Darktrace white paper_ics_final

Cyber Security for Corporate and Industrial Control Systems
WHITE PAPER
Darktrace Industrial Immune System Provides Continuous Threat Monitoring for Oil &
Gas, Energy, Utilities, and Manufacturing Plants

"CISOs responsible for cyber security strategies should
consider solutions for advanced persistent threat detection and
analysis that can be used across both IT and OT environments."
Earl Perkins, Gartner Cool Vendors in Energy and Utilities, 2015
Darktrace named Gartner 'Cool Vendor' 2015
"The worst case scenario is a critical infrastructure attack, and
these organizations are ill prepared to deal with it."
Dr Larry Ponemon, Founder of Ponemon Institute

White Paper
3
Executive Summary
Industrial Control Systems (ICS) underpin both individual businesses and large parts of the National Critical Infrastructure.
They maintain control over facilities such as power stations, water distribution and car production lines. Historically they
were kept separate from corporate networks, but significant achievable business benefits are driving a convergence
between Operational Technology (OT) systems, such as ICS, and the corporate Information Technology (IT) environment,
and hence the wider internet.
The business of cyber security has changed dramatically in the past few years, presenting a significant challenge to
managementteamsacrossallindustriesandbusinessdomains.Organizationstodayareinauniquepositiontoquantifiably
outpace threats and manage them to minimize organizational impact, whether that be reputational, financial or physical.
We see an increasing trend toward IT security teams taking on more accountability and responsibility for securing the OT
systems, which require different specialist skills and working practices. This cultural and technical convergence will be a
steep learning curve, one to be overcome.
Now open to the same attack vectors used in the majority of cyber-attacks, ICS devices are inherently much less secure but
their compromise can lead to enormous physical damage and danger to human lives. Ever since the Stuxnet malware was
widely reported in 2010, threats to industrial systems have grown rapidly in both number and capability. This was made
clear in the 2014 compromise of a German steel mill that caused massive damage to a blast furnace. Ongoing malware
campaigns such as ‘Energetic Bear’ are actively acquiring critical data about control systems, while quietly maintaining
persistent access. Existing defenses such as firewalls have repeatedly proven inadequate on their own, especially against
insiders who may already have privileged access.
Darktrace’s Industrial Immune System is a fundamental innovation that views data from an ICS network in real time,
and establishes an evolving baseline for what is normal for operators, workstations and automated systems within that
environment. Advanced Bayesian mathematics and cutting-edge machine learning detect abnormal behavior and flag
it for investigation, capable of discovering previously unknown attacks as they emerge. Total prevention of all cyber
compromises is not a realistic goal but, if identified early enough, threats can be mitigated before they become a full-
blown crisis. Darktrace’s technology can be deployed across both IT and OT environments to provide full coverage of an
organization.

4
•• Enhanced performance through cost and time
saving which allows for the smooth transition
of newly-developed products into existing
manufacturing operations, reducing time to market
•• Business optimization using data transferred
between IT and OT environments
The breakdown of this cultural divide between OT and
IT staffs will often require CISOs to manage across
teams that historically have different approaches to
cyber security. During this convergence the assurance
of long-term reliability and safety requires CISOs to
reshape enterprise security practices. The merging
of specialized OT systems with IT technologies and
endpoints will require CISOs to assume responsibility
for OT cyber security without specialized OT skills or
in matrix-based organizational environments, thereby
exposing new technology and change-management
risks. This gap in skill sets as IT and OT systems converge
will generate new cyber security problems as attacks
become more focused and sophisticated. A strategic
and unified approach to cyber security will inevitably
benefit organizations, allowing them to operate in a
more reliable and efficient manner.
Industrial Control Systems face numerous cyber-
security threat vectors with varying degrees of potential
loss, ranging from non-compliance to disruption of
operations which could result in destruction of property
and, unfortunately, potential loss of human life.
Examples of potential ICS-related threats include:
•• Advanced Persistent Threats (APTs)
•• Unintended spillover of corporate network
compromises
•• Disruption of voice & data network services
•• Coordinated physical & cyber-attack
•• Insider sabotage
•• Hacktivist attacks
•• Supply chain disruption or compromise
•• Catastrophic human error
•• Distributed Denial of Service (DDOS)
The cost is significantly higher to remediate a system
than to detect a cyber threat early, not only in time
and money, but also in safety and reputation. Legacy
approaches have fallen short as evidenced by cyber-
attacks ranging from the infamous Stuxnet to a recent
German steel mill compromise. What if the Saudi
Aramco attack had been aimed at critical infrastructure
instead of business workstations?
Enabling Modern Industry
Industrial Control Systems (ICS) are at the heart of
modern industry, monitoring and controlling complex
processes and equipment. Many businesses are
wholly underpinned by the reliable functioning of
this Operational Technology (OT), such as automated
production lines at car manufacturing plants. For
organizations that form part of the National Critical
Infrastructure, the consequences of unplanned
outages are far-reaching, being responsible for
maintaining utilities such as power, heating and clean
water to huge numbers of households and places of
work. In many industrial processes, reliability of an ICS
has a direct and immediate impact on the safety of
human lives.
ICS and SCADA
ICS is an umbrella term covering many historically
different types of control system such as SCADA
(Supervisory Control and Data Acquisition) and DCS
(Distributed Control Systems). Also known as IACS
(Industrial Automation and Control Systems), they are
a form of Operational Technology. In practice, media
publications often use “SCADA” interchangeably with
“ICS”.
Corporate Information Technology (IT) systems and
Industrial Control Systems have different objectives,
even when operating within the same organization.
While IT and OT often speak different languages, cyber-
attacks across both environments have continued to
evolve to become more targeted and destructive. When
it comes to ICS, reliability is the primary concern as
attackers aim to disrupt the critical services customers
rely upon.
IT and OT systems are converging, driven primarily by
economic pressures resulting from globalization and
intensifying competition, along with the benefits and
eventual competitive advantages that stem from the
integration of these disciplines. These benefits include;
•• Cost reduction by applying similar technology,
standards and governance principles for IT and OT,
including remote management
•• Risk reduction through jointly addressing safety
issues, leading to an integrated approach that
provides enhanced security against cyber intrusions
from outside the company and to central cyber-
security governance within the company

