Resilient and secure remote monitoring for a class of cyber-physical systems against attacks

Abstract

This paper is concerned with the resilient and secure remote monitoring of a cyber-physical system of a discrete time-varying state-space form against attacks. The specific statistical characteristic, magnitude, occurring place and time of the attack signals are not required during the monitor design and attack detection procedures. First, an optimal ellipsoidal state prediction and estimation method is delicately developed in such a way that the recursively computed prediction ellipsoid and estimate ellipsoid can both guarantee the containment of the true system state at each time step regardless of the unknown but bounded input signal. It is expected that the two ellipsoids can resist certain attacks as the calculated state prediction and state estimate are sets in state-space rather than single pointwise vectors, thus potentially enhancing the resilience of the remote monitoring system. Second, a set-based evaluation mechanism in combination with a remedy measure are proposed to provide timely detection of certain attacks. Furthermore, a numerically efficient algorithm is established to achieve resilience and attack detection of the remote monitoring system. Finally, it is shown through several case studies on a water supply distribution system that the proposed methods can provide quantitative analysis and evaluