devices becomes exceptionally difficult.
In these circumstances, the expertise required to support cyber forensics investigation will probably
be within the province of one or two veteran engineers or systems administrators who have a very rich
experience and history with the system in question. Undoubtedly, investigators operating in these types of
architectures should not have high expectations for being able to collect detailed incident information or
artifacts. As a system’s age approaches 20 years or beyond, historical analysis has shown that
undoubtedly these aged technologies do not have the capability to support an extensive forensics
investigation. In many cases, all that can be done is to draw on the network-based communications (if
indeed a network is involved) and try and extrapolate specific information as to how the incident
occurred. In addition, using process system reports, trending graphs, and snapshots of system activity
(event logs) over time can provide valuable data points. Table 3 illustrates some of the more important
aspects associated with Legacy/Proprietary systems.
u. The diagram is notional in nature and does not account for all possible architectures. As an example, it is not uncommon for
modems to be connected directly into applications servers as well as PLCs, especially in legacy environments.
22
Solutions that utilize third-party log monitoring can in many cases help with log data
analysis. In some cases, the usage of these tools can help with investigative analysis.
23
be tuned. In doing so, several key considerations can help ensure the forensics plan is adept at supporting
both the IR plan and the needs of the business.
24
auditing, monitoring and access control in critical systems may not be as simple as using the base
operating systems security functionality.
For those entities that require authentication or authorization for each modification that is to be made
in the control system domain, the concept of “change point verification” may very well introduce the need
for the instantiation of unique audit trails. This information can include such detailed information as
reason codes, overrides, decision approval codes or other information that could be closely tied to an
incident source or incident occurrence. In some solutions, the security extends from the localized user out
to nodes and points-of-presence in the control system domain.
Clearly, this information is extremely pertinent to the investigator during evidence collection or
incident support, and could reside on any of the primary resources found in the control systems
environment (human-machine interface [HMI], engineering workstation [EWS], Historian, etc.). Of
course, the more proprietary in nature the solution is the more reliance the investigator will have on input
from the developers and the vendor of the solution.
During a forensics investigation, the investigator may find the largest concentration
of modifications to the OS will be in the HMI.
25
2.3 Identification and Collection of Data
Assuming the investigator has been able to classify the type of environment to be evaluated, the
identification and collection of evidence from key information sources can be started. Accordingly, the
forensics investigator should be able to ascertain which components of the architecture can be assessed
using contemporary forensic technologies versus those that may not. The forensics guidelines and best
practices as they apply to contemporary technologies (i.e., where they have been proven effective before)
are beyond the scope of this report, so this section will focus on the identification and collection of
pertinent forensics evidence within a control systems environment that will need special attention and
methods.
The basic framework for any investigation, as it pertains to the identification and collection of digital
evidence (whether it is in the control systems environment or not) will have several core components or
elements that must be adhered to by any investigator. To ensure the investigator has a concise and
effective framework for executing a forensics program in a control systems environment, the following
traditional forensics elements will be examined and the uniqueness of a control systems environment and
the impacts on these elements will be discussed. These elements are:
• Reference clock system
• Activity logs and transaction logs
• Other sources of data
• General system failures
• Real time forensics
• Device integrity monitoring
• Enhanced all-source logging and auditing
26
elements in the control systems domain.x It is advised that the investigator, prior to commencing
investigation, works closely with control systems network administrators and engineers to confirm the
centralized clocking mechanism is trustworthy, and determines the extent to which it influences timing on
the other network components. In addition, to compensate for the possibility of there being multiple
centralized clock mechanisms for each of the control systems (and IT functions within a control systems
domain), the forensics investigator is strongly advised to ascertain if more than one clocks exist. If so, it is
imperative to determine if theses clocks are synchronized and which Network Time Protocol (NTP)
server each system is statically set to resolve to.y
Many organizations use global positioning system (GPS) clocks to ensure ubiquitous
time across the domain. However, it is not uncommon that network latency can introduce
considerable time differences in amongst devices. Network Time Protocol (NTP) can be
used to estimate and account for this latency, and some organizations incorporate both
methods. Caution is advised, because NTP can permit for network spoofing attacks,
causing severe discrepancies in timing and time stamping.
x. “Centralized time function” refers to a master time source in the system being investigated, not a centralized system in terms of
geography. Also, investigators should be aware that the timing mechanism for the control domain may itself have been impacted
by the cyber incident and thus should be deemed unreliable.
y. Creating effective clocking mechanisms for forensics investigations is beyond the scope of this report.
27
the success of a forensics investigation. Although specific vendors may have provided the operational
software on the systems, the underlying network communications capability provides not only vital
communications to the control systems domain, but possibly exploitable pathways for attackers.
This approach empowers the investigator to be able to expedite the investigation as it pertains to
whether or not it can be determined which systems have been impacted. If the impacted systems are
identified, OS logs and transaction logs directly relevant to control systems elements can be collected and
investigated using OS-centric guidelines that are widely available. If the incident response activity is
unable to define any specifics effectively as to the incident location (and as such specific platforms or
technologies) it forces the investigator to identify all possible audit logs and transaction activity logs, thus
creating a very extensive collection profile. Clearly, without an effective incident response capability for
an operator’s control systems environment, this latter scenario would probably be prevalent.
To assist the reader in understanding where evidence could be collected in the control systems
environment, as well as help them understand plausible unique relationships in the control systems
environment, several key components are considered from the reference architecture. In the interest of
brevity, it is assumed that the investigator has been able to ascertain the environment classification of the
control systems domain.
28
HMI
The criticality of the HMI in the control systems environment cannot be overstated. As the primary
point for all command and control activity within the control systems environment, the HMI will demand
special attention in a forensics investigation. Common/Modern deployments of the HMI allows the
investigators significant leeway in trying to understand what, if any, elements of the incident impacted the
production environment. Although consideration must be given for some vendor-specific software
applications that are tied uniquely to the control process, it is often the case that the underlying functions
of the core operating system will allow for forensics analysis.
However, like all elements in the domain, attention must be paid to the volatility of the different types
of information available from this resource. In lieu of any vendor interaction upon the immediate
evidence collection activity, the investigator will need to consider a number of elements that may impact
the using of standard forensic methods.
Of concern to the investigator is the possibility that the version of the HMI software may have
required initial hardening of the operating system (kernel) or use a standard “build” that removed non-
essential services and/or files from the base operating system. To that end, some of the more common
features and capabilities associated with transaction monitoring, alarm and event logging, or diagnostics
may be modified or absent all together. Although the core drives and resident data could be harvested for
offline investigative analysis, key data stores and file structures may be so different that a vital evidence
collection may be impossible. Furthermore, without in-depth understanding of how the HMI is executing
the command and control function in the environment; the investigator may be unable to locate pertinent
evidence that is in the HMI data stores.
Investigators are reminded that if the HMI has extensive customized applications that
augment the core OS kernel, there is an elevated risk of data mingling that will cause the
investigator problems in trying to isolate key incident-specific data.
29
Whether or not a field device is on or off, or available for either online or offline investigation, until
the extent of the incident is fully understood, the investigator may have no insight as to what the best
method might be. When it comes to field devices, many elements give rise to added complexity in the
investigation. If a field device is available for offline investigation, it is unknown (prior to analysis)
whether such a situation is an advantage or a disadvantage to the forensics investigator. In addition, as
may most often be the case, the field technology will not necessarily be made available due to the
requirement that must remain operational to supporting real-time operations. However, if the system is
available for analysis, the investigator may have some opportunities, depending on the classification of
the device. All of these factors can contribute to making each investigation, unique. However, some
common practices can be used in all cases.