White Paper
5
ICS Cyber Security Issues
Historically, industrial control environments were ‘air-
gapped’; physically isolated from corporate networks
and the internet. However, computer viruses and
other forms of cyber-attacks such as Stuxnet [1] and
“agent.btz” [2] have been known to bridge the gap
by exploiting security holes related to the handling
of removable media, or simple human error. While
security is an upside of having a seemingly closed or
isolated system, the downside includes the limited
access or inability to access enterprise decision
making data or to allow control engineers to monitor
systems from other networks. Additionally, ICS often
tie together decentralized facilities such as power, oil
& gas pipelines, water distribution and wastewater
collection systems, among many others, where the
network is hard to physically secure.
ICS systems, whilst effectively designed to be
interoperable and resilient, are not necessarily
secure. With the increasing number of connections
between ICS systems, corporate networks and the
internet, combined with the move from proprietary
technologies to more standardized and open solutions,
they are becoming more susceptible to the kind of
network attacks that are found more commonly in IT
environments.
Cyber-security researchers are particularly concerned
about the systemic lack of authentication in the
design, deployment and operation of some existing
ICS networks and the belief that they are completely
secure simply because they are physically secure. It
has become clear that any possible connection to the
internet can be exploited, even if it is not direct. ICS-
specific protocols and proprietary interfaces are now
well documented and easily exploited. The use of a VPN
(Virtual Private Network) is also not sufficient protection
for ICS users as this can be trivially bypassed with
physical access to network switches and never provides
end-to-end coverage. ICS vendors are increasingly
urging CISOs to converge their approaches to IT and
OT cyber security, with an equal level of caution and
depth in defense strategy.
Challenges Facing Industry
Industry faces a growing challenge in dealing with
cyber threats, both external and internal. There are
an increasing number of threat actors with both the
motivation and capability to compromise industrial
control networks and devices. The consequences of
compromise range from damaging to catastrophic,
from immediate physical harm to long-term industrial
espionage.
Control engineers historically have not had to worry
about cyber threats coming through corporate IT
systems, while IT security staff have had little to do with
the fundamental differences in control systems or the
physical equipment that those systems manage. ICS
devices are inherently insecure, and extremely difficult
to update with even the rudimentary protections that
are possible.
A New Approach: Darktrace and the
Immune System
Utilities, OT-centric industries and other national
infrastructure organizations, are challenged with
rethinking cyber security across all technologies to
deliver continuous insight that provides early warning
of both indiscriminate and targeted compromises,
supported by mechanisms that can manage incidents
before they become a business crisis. Total prevention
of compromise at any cost is untenable, however,
detection and response to prevent a crisis from
developing is an achievable cyber security goal in an IT/
OT environment.
Darktrace’s Industrial Immune System for ICS is a
fundamental innovation that implements a real-
time “immune system” for operational technologies
and enables a fundamental shift in the approach to
cyber defense. Based on groundbreaking advances in
Bayesian probability theory and powered by cutting-
edge machine learning, Darktrace analyzes data and
"Darktrace adds another level
of sophistication to our defense
systems, and had already
identified threats with the potential
to disrupt out networks."
Martin Sloan, Group Head of Security, Drax

6
creates a unique behavioral understanding of “self”
for each user and device within the network and, like a
biological immune system, it detects threats that cannot
be defined in advance by identifying even subtle shifts
in expected behavior. People and devices all behave
in a unique way that necessarily differs from their
peers to varying degrees. However, their behaviors are
significantly more predictable when compared to their
historical behaviors and patterns of change.
With Darktrace’s self-learning “immune system”,
organizations are able to detect and respond to
emerging threats, even if novel or tailored, and
regardless of whether they originate in either the IT
or operational domains, or traverse between them.
By identifying unexpected anomalies in behavior,
defendersareabletoinvestigatemalwarecompromises
and insider risks as they emerge and throughout stages
of the attack lifecycle. Darktrace provides the real-time
visibility required to make intelligence-based decisions
in live situations, while enabling in-depth investigations
into historical activity.
Real Vulnerability
While it is likely that many attacks are never revealed to
the public, the list of known compromises is growing.
The most notorious incident that arguably propelled the
vulnerability of ICS into the mainstream consciousness
was the discovery of the Stuxnet attack in June 2010
[1], a “weaponized” form of malware. Stuxnet targeted
the Natanz nuclear facilities in Iran with great precision,
causing nuclear centrifuge equipment to wear out at a
vastly increased rate.
Sabotage and Shutdowns
Significant attacks have been made by former
employees who wrongfully retained access following
dismissal, such as Mario Azar, who was indicted for
disabling a computer system detecting pipeline leaks
in Southern California [3]. Attacks by individuals
who never possessed legitimate access include the
compromise of the South Houston, TX water system.
[4].
ICS networks have also been damaged as unintended
side effects of problems starting in corporate networks
that took advantage of the increasing connectivity,
proving clearly that the standard PCs which now form
part of a typical ICS are open to the same compromises
as their enterprise counterparts. At least three
problems at major power stations have been publicly
attributedtothis;theDavis-Bessenuclearpowerstation
(Ohio, USA) when safety systems were crippled by the
Slammer worm [5], the Browns Ferry nuclear power
station (Alabama, USA) being manually scrammed as a
result of a drastic increase in network traffic [3], and the
Hatch nuclear power station (Georgia, USA) due to a
faulty software update on a business network machine
that communicated with the control network [6].
German Steel Mill
At the end of 2014, the most significantly publicized
attack since Stuxnet was revealed in a German report
disclosing that hackers had struck an unnamed steel
mill in Germany [7]. This was a targeted Advanced
Persistent Threat (APT) compromise, beginning with a
spear-phishing attack that enabled the hackers to gain
initial access onto the office network of the steelworks.
From there, they were able to successfully explore the
company’s networks and eventually manipulate and
disrupt the production networks. Failures of individual
control components accelerated, resulting in a blast
furnace being unable to shut down which caused
“massive” damage to the installation.
Reconnaissance and Pre-positioning
The goals of the largest known ongoing ICS attack
campaigns have mostly shifted away from active
sabotage to long-term persistent compromise and
reconnaissance. Recently the Energetic Bear campaign
has used the Havex [8] Remote Access Trojan (RAT) and
the Sandworm APT group have been using a variant of
the BlackEnergy malware [9]. In both cases ICS-CERT,
the USA’s Industrial Control Systems Cyber Emergency
Response Team, have long-running alerts tracking
them [10][11]. Both provide persistent external access
to compromised control networks and are capable
of downloading additional modules to enhance their
capability. Having identified all of the devices in a
network, it would be simple for them to download
additional sabotage modules and cause immediate,
widespread damage.
Havex was targeted against ICS customers by using
a highly effective ‘watering-hole’ attack, where the
attackers compromised three legitimate ICS vendor
websites and replaced real software updates with
versions already containing the malware. There was no
possible way for traditional network defenses such as
border firewalls to protect against this, and standard
procedures employed in many corporations would
have trusted the trojanized updates and added them
to internal whitelists of software for authorized use.

