Protection against cyberattacks in industrial control systems

Industrial control systems (ICSs) are cyber-physical systems that combine computational, communication and physical layers to perform a set of tasks. They are used to monitor and control various installations, services, processes, applications, and systems such as energy production and distribution (electricity, gas, etc.), transportation networks, and advanced communication systems. The use of ICSs is especially critical in the context of energy transition owing to the increasingly dense communication and exchange of data and information between consumers and utilities, and the services provided (micro-grids, demand side management, etc.). The role of ICSs thus becomes critical since they monitor the stability and reliability of energy production and distribution, optimize the energy consumption and production at centralized/distributed scales, increase the penetration of renewable energy into the grid, etc. However, ICSs are vulnerable to cyberattacks because they are designed to perform predefined tasks of production and ensure safety, but not to address security issues. These attacks modify controller programming to cause serious damage to physical equipment. The Stuxnet attack against Iran’s nuclear program is a typical example. It modified the program controlling the uranium centrifuges to make them spin too fast and for too long, causing their destruction. It is accordingly of prime importance to develop cyberattack detection systems that will spot these attacks early to stop them or limit their catastrophic consequences for critical physical equipment. This paper deals with the problem of cyberattack detection in ICSs in the context of energy transition. It is structured as follows. First the structure of a cyber-physical system and the components of its cyber and physical layers are presented, the different cyber-attacks made against the various components of a cyber-physical system are classified, and their impacts on the system integrity and performance are assessed. The approaches used to design an intrusion detection system are then listed. They belong to three main families: model-based, signature-based and anomaly detection approaches. The advantages and drawbacks of the approaches in each family are discussed according to the attack type and the information available to the attacker about the system structure. An example of a cyber-physical system composed of a pump, a valve, a tank, sensors, and a controller is used throughout to illustrate the different concepts and methods presented.