If the impacted field devices belong to this category, chances are there is some useable capability
within the field device that supports forensics investigation. Investigators need to understand that
historically the deployment of these field devices is often very specific as it pertains to business
operations and the requirements of the control systems. The specific configuration of these devices should
be readily available to any investigator assuming that the configuration files have been kept, maintained,
and stored in a relatively secure environment. From the previous section, not only should the
configurations and any logic operations associated with the devices be readily available, any proactive
security actions such as hashing, checksums, or integrity verification should be used to allow the
investigator to cross-reference findings and ascertain if any tampering has been done.z
If the control system is online, the investigator is reminded of both the order of volatility of the data,
as presented in Section 1.1.aa If the incident response team has been able to verify that the incident is
perhaps still ongoing or the control system is back to a normal operating state, the forensics investigator
has the opportunity to recognize other useful files and information in the field device environment. This
often becomes a matter of urgency because this volatile system information, which includes running
processes, current connection states, and memory content, can often be deleted, overwritten, or destroyed
instantly. Due to this concern, it is recommended that wherever possible that the incident response team
collect volatile information available from the field devices.
Significant computing horsepower can outfit some modern field devices, and a live forensics
investigation may be able to obtain critical incident information on a running system that would be lost if
the system was shut down or rebooted. This includes:
• The device date and time
• Current active processes
• Current running processes.
In addition to having access to field device configuration files for analysis, modern networked field
technologies (whether online or offline) should be able to provide information similar to their common
enterprise network counterparts that may include open ports, applications associated with open ports, and
network connections. Although this type of information may not be available on all of these types of
devices, the control systems vendor community is producing field systems with such capabilities at a
rapid rate. In the same way that an attacker may be able to leverage both the network functionality and
device capability of the elements within a control systems environment, the forensics investigator should
be able to use those same types of technologies to aid in an investigation. When embedded services are
z. This method is effective for the detection of modifications after the technology has been deployed into operation, but cannot
detect if the firmware had been tampered by the vendor (i.e., insider attack during production or external attack on vendor
production systems).
aa. The issue of non-persistence is also critical, as the longevity of the data on a particular device also impact investigation.
30
available on the devices, such as Web or network management tools, the investigator should have an
opportunity to harvest current state information relevant to those services.
Note that regardless of the classification type of the field device technology, at some level the vendor
should be involved in the investigation. Extending this ensures the Service or Security Level Agreements
(SLA) with the vendor include verbiage for support during forensic investigations. Although the vendor
does not need to be advised of the investigation specifics, or be the recipient of any acquired information,
the vendor can certainly provide direction as it relates to logging, transaction activity, embedded service
functionality, or how the field devices store, handle, and write to volatile memory space.
The forensics plan should call for a detailed investigation of the process logic
internal to the field devices. Of particular importance is the historical administrative
record that may indicate subtle changes to the logic made to accommodate changes in
the physical system. bb
When dealing with modern but proprietary control systems technologies, especially
those dedicated to the command and control function in an operational environment,
interaction with the vendor prior to the investigation is strongly recommended. Although
it may be obvious what file systems are in use, it is not uncommon for standard file
systems to be modified by the vendor to accommodate unique control capabilities. These
modifications can impact the functionality of both the forensics investigator’s tools as
well as any inherent operating system specific auditing functions, leading to poor
evidence collection or added complexity due to data mangling.
bb. The reader can review the official Taum Sauk incident report, highlighting some detailed information about the modifications
made inside the vice logic to accommodate for subtle changes in how water levels were being recorded.
www.ferc.gov/industries/hydropower/safety/projects/taum-sauk.asp
31
• Fault tolerance that automatically kills jobs
• Real-time reallocation of memory space
Such modifications can work both for and against an investigator; depending on whether or not the
system is online (alive) or offline (dead), the activity of these modifications needs to be fully understood.
