Guest talk by Dr. Alexander Gallo
The increase of networks connections over which to close the loop in control systems, has exposed these cyber-physical systems to the effects of cyber-attacks: malicious agents can target communication channels and disrupt the controller’s performance. Diagnosis is a critical step towards establishing resilience in control systems. Of many diagnosis strategies that have been proposed in literature, several rely on actively perturbing the control system, thus changing its behavior, which leads to detection of even the most sophisticated attacks. In the context of multiplicative watermarking, this is achieved through the definition of time-varying parameters, which must be shared by both watermark generators and removers. In this talk, we address how these parameter switching functions can be designed to reduce the amount of shared knowledge necessary by different elements of the diagnosis scheme, and show how these can be designed to ensure that they cannot be predicted by malicious agent with system identification capabilities. We propose two methods, inspired by cryptography and systems theory, that ensure secure switching with limited shared information and guaranteed performance against eavesdropping attackers.
Alexander J. Gallo is currently a postdoctoral researcher at the Delft Center for Systems and Control at TU Delft, the Netherlands. He received his PhD as well as his MEng from Imperial College London in 2021 and 2016, respectively. He is interested in the design of distributed methods for resilient control of networked control systems, active diagnosis of cyber-attacks in cyber-physical systems, and control of systems under ageing. His main fields of application are energy distribution networks and wind turbines.