Resilient State Estimation and Attack Mitigation in Cyber-Physical Systems

Smart and Cyber-Physical Systems (CPS), e.g., power and traffic networks and smart homes, are becoming increasingly ubiquitous as they offer new opportunities for improved performance and functionalities. However, these often safety-critical systems have also recently become the target of cyber or physical attacks. This chapter contributes to the area of CPS security from the perspective of leveraging the knowledge of physical system dynamics as an additional “sensor” to mitigate the effects of false data injection attacks on actuator and sensor signals as well as attacks on the switching mechanisms, e.g., circuit breakers, on the state estimation and control algorithms in these systems, in order to ensure continued safety, reliability, and integrity of systems despite attacks. Specifically, we consider physical models with switched linear dynamics subject to two classes of uncertainties: A) stochastic uncertainty (aleatoric), i.e., with (unbounded) stochastic process and measurement noise signals, and b) set-membership uncertainty (epistemic), i.e., with distribution-free bounded-norm process and measurement disturbances. In both settings, we model the system under attack as a hidden-mode switched linear system with unknown inputs (attacks) and propose multiple-model inference algorithms to perform attack-resilient state estimation with stability and optimality guarantees. Moreover, we characterize fundamental limitations to resilient state estimation (e.g., upper bound on the number of tolerable signal attacks) and discuss the topics of attack detection, identification, and mitigation under this framework. Simulation examples of switching and false data injection attacks on a benchmark system and an IEEE 68-bus test system show the efficacy of our approach to recover resilient state estimates as well as to identify and mitigate the attacks in the presence of stochastic and set-membership uncertainties.