A simple example would be in a situation where fault tolerance has overridden the need for a supporting
system process, which was truncated by malicious activity. The vendor’s solution to kill the process
instantaneously removes any evidence that could be used by the investigator. Although most of the
proprietary influence could very well be on the HMI, it is not unusual for vendors to push proprietary
solutions into the acquisition or application servers as well.
Modern/Proprietary Field Devices
Field devices under this category may have some inherent technology or capability that can aid a
forensics investigation, but it may be substantially less than in the Modern/Common category. As
industry moves into this category, the dependence on input from the vendor will grow. It is also important
to note that many of the proactive strategies associated with understanding the integrity of the core files in
these devices can still be used effectively.
This report is directed to those environments that have network-based communications and are open.
However, some vendor solutions for field devices are proprietary in nature. They are often deployed using
a network address schema provided by the vendor. Therefore, a good investigation will need to know the
entire architecture including how the system is addressed. For organizations that use proprietary field
devices, there is a good chance that these modern (but proprietary) devices may very well have some sort
of embedded vendor-specific security mechanism. If this is true, the investigator should be alert to the fact
that there will be a relationship between the activity at the field devices and the commanding control
equipment somewhere else in the network.
32
to providing for immediate recovery than to collecting log data for analysis. The dependence on input
from the vendor will be beneficial and unless trained in the actual device technology, the investigator will
most likely be unable to obtain any viable incident evidence. Such restrictions (as alluded to in previous
sections) will also limit the feasibility of any proactive strategies associated with understanding the
integrity of the core files in these devices.
From a networking perspective, legacy and proprietary field device structures will most certainly be
based on serial connections. For those installations that are trying to migrate to modern networking
infrastructures, protocol servers will augment serial-based connections to the field devices, subtly
translating between proprietary port serial-based communications and Ethernet-based mechanisms or
even wireless. For the most part, however, the information available from field devices in these types of
architectures will be able to provide little or no information beyond the vendor-specific fault tables in the
devices. The rapid sampling and data overwrite rates for these devices, in combination with the often-
trivial amount of memory inside them, combines to make a very difficult situation for the forensic
investigator. These issues are of course compounded by the fact that if there is indeed some sort of
incident it may interrupt the capability of the field device to work properly. Normal operations usually
demand the instantaneous reboot or the swapping out of the device so that process operations can
continue.
As the age of the system increases, it becomes more probable that the original vendor
responsible for the development of the technology is either no longer in business, the
contracts have expired, or there is simply no information about the device available. This
drives demand for “community-level” support, and as such, peer networks can become
one of the few remaining support mechanisms.
When faced with the challenge of working with legacy and proprietary field devices,
whether or not the devices are available online or offline, the vendor should be contacted
and an experienced engineer should be made available to support the investigation.
33
Table 4. Sample of possible artifacts and relevant forensic information.
The recording of system failures and event-based incidents are often required for
event recreation activities. This data can be of particular importance during the initial
steps of an investigation.
When reviewing modern cyber attack vectors, one of the key components of attack includes the
forced failure and rebooting of a compromised system, which allows any malicious code to impact the
function of the information resource attacked. Thus, as part of the cyber forensics plan that addresses how
systems will fail and recover (which may span several components), the investigator should assume that
the design and maintenance phases of the system development lifecycle may have been used to
incorporate safeguards as they relate to recovery. Although such methodologies may be vendor specific,
the inherent capability to restore system functionality as soon as possible will probably be within most
control systems technologies regardless of their classification.
34
In general, how a control systems component’s operating system responds to a type of failure can be
of value to the user/operator, and an understanding of what failures mean can contribute to an overall
better understanding of cyber security in the control systems domain. Moreover, an effective cyber
forensics plan that includes training, response, and management practices reduces system downtime and
increase overall security posture.