White Paper
7
If an environment is infected in this manner, only its
unique behavior, once installed on the ICS network,
could be used to detect Havex’s presence.
A survey published in April 2014 by the SANS Institute
[12] reported a significant increase in the number of
identified or reported breaches of control systems
over just the previous twelve months. Respondents
also noted that their ability to protect these systems
had not improved within the same period. This is a
chilling indictment of the challenges facing the OT
cyber-security efforts of organizations today.
Darktrace Technology
New vulnerabilities are emerging at a pace that is
difficult to keep up with, and looking only for published
historical attack types is an unsuitable approach for
operationally important environments. Darktrace does
not require a priori assumptions about environments
or threats, and can therefore detect the ‘unknown
unknown’ threats that are as yet unidentified, either
because they are novel or have been tailored to a
particular defender.
The Darktrace architecture continues to adapt and
self-learn throughout its entire deployment. Its
understanding is constantly being revised and refined
in light of new evidence as it ingests and analyzes new
information - the more data it sees, the more it learns.
This adaption means that no new or customized threat
has the ability to hide from Darktrace. Whenever an
abnormal change to behavior takes place within the
environment, the Industrial Immune System identifies
deviations from the learned ‘pattern of life’ and alerts
the organization to the possible threat. Changes that
are not real threats are incorporated into Darktrace’s
evolving understanding of normality.
The advanced mathematics inside Darktrace make it
uniquely capable of highlighting significant potential
threats without burying them beneath many
misguided, insignificant or repeating alerts. Far more
than a set of simple rules applied to network traffic,
it can correlate many subtle indicators separated by
location or time into strong evidence of a real emerging
threat, meaning that security analysts are not flooded
with false positives.
Darktrace’s Threat Visualizer interface can be used
to manage these detections, but it is also possible to
route the output to an organization’s existing Security
Information and Event Management (SIEM) system, to
integrate with established processes and procedures.
Passive Observation
Connecting new devices into a corporate network is
straightforward and routine, with little attached risk.
The same is not true of industrial networks, where for
many applications even the slightest interruption in
service could be damaging. This is why larger and more
critical networks are left as untouched as possible
between planned outages.
The Darktrace appliance runs on a server that is
connected completely passively to an ICS network,
receiving copies of as much communication traffic
as possible. It does not interfere with the operation
of the control network in any way, flagging anomalies
for investigation but not attempting to influence the
situation. The appliance receives copies of raw network
data using the built-in port mirroring or “spanning”
capabilities of network switches, or using fail-safe
taps, sometimes via an aggregator to bring together
numerous connections in one location.
ICS networks are deliberately segregated into Trust
Levels as defined by the ISA95/Purdue reference model
[13], depending on how much each device on the
network is trusted to behave as expected. Darktrace
can be connected at Level 2 (supervisory control), Level
3 (data servers) and Level 4 (IT networks) to provide
defense in depth. It also extends cyber-security
coverage down into Level 1 (field devices).
A highly flexible, distributed architecture allows
Darktrace to securely cover multiple Trust Levels and
the wide variety of network topologies within and
between them. Examples include wholly separate
appliances for each Trust Level, or multiple appliances
within a widely distributed single Trust Level with
a master appliance providing a single interface. If
required, a network diode device could guarantee that
a channel for moving data from one Trust Level to a
higher Trust Level to reach a single appliance covering
both cannot be used to communicate in the other
direction.

