Abstract

Cyber-Physical System (CPS) is the term used to describe the physical systems equipped with computation and communication capabilities. CPSs can be used in different applications e.g. autonomous vehicles, water distribution systems, smart grids, industry 4.0 and Internet of Things (IoT). CPSs have expectation of improving the capability of traditional engineering system but on the other hand, they arise several concerns about their security against cyber-attacks. In the last decade, several cyber-attacks targeting SCADA systems have been reported, see e.g. Maroochy water breach and the Stuxnet worm aimed Iran's nuclear facility. From a control point of view, a CPS can be interpreted as a Networked Control System (NCS) where the risk of cyber-attacks can be modeled as the possibility that malicious agents could compromise the communication channels. In order to bene�t from CPSs, specially in safety critical systems, their vulnerabilities to cyber-attacks must be properly faced. In this thesis two control architectures for CPS are developed. In the first, starting from the analysis of active detection mechanisms available in the literature, we propose a novel architecture capable of detecting a broad class of False Data Injection (FDI) attacks. Such strategy has been contrasted with the well-known watermarking detection mechanism and it is shown that our solution is capable of detecting replay attacks without degrading the closed-loop performance of the system. Moreover, it is shown that compared to detection schemes resorting to auxiliary systems, the proposed strategy is less involved and of easier implementation. In particular, it can be installed on the existing NCS infrastructure without changing communications, controller or state estimator. In the second architecture, we propose another novel architecture capable of detecting and mitigating a broad class of FDI attacks. First, we propose a detection mechanism based on a coding scheme to limit the attacker's disclosure and disruptive resources and prevent the existence of stealthy attacks. Second, we propose an emergency local controller that is activated when an attack is detected or the plant's safety is in danger. It is proved that the proposed architecture always guarantees the safety of the system, regardless of the attack actions and detector performance. Moreover, plant's normal operation recovery is ensured once the attack is terminated.

Available Versions of this Item

• Cyber-Attack Detection and Mitigation in Networked Control Systems. (deposited 30 Jun 2021 15:04) [Currently Displayed]
  • Cyber-Attack Detection and Mitigation in Networked Control Systems. (deposited 27 Oct 2022 13:52)