System failures may be categorized as:
• System reboot
• Emergency system restart
• System cold start
A system reboot takes place after the system shuts itself down (or is forced to shutdown) in a
controlled manner in response to a trusted computing base (TCB) failure. Also, if the system finds
inconsistent object data structures in its environment or if there is not enough space in some critical tables
to perform key tasking, a system reboot may take place. The reboot often releases resources and returns
the control systems component to a more stable and safer state.
Emergency system restarts often occur after a system failure happens in an uncontrolled manner. The
cause of these restarts may be anything from a core operation failing to work or a lower-privileged user
process attempting to access memory segments that are restricted. The system may see this as an insecure
activity that it cannot properly recover without rebooting. When this happens, the system enters a
maintenance mode and recovers from the actions taken and then is brought back online in a consistent and
stable state.
A system cold start takes place when an unexpected activity occurs and the regular recovery
procedure cannot recover the system to a more consistent state. The system and user objects may remain
in an inconsistent state while the control systems attempts to recover. The control systems user or
administrator may require intervention to restore the system.
From a forensics perspective, having the capability to monitor key file structures for integrity as well
as functionality is advantageous. Moreover, for systems with a high level of observed “uptime,”
unscheduled restarts could be indicative of a serious security issue. The importance of being able to
recognize and understand system faults is critical to forensics investigation in the control systems domain.
The cyber forensics plan should have instruction and guidance on how these incidents are observed and
reported, and can be deployed as new operational reporting standards or augmentations to existing
operator reporting practices. In either case, cyber security applicability to traditional recovery may help
contribute to a positive cyber security culture being developed and will support the overall reliability of
the control systems information architecture. Like all failures and reboots, the cause should be
investigated as required because there may be security issues requiring immediate attention.
Modern control systems domains have redundancy built in, with backup networks and key resources
mirrored to accommodate for any catastrophic failure. Often, in the case of networked infrastructures,
facilities have secondary ready-to-go systems, known as “hot standbys,” that are resident online
information and control systems assets ready to come online in the event of primary system failure. The
key to operations (and cyber security) is ensuring that these secondary systems are fully compliant with
current configurations and, if needed, can become operational with the exact same configuration and
system upgrades as the primary system. Therefore, if an event requiring a switchover to the redundant
system is required, operation can and will be maintained as if the main system was still online.
Unfortunately, many organizations do not ensure that key secondary systems are upgraded and configured
to the same cyber security requirements. There have been instances where secondary systems have been
35
vulnerable, and as such, ensuring secondary systems are secured prior to deployment becomes
important.cc
Being able to have an effective history of system faults can aid investigators in
pinpointing abnormal system activity and cross-reference with other obtainable
information such as time of day, operator actions, and device activity.
cc. Potential Vulnerability of Plant Computer Network to Worm Infection Notice, http://www.nrc.gov/reading-rm/doc-
collections/gen-comm/info-notices/2003/in200314.pdf.
dd. This is a very difficult issue to address, as leaving the impacted system online may allow the attack to perpetuate, while
failing over to back up systems may introduce the same vulnerability into the systems due to identical technology being used. In
addition, disconnecting the system causes system productivity to drop to zero.
ee. It would prove prudent to use tools that are based on industry standards, and these standards should be a combination of those
forming both the IT and Control System domain. In addition, taxation on system resources should be tested prior to installation
on a control system machine as minor delays may not be acceptable in a control system environment that may require rapid
execution of commands to end point field devices
36
not change very much. In addition, the data that is pervasive on the network is often very predictable, and
in many cases somewhat deterministic. All of these factors can combine to ensure that the network
administrators in the control systems domain have a well-structured, common operating picture as to how
the system is operating and what it does when it is operating. In reviewing the control systems reference
architecture, it shows that many of the vital technologies in the control systems network will have some
sort of application, operating system, or logic associated with its normal function. These elements are well
known by the operators, engineers, and administrators and are most likely to have copies and support
backups to aid in the reconstitution of the system in the event of failure.
