Managing Cyber Security
Managing Cyber Security
Since the White Paper, there has indeed been an increase in awareness, although this has arisen
out of necessity, in response to attacks and incidents experienced by the most developed
countries. Network unavailability on a massive scale, attacks on government information
systems, espionage targeting strategic companies, destabilization operations and failures of all
types have unfortunately been in the news in the last three years.
Industrial Control Systems (ICSs), even when not connected to the Internet, are exposed to such
threats. The Stuxnet worm, which appeared in 2010, provides tangible proof that our worse fears
regarding attacks on sensitive plants may become a reality. In 2009 an ingenious and unwitting
adolescent was able, via the Internet, to derail a tram in Poland, demonstrating the vulnerability
of its signalling system. Other cases include pipeline break and water pollution. Unfortunately,
examples abound.
Cybersecurity concerns are therefore just as relevant to ICSs as they are to other information
systems, and probably even more so.
The issue is: just like society as a whole, industries have in many cases adopted digital
technologies on an ad hoc basis and without any prior strategy, and a variety of different
systems are interconnected, with the main concerns being productivity, efficiency and safety –
but rarely security.
Industries possess one advantage in terms of the security of their information systems: they have
a robust culture of dependability for their plants, and in most cases they possess in-house
cybersecurity competences for their office systems. These two cultures must now come together
and efforts must be combined in order to protect ICSs appropriately.
The guide on the cybersecurity of ICSs published by the French Network and Information
Security Agency, ANSSI (Agence nationale de la sécurité des systèmes d’information), has been
created to support companies in this endeavour. This is the first publicly available version of the
guide. It is designed to evolve as standard practice changes, and in response to your
In addition, it is a practical case study designed to illustrate scenarios posing a risk to companies
and to show how these are to be dealt with.
Finally, this guide is not solely intended for ICSs; its content also applies to the following non-
exhaustive list of systems: data centres, smartgrids, building management systems (BMS),
centralized technical management systems (CTM), and many embedded systems.
The purpose of the guide is to assess the cybersecurity of industrial control systems. Although
specific to each facility, ICSs are in most cases made up of the following components:
Programmable Logic Controllers (PLC);
Distributed Control Systems (DCS);
Safety Instrumented Systems (SIS);
Sensors and actuators (intelligent or non-intelligent);
Fieldbus;
Supervisory control software: SCADA;
Manufacturing execution system (MES);
Engineering and maintenance software;
Embedded systems.
ICSs currently make abundant use of information technologies, but they were not designed to
deal with threats that the latter present. There are now numerous examples of published ICS
vulnerabilities (for example, concerning Modbus and OPC protocols).
This is why they need to be included in general discussion on the security of company
information systems1.
The purpose of this guide is not to set out an exhaustive list of recommendations or to set out all
of the components of an ICS. Rather it proposes a strategy for awareness-raising and
implementation of good practices to support companies in the implementation of security.
There are no ideal or "one-size-fits-all" solutions. Appendix B sets out potential good practices.
Each system has its own unique characteristics and risks that need to be analysed in order to
implement suitable solutions whilst limiting the impacts on the core activity of the company.
Securing a system entails costs that are often difficult to calculate. So are the resulting gains.
Nevertheless, this securing process will protect company investments and production. This is
why it is important to define the right objectives and adapt these to requirements. Paradoxically,
"over-security" can result in effects contrary to those sought and undermine industrial
1
A company information system includes all of a company's information systems, i.e. management information systems
(information systems for departments and office applications, human resource management, customer relations and
integrated management) and industrial control systems.
The abbreviations and acronyms used in the document are set out in Appendix C.
There are number of myths about industrial control systems. The most commonly held are
examined here.
Myth Reality
"cybersecurity is not compatible with On the contrary, cybersecurity and dependability are
dependability." consistent in a number of ways, see §2.3.3.
2
Some "exploit kits" are officially sold as cybersecurity tools with the aim of detecting vulnerabilities during audits, for
example. These "kits" can of course also be used by attackers.