8
Darktrace Proof of Value
Darktrace's Proof of Value (POV) allows organizations
to experience first-hand its Industrial Immune
System's ability to detect previously unseen threats
and anomalous behaviors within a customer’s own
environment. Along with the POV, Darktrace provides
access to our Threat Visualizer (below) for use during
the POV as well as weekly Threat Intelligence Reports
produced by its team of cyber security specialists.
Some organizations prefer to trial Darktrace on their
corporate IT systems to confirm the passive and
secure operation before engaging installation into ICS
networks.
Visibility Into Industrial Control Systems
Architectures of ICS systems and their operational
networks are often documented to a standard that
exceeds corporate equivalents, but these long-lived
environments are complicated and will typically have
undergone many changes by multiple individuals over
their lifetime. Knowing and understanding what is
genuinely happening inside the environment can be a
real challenge. Darktrace addresses this challenge by
observing, analyzing and capturing communications
along with their associated metadata.
In addition to its core identification of anomalous
activity and possible compromise, Darktrace’s Threat
Visualizer interface uniquely displays all this rich
information in an intuitive 3D dashboard that allows
the operator to get a true and real-time overview of
what is happening. This can be used to investigate
whether the control system’s real behavior matches its
intended design.
Darktrace’s Industrial Immune System retains all of the
capabilities of Darktrace in the corporate environment,
and will ideally be deployed observing both the ICS
and corporate networks. The most likely attack vector
for ICS compromise is the IT network. Discovering
threats while still within the corporate network vastly
increases the defense-in-depth of the control system.
This also protects confidential data about the control
system stored on corporate servers, which might
include detailed operational diagrams, device details
or efficiency and safety reports.
Fig. 1 - Threat Visualizer

White Paper
9
Insider Threat
Threat from ‘trusted’ insiders is an important consideration for OT environments. Over the long lifecycles
involved with the building and utilization of infrastructure and manufacturing equipment, a large number of
different individuals, including both permanent staff and short-term contracted specialists, will usually have
interacted with control systems. Many of them will have had privileges that allow them to modify configurations
or the underlying software and hardware.
Vetting and training staff can reduce but not eliminate the risk of insider incidents from occurring. These
incidents can be unintentional due to a mistake or intended short-cut that puts something important at risk, or
a deliberate act by a disaffected or ideologically motivated individual. The increased access and organizational
familiarity that insiders have means their malicious actions can be very well targeted and effective at disrupting
operations. They also have a greater ability to interfere with monitoring or masquerade as others, making their
activities harder to identify and attribute.
Insider risk is a serious challenge often underestimated in breadth. When supply chains or contractors are
involved, it becomes impossible to draw a neat line between ‘inside’ and ‘outside’. We need to trust people in
our extended organizations with the access and privilege that they require to do their jobs, but we also need
mechanisms to identify when something is going wrong and needs to be corrected.
Traditional network border defenses such as firewalls perform an important function in a complete cyber-
security solution, but insiders are a key example of their limitations. Insiders do not have to pass through
border defenses to accomplish most of their potential goals, meaning that those defenses have no chance at
all to prevent or identify their actions [14].
Given the complexity and the variety of people and processes that make up an organization, any monitoring
approach needs to start from a complete understanding of what is normal for the unique environment. Only
then can it have the insight to identify subtle patterns and correlated action over time that can be the only
early signs of emerging issues, and allow them to be handled before they become major crises.