Many of the key elements in the control systems architecture, specifically the field devices, will have
logic associated with them that does not change. Although in some network entities the state information,
resident memory, connection entries, and other service data will change, some entities may have datasets
in the form of firmware or logic that are not intended to be changed. This static nature of control systems
provides the opportunity for the organization to take specific baseline measurements of key system
internals that are not intended to change. These baseline measurements should be taken both prior to
deployment and following each authorized change in order to verify a device’s integrity.
A variety of different means can complete these specific measurement activities, including
checksums, hashes, or some other type of quantitative measurement associated with that particular
instance of data logic. A simple example would be to run a hash algorithm, such as SHA-1, on the logic
found in the critical field devices. The hash associated with the field devices can be kept offline or stored
in a secure read-only environment. In the event of a cyber incident considered to have played a part,
simple calculation of the hash on the existing logic inside the device would provide comparison for the
known hash that was previously calculated. If such an activity showed the hash as different, the
investigator would immediately know that the tampering or changing of the logic within the field device
was a component of the incident. Such knowledge would greatly empower the investigator, who would
quickly be able to develop an information flow between the field devices and other control systems
information resources. This would allow for the targeting of specific entities that may have appropriate
logging or transaction information, or provide insight as to what information resources should be
processed for evidence collection.
Forensics practice includes the process of hashing, which is considered one of the mandatory steps in
evidence collection. Once an investigator has actually acquired hard drives or other memory devices for
evidence, the investigator needs to be able to make an exact copy of the device so the original may be
preserved. By using various copy mechanisms in combination with proven signature and hash methods,
the investigator is able to make exact copies of the information to allow for the extraction of evidence in a
manner that would be admissible in a court of law. For certain data, logic, or other operational
instructions sets that are deployed to be static in nature, the use of hashing can be extended to help
investigators in the event of a cyber incident.
37
Additionally, this is a very simple task for an experienced administrator that can have far-reaching
benefits (assuming there is no adverse effect on the performance of the system).
In the event that a field device lacks sufficient logging or audit capabilities, it is encouraged to log all
network traffic to and from the device. Logging network traffic may aid the forensics investigator in
determining a specific device’s role in a cyber event. A passive network capture device can accomplish
this. Most modern operating systems allow for this functionality with the installation of a software
package such as Wireshark TM or TcpDump TM.
Lastly, it is highly recommended that all audit, network, and device logs are stored securely either
offline or in a read-only capacity. This is a critical step in maintaining the integrity of the logs for an
examiner to utilize during an investigation as an attacker will often try to destroy or modify these logs to
cover their tracks.
38
3. ACTIVATING AND SUSTAINING
A CYBER FORENSICS PROGRAM
A cyber forensics program in any organization must be very closely tied to the incident response
capability of that organization. Although this document does not extend to the recommendations and
practices associated with providing evidentiary information for legal proceedings, the steps and methods
suggested in this recommended practice clearly outline functionality that can be used to support incident
response activities. In most cases, following the detection and correction of a cyber incident, forensics
investigation will be used to ascertain the cause of the incident, as well as other fundamental attributes
that can be collated to provide a full cognitive picture surrounding the cyber event.
It is clear that for executing successful forensics operations within a control systems environment,
there should be some sort of pre-deployed and proactive effort to assist investigation in the event of a
cyber incident. The volatility of data, the uniqueness of the technology, the proprietary components of the
control systems environment, and the almost certain need to have the equipment running and available all
combine to create substantial hurdles for the forensics investigator. As such, it becomes paramount to
create activities for introduction prior to the cyber incident, such that following an incident response
investigation the forensics program will have a pre-established starting point. It is also very important to
introduce these types of solutions so that they are not intrusive, and in no way impact the business
operations, stability, or the critical functionality related to the control systems environment.
39
Table 5. Roles matrix for incident response and forensics in control systems.