Negligence is not the result of any wilful or malicious action, but its effect can be similar to that
of an attack. It can create vulnerabilities that are difficult to detect, which can be exploited by
attackers or which can simply impair system availability.
For example, the involuntary modification of setting of the setpoint for regulation or
modification of an alarm may have disastrous consequences for the quality of products, services
provided, the environment or the health and safety of individuals.
The use of a USB stick – whether personal or not – to transfer data between isolated ICSs can
cause system unavailability if the stick is infected with a malwares.
In these two very concrete cases drawn from real-life experience, the individuals involved had
not intended to cause any harm. However, the impact on the ICSs was very real.
Such negligence may arise due to a lack of personnel training and information on the issues.
Vulnerabilities may have multiple origins and it is not the purpose of this guide to enumerate
them. Some examples of vulnerabilities frequently encountered in industrial systems are listed
in Appendix A.
The increasing need for consolidation of company data and access to it in real time from any
point on the planet, and the cutting of development and possession costs and planning
constraints have precipitated the convergence of the industrial and management IT fields.
Ethernet networks are now employed in ICSs and even as fieldbuses. They offer functionalities
such as a shared network infrastructure and the possibility of using IP layers (for example, for
remote maintenance).
Development, maintenance and remote maintenance are currently entirely developed on generic
platforms created from management IT (.Net or Java platforms, for example).
Systems standardisation and new functionalities have pushed vulnerabilities from the realm of
management IT to ICSs. The systems, referred to as proprietary, often lacking in security
Lack of training of users, cultural differences or lack of awareness of the risks associated with
cybersecurity may constitute a further significant vulnerability.
Numerous incidents on ICSs occur each year, but few receive media attention, such as the
incident with the nuclear power station in the United Kingdom linked to Conficker, the incident
associated with the Slammer worm in the USA, or, in 2010, the generalised propagation of the
Stuxnet3 worm. Their impacts may be analysed along a number of different lines, which are set
out below:
Impact on the A system failure after the malicious gaining of control can result in
environment system dysfunction (e.g. opening of pollutant containers) and
cause the site and its environment to become polluted. Such an
incident occurred in Australia in recent years.
3
Stuxnet is malware that targets ICSs. It exploits multiple vulnerabilities that exist in the Microsoft Windows operating
system and the SCADA WinCC software created by Siemens. The malware modifies the programme executed by certain
industrial PLCs from the Simatic S7 range manufactured by Siemens. The modifications made can lead to a slowdown in
production but also to the physical destruction of the plants run by the PLC.
These various impacts generate financial losses associated with the loss of activity or the
payment of compensation to potential victims (customers, individuals, local governments,
associations, States and even the European Union) and impair the image of the company.
4
See the ANSSI website: http://www.ssi.gouv.fr/rgs/
A large proportion of incidents are linked to lack of user awareness of the risks associated with a
system. Ensuring that they are aware of the rules governing "Healthy networks" will help reduce
vulnerabilities and opportunities for attack5. Awareness-raising must be consistent since the
risks are constantly evolving.
There is no point in trying to bolt a cybersecurity strategy onto an industrial system without any
prior understanding of the requirements of the core business that it serves. It is therefore
important to determine:
the business objectives (production, distribution, protection of assets and persons etc.)
and the services provided;
impacts in the event of service interruption;
functions vital to reach objectives, and specifically:
their degree of involvement and criticality in service provision,
the systems used to provide these functions,
whether these systems are centralised, distributed, remotely accessible, etc.;
An inventory of physical plants, systems and critical applications is a vital prerequisite for the
implementation of cybersecurity in industrial systems. This inventory is the first stage in risk
analysis, and makes it possible to determine the various different levels of criticality, safety,
availability and integrity that are required for the mapped items.
Any project must include a risk analysis in order to identify the sensitive elements within the
system, and their requirements and security objectives when faced with the threats identified.