10
Conclusion
Businesses face many challenges as we move into an era of ever increasing connectivity and standards of communication.
Those trying to secure industrial control systems as well as corporate networks face additional and substantially different
problems, as the devices involved are far less secure than their corporate counterparts.
There is public evidence of growing motivation and capability of threat actors towards control systems, a trend likely to
continue and brought into sharp focus by the 2014 cyber sabotage of a German steel mill. This attack used state-of-
the-art methods to reach the control system of a target with little political or ideological significance, a combination not
previously observed.
De-risking the OT environment is a perpetual challenge requiring new technologies that will deliver continuous insight
and provide early warning of both indiscriminate and targeted compromises. Total prevention of compromise seems
effectively impossible for the foreseeable future, but prevention of crises is an achievable goal across both corporate IT
and operational technology environments.
A new approach that can manage incidents across corporate IT and OT before they become an operational crisis is
required. With Darktrace’s self-learning immune system, organizations are able to detect and respond to emerging
threats in real-time. Advanced behavioral analysis mathematics can detect even previously unseen novel or tailored
attacks, regardless of whether they originate in the corporate IT or OT domains or traverse between them.

White Paper
11
Resources
These additional resources, available from our website, complement the information in this white paper.
Data Sheet: Stuxnet Example and Full References
For a detailed example of how Darktrace can detect previously unseen ICS cyber-threats like Stuxnet, and the full list of
supporting references to this White Paper, please email ics@darktrace.com
Data Sheet: Standards and Compliance for ICS: NERC CIP V5
Darktrace Industrial Immune System can help organizations transition to the new cyber-security standards set by the
North American Electric Reliability Corporation can be found at www.darktrace.com/resources/data-sheets
Case Study: Darktrace at Drax
A case study of the use of Darktrace at Drax can be found at www.darktrace.com/resources/case-studies
White Paper: Enterprise Immune System White Paper
A white paper covering the wider use of the Darktrace Enterprise Immune System across a whole organization can be
found at www.darktrace.com/resources/whitepapers

ICS-001r2en Darktrace © Copyright 2015 Darktrace Limited. All rights reserved. Darktrace is a registered trademark of Darktrace Limited. Enterprise Immune System, and Threat
Visualizer are unregistered trademarks of Darktrace Limited. Other trademarks included herein are the property of their respective owners.
About Darktrace
Darktrace is one of the world’s fastest-growing cyber threat defense companies and the leader
in Enterprise Immune System technology. Darktrace detects previously unknown threats in real
time using advanced machine learning and mathematics developed at the University of Cambridge
to analyze the behavior of every device, user and network within an organization. Some of the
world’s largest corporations rely on Darktrace’s self-learning appliance in sectors including
energy and utilities, financial services, telecommunications, healthcare, manufacturing, retail and
transportation. The company was founded in 2013 by leading machine learning specialists and
government intelligence experts, and is headquartered in Cambridge, UK and Washington D.C., with
offices in London, Milan, New York, Paris, San Francisco, and Singapore.
Contact Us
US: +1 (917) 363 0822
Europe: +44 (0) 1223 350 653
Email: info@darktrace.com
www.darktrace.com

Darktrace white paper_ics_final

More Related Content

What's hot (20)

Similar to Darktrace white paper_ics_final (20)

More from CMR WORLD TECH (20)

Recently uploaded (20)

Darktrace white paper_ics_final