The forensics function inside an organization can be designed to support the incident response
function, both during and after initial response phases. Those overseeing the forensics investigation will
require access to all of the information obtained by the incident response team, as well as all reporting,
ideas, and after action activities that have been developed. If the organization is interested in ascertaining
if any legal wrongdoing has occurred, the analysis of the evidence will be critical to the organization’s
wish to advance with prosecution.
The collection of the evidence, as well as the adherence to appropriate chain of custody practices, will
be as important in a control systems investigation as in any other cyber forensics investigation. Thus, it is
recommended that the organization include members from the forensics team to be active or, at the very
least, passive participants in an incident response activity. By doing this, the investigator has a much
clearer understanding of the incident from discovery to remediation and can focus on the impacted
elements. The team members will include:
Control Systems Incident Manager (CSIM) – This team leader will be responsible for coordinating
the response with control systems personnel and those responsible entities who oversee IT operations (and
security operations) with the control systems domain. The CSIM will engage for the entire response
portion that involved control systems, and will being involvement after the Detection phase. Each of the
core activities will be primary except for the Summary Report, with special emphasis on ensuring tactical
involvement at all phases of the response. The CSIM will oversee the translation of duties and activities
from the primary IR function to the Control Systems Security Specialist (CSSS), the CS Engineering
40
(CSE) team, and any vendor coordination that is required. The CSIM will interface directly with both the
IR Director and the IR Coordinator.
Control Systems Security Specialist (CSSS) – This function will provide critical support from a
security perspective, and will aid in both the recovery and mitigation phases of the incident response. The
CSSS will also be involved in ascertaining what critical assets may have been impacted, and will work
with control systems engineering, vendors, and other members of the incident response team as required.
The CSSS will initially have secondary functions during both the Response Initiation and the Forensics
Collection phases, migrating to a primary role during the latter part of the incident resolution lifecycle.
The CSSS will work closely with both engineers and incident managers supporting both investigation and
containment activities, and will have specific tactical activities supporting restoration, reporting, and
analysis. This role will also play a significant part in the in-depth analysis of the acquired data and be very
familiar with methodologies that can be used to overcome the challenges associated with data collection.
Control Systems Engineering Support – The importance of having control systems engineering
support an incident response and forensics function cannot be overstated. As in most control systems, it is
the engineering function that understands the control system operation better than anyone understands,
and can work effectively with both the primary incident response team and the control system incident
manager. Being able to have the control systems engineering capability support primary functions such as
containment, recovery planning, and restoration (as well as system upgrade), will provide significant
value to both incident response and forensics activity. In addition, it is the control systems engineering
function that may be able to facilitate more effective liaising with the vendor community.
Due to the uniqueness of the data and the relationships amongst the information resources in the
control systems domain, a team comprised of individuals that have an advanced understanding of the
system should complete an analysis of collected evidence. This analysis team needs to be vetted to ensure
that there is a low risk of one on the team members being actually involved in instigating, propagating, or
trying to conceal the incident.
41
4. REFERENCES
AGA-12: Cryptographic Protection of SCADA Communications General Recommendations,
http://www.aga.org/NR/rdonlyres/B797B50B-616B-46A4-9E0F-
5DC877563A0F/0/0603AGAREPORT12.PDF, Web site accessed July 2008.
ANSI/ISA-TR99.00.01, Security Technologies for Manufacturing and Control Systems, 2007.
Braid, Matthew, “Collecting Electronic Evidence after a System Compromise,” Australian Computer
Emergency Response Team, December 2001.
Barrett, Neil, Computer Forensics Jump Start: Computer Forensics Basics, Solomon, Sybex Printing,
2005.
Cyber Security Procurement Language for Control Systems Ver. 1.8
http://www.msisac.org/scada/documents/4march08scadaprocure.pdf. Websiet accessed July 2008
Department of Energy, 21 Steps to Improve Cyber Security of SCADA Networks,
http://www.oe.netl.doe.gov/docs/prepare/21stepsbooklet.pdf, Web site accessed July 2008.