These objectives are then broken down into security requirements, which pertain to the system
itself (intrinsic robustness), and its design, construction and operating environment. These
5
See the ANSSI website: http://www.ssi.gouv.fr/fr/bonnes-pratiques/principes-generaux/
They are clearly formalised in a security target, and constitute a benchmark for system security.
The risk analysis principles for cybersecurity are not different from those for dependability,
though if the terminology used may not be identical.
A number of risk analysis methods exist. ANSSI suggests the EBIOS 6 method.
The conclusions of the risk analysis lead to the definition of adequate security measures, for
which the effort is commensurate with the challenges and adapted to the real needs. These may
be technical but also organisational.
Software and embedded hardware intrinsically contain bugs and vulnerabilities. Some are
known and others are discovered over time. Attack surface must be reduced.
Adopting a defence in depth strategy provides protection against threats that are not yet known,
to reduce the perimeter on which a threat may directed, or to mitigate its impact.
Networks segregation with firewalls is not sufficient. Other mechanisms must be used alongside
this at different levels (e.g. physical access control, configuration hardening, antivirus
protection).
On legacy systems, automatic updates may be incompatible with the availability constraints and
antivirus software may disrupt the operation of business applications that are not designed for it.
Other mechanisms may be established such as operating system hardening and intrusion
detection. Multiple solutions exist. It is good practice to use barriers that are suited to the
system, and that do not negatively impact upon business objectives, and then to assess the
residual impacts.
It may also be useful to consult white papers and manufacturer's recommendations as well as
consulting the ANSSI7 website with regard to this area.
6
See the ANSSI website: http://www.ssi.gouv.fr/ebios/
7
See the ANSSI website: http://www.ssi.gouv.fr
Within an industrial environment, it can be complicated and even impossible to set in place
certain protective barriers without negatively affecting business functions. Countermeasures
must include detection and system surveillance mechanisms. Their operation must be
transparent so as not to disrupt any business functions.
Such measures will not prevent an incident but will enable its detection and limit its effects
as far as possible.
The earlier an incident is detected, the greater the possibility of implementing measures to
reduce and confine its effects, for example by:
physically isolating systems in the case of a virus attack so as to limit propagation
risks;
stopping a systems before it is degraded if the integrity of configuration data has been
compromised as a result of errors or intentional changes (this will, for example,
prevent systems destruction that could be caused by propagation of a worm such as
Stuxnet).
Finally, the collection of information from alarm and event logs is vital for subsequent analysis.
In some cases these logs may provide useful information and proof within the context of a legal
inquiry.
Incident management must also include a post incident (forensic) analysis phase to improve the
effectiveness of the cybersecurity measures initially implemented.
Keeping up to date on threat developments and vulnerabilities by identifying the incidents they
give rise to and their potential effects is a fundamental defence measure.
Websites such as that of the ANSSI9 operations centre or manufactures' website are important
sources of information on identified vulnerabilities, and on any existing corrective measures or
countermeasures that can be implemented.
Firmware updates for PLC and other devices, and fixes for operating systems and applications
are signalled by means of alerts and notices. RSS and Atom feeds often make information
rapidly available.
It may be worthwhile contacting manufactures to find out the best way of keeping up to date.
Also consider asking suppliers, contractually, to keep themselves up to date regarding
vulnerabilities. The more exhaustive mapping is, the more effective monitoring will be.
Preparing to deal with exceptional events against which all previous measures have failed will
minimise impacts and allow activity to be restarted as swiftly as possible.
Company Business Continuity Plans (BCP) must therefore include industrial control systems.
These include the definition of Disaster Recovery Plan (DRP), identifying the means and
procedures needed to return to a normal situation as quickly as possible, in the event of loss or
exceptional events. These should set out how to reconstruct the system following a virus attack,
fire, flood or loss of data.
8
This can be consulted on the CERT-FR website: http://www.cert.ssi.gouv.fr/site/CERTA-2002-INF-002/index.html
9
See the CERT-FR website: http://www.cert.ssi.gouv.fr/
Setting objectives that are appropriate for the priorities involved, formulating a strategy, raising
personnel awareness and providing training are all tasks that are incumbent upon the
management. The process can be progressive, carried out in a number of phases over time and
dealing with aspects ranging from the simplest to the most complex and from the most to the
least obvious, but it must be implemented system-wide. It may draw upon existing
cybersecurity standards taking into account the constraints specific to ICSs.
Industrial control systems must be included within a company information systems security
policy, as it is the case with any other information system, from the outset of the project.
Security issues are not specific to this field, but the implementation of solutions requires that
these be tailored to suit the industrial context.
Frequently, organisations do not promote increased proximity between conventional IT and ICS
environments. This is why a security implementation project for industrial control systems
cannot succeed without the involvement of the top management of the company.
System cybersecurity must be considered from the outset of the project by the end user who
must express their requirements.
In such cases, the company with responsibility for the future operating contract must carry out
an exhaustive inventory in terms of system security and of the means available to maintain it at
an acceptable level. This inventory must be accepted or be negotiated with the company
awarding the operating contract such that all parties are in agreement as to the cybersecurity
level for the "as built" plant, and the current means that it includes to maintain an acceptable
level of security. The outsourcing guide10 published by ANSSI may provide solutions to this
issue.
Finally, it should be remembered that changing passwords is vital upon operational transfer.
The integrity of a PLC safety program, for example, is now a priority in both the cybersecurity
and the safety fields. An attacker or a malware modifying a security program concerns both
fields. Stuxnet has shown that this type of scenario is entirely plausible.
The absence of cybersecurity or insufficient cybersecurity can therefore be the potential cause
of a failure mode in industrial system. Failure mode analysis is dealt with using dependability
10
See the ANSSI website: http://www.ssi.gouv.fr/externalisation
To cover all risks, cybersecurity and dependability must be dealt with together, using a joint
approach.
For example, the potential causes of a temperature increase at a plant above its nominal
threshold may be:
a reading issue linked to the failure of a sensor:
physical failure of a sensor,
incorrect calibration of the sensor,
an intentional change made to the parameters of a sensor by an unauthorised
person (gaining of control by a hacker, or a virus) or as a result of negligence;
a problem associated with a cooling circuit valve:
mechanical failure,
servo-motor failure,
intentional forcing of the command valve value by an unauthorised person
(gaining of control by a hacker, a virus) or as a result of negligence,
a problem with the setting of the setpoint for regulating the cooling system,
an input error made by an operator,
a change made to the setpoint by an unauthorised person.
FMECA and HAZOP type analyses are often complicated to carry out and time-consuming.
Including management information system and cybersecurity competence in teams conducting
these tasks will be much more effective than a stand-alone, piecemeal approach in which each
individual understands their own field, but is unable to see the big picture.
The cybersecurity of industrial system must be taken into account at the time the maintenance
plans are designed. These plans must include the operations that are necessary in order to
maintain the level of systems security in the long-term:
defining cybersecurity-specific maintenance operations that are necessary for
maintenance of operational conditions (MOC) and maintenance of secure conditions
(MSC): specifically, provision must be made for integration of fixes provided by the
manufacturer;
11
FMECA: Failure Modes, Effects and Criticality Analysis. Dependability and quality management risk analysis tools.
12
HAZOP : HAZard and OPerability study. Risk analysis method used in the field of dependability and safety.
Industrial control systems maintenance cannot be separated from the maintenance plans for the
plants that they command. Cybersecurity operations should be monitored in the plant
maintenance management tool (CMMS13).
The security requirements of the procured system must be studied and be clearly formalised (in
a security target or in the STC14) and included in calls for tenders as is the case for functional,
performance, quality, environmental, and safety requirements and also regulatory compliance
requirements.