Department of Energy, Common Vulnerabilities in Critical Infrastructure Control Systems,
http://www.oe.netl.doe.gov/docs/prepare/vulnerabilities.pdf, Web site accessed July 2008.
DHS United States Computer Emergency Readiness Team (US-CERT), http://www.us-cert.gov.
DHS Control Systems Security Program, http://www.us-cert.gov/control_systems/, Web site accessed
July 2008.
Electric Power Research Institute, http://www.epri.com/, Web site accessed July 2008.
Executive Order 13231: Critical Infrastructure Protection, http://www.fas.org/irp/offdocs/eo/eo-
13231.htm, Web site accessed July 2008.
Federal Energy Regulatory Commission Order 702 Final Rule
www.ferc.gov/EventCalendar/Files/20071030162551-RM06-23-000.pdf, Web site accessed July
2008.
Information System Security Association, http://www.issa.org/, Web site accessed July 2008.
Information Systems Audit and Control Association, http://www.isaca.org/, Web site accessed July 2008.
Instrumentation, Systems, and Automation Society, http://www.isa.org/community/SP99, Web site
accessed July 2008.
National Association of Regulatory Utility Commissioners, http://www.naruc.org/, Web site accessed
July 2008.
NIST PCSRF, http://www.isd.mel.nist.gov/projects/processcontrol/, Web site accessed July 2008.
NIST 2nd Public Draft: Guide to Industrial Control Systems (ICS) Security
http://csrc.nist.gov/publications/drafts/800-82/2nd-Draft-SP800-82-clean.pdf. Web site accessed July
2008
NIST Special Publication SP 800-53 Revison 1 Appendix I http://csrc.nist.gov/publications/nistpubs/800-
53-Rev1/800-53-rev1-final-clean-sz.pdf. Web site accessed July 2008
NIST Special Publication SP 800-53 Revision 1 Appendix F Augmented for Industrial Control Systems
http://csrc.nist.gov/groups/SMA/fisma/ics/documents/papers/ICS-Augmentation-Appx-F-800-53-
rev1_clean_22jun07.pdf. Web site accessed July 2008
42
North American Electric Reliability Council (NERC), http://www.nerc.com/, Web site accessed July
2008.
Partnership for Critical Infrastructure Security, http://www.pcis.org/, Web site accessed July 2008.
Presidential Decision Directive 63: Critical Infrastructure Protection,
http://www.fas.org/irp/offdocs/pdd-63.htm, Web site accessed July 2008.
Process Control Systems Forum (PCSF), http://www.pcsforum.org/, Web site accessed July 2008.
RFC 3227 Guidelines for Evidence Collection and Archiving, http://www.faqs.org/rfcs/rfc3227.html Web
site accessed July 2008.
Sandia National Labs Center for SCADA Security, http://www.sandia.gov/scada/home.htm, Web site
accessed July 2008.
Thiagarajan, Val, Information Security Management BS 7799.2:2002 Audit Check List for SANS, 2003.
The White House, Presidential Directive on Critical Infrastructure: Identification, Prioritization, and
Protection - HSPD-7, http://www.whitehouse.gov/news/releases/2003/12/20031217-5.html, Web site
accessed July 2008.
The White House, The National Strategy for the Physical Protection of Critical Infrastructures and Key
Assets, http://www.whitehouse.gov/pcipb/physical.html, Web site accessed July 2008.
The White House, The National Strategy to Secure Cyberspace, http://www.whitehouse.gov/pcipb/, Web
site accessed July 2008.
Technical Support Working Group (TSWG) SCADA Security Web site,
http://www.tswg.gov/subgroups/ps/infrastructure-protection/products.html#, Web site accessed July
2008.
Government of Western Australia Department of the Premier and Cabinet Office of e-Government
Forensic Plan, http://www.egov.dpc.wa.gov.au/documents/forensic_plan.pdf. Web site accessed
February 2008.
43