These pertain to the system that is the object of the consultation and to management of the
project itself (training and accreditation of installers), including operating and maintenance
phases. It is therefore prudent to:
check in tender responses that the security requirements set out in the consultation are
met;
draft clauses for device maintenance:
request the maintenance plans needed in order to maintain the plant in an
operational and a secure condition,
define procedures for incident handling and provision of security fixes: who takes
the initiative, who implements, by which deadline, who carries out functional testing
and how, etc.;
specify clauses governing the terms of external service provision by subcontractors:
specify the terms of onsite support and intervention: is remote maintenance
accepted (if so, under which conditions)? can the service provider leave the site with
faulty devices and their configurations? can external providers use their own tools?
are external providers required to have specific qualifications?
determine the legal clauses to be included in contracts;
13
CMMS: Computerized Maintenance Management System
14
STC: Specific Technical Clauses
15
See the ANSSI website: http://www.ssi.gouv.fr/externalisation
16
MMC: Multi Media Card.
In many cases, good practices are similar to those that apply to management information
systems, although their implementation must be tailored to industrial sector constraints. They
are not ranked in order of priority and their ease of deployment will depend on the plant
concerned.
Method Identify who needs access to devices, why and how frequently.
Install servers in controlled access premises (where possible within IT rooms).
Place work station central units, industrial network devices and PLCs in
locked cabinets.
Protect access to the network cable and sockets.
Ways of Install a door “dry contact” to generate an alarm on the SCADA system upon
managing opening.
constraints "Break glass" type procedure
Ways of Filtering is applied upstream of these networks. Physical access to the process
managing network is limited and controlled.
constraints
Scope Work stations, servers, programming and maintenance console, touch screens.
Constraints It may be necessary to exchange data between networks that are not
interconnected.
Ways of The installation of clean machines - dedicated data transfer machines - may be
managing a way of meeting user requirements. These machines must be reinforced,
constraints regularly updated, have antivirus software installed and be under high
surveillance.
Method Define a management policy for user accounts and application accounts.
Do not leave default accounts on devices, machines and applications
(admin/admin).
Favour strong passwords (see http://www.ssi.gouv.fr/mots-de-passe, ->
calculate the "strength" of a password and the CERT-FR information note on
Ways of Trace actions performed with these logins to detect any diversions or
managing abnormal behaviour. Use strict organisational procedures (logbooks) in order
constraints to determine the identity of individuals using generic accounts on a moment-
by-moment basis.
Ways of If, ultimately, certain unsecured functionalities are needed, then an in-depth,
managing documented analysis (for example, exception handling) must provide
Method Activate traceability functions where devices and software permit this (syslog,
SNMP V3, "Windows Event", text file, etc.)
Select relevant events and organise event storage (volume, storage time).
Centralise logs and generate alerts for certain events or event outcomes.
Ways of Tools exist to help manage events and sort these according to predefined
managing criteria.
constraints
Method Define a back-up policy. Which data needs to be backed up in order to meet
the needs of users, reconstruct a plant following an incident or meet
regulatory requirements?
Constraints It is not always possible to automatically back up data, particularly for sensors
and actuators and for PLCs.
GP09: Documentation
Reason Control documentation in order to have an exact representation of the ICS and
avoid operating errors.
Control dissemination so that only those persons requiring information
receive it.
Ways of Make users aware of the risks associated with documentation. Leaving
managing documents in view on a desk or in a car boot (next to the duty computer for
constraints example) is not good practice.
Constraints Incompatible with certain older generation SCADA applications for example,
no antivirus updating mechanism (stand-alone machines for example).
Contractual issues such as loss of manufacturer's guarantee.
Method Define a management policy for patches (systematic, periodic or ad hoc) that
is suited to the functional constraints and risks identified. For example, define
priorities for deployment of patches, verify ascending compatibility, and
interoperability.
Systematically apply patches to engineering stations and portable machines.
Periodically apply patches to operator stations.
Apply patches during maintenance on sensitive plant.
Scope Production PLCs, PLC programmes that are backed up or under development
and companies
This document is a courtesy translation of the guide Cybersécurité des systèmes industriels:
Maitriser la SSI pour les systèmes industriels. In case of divergence, the French version pre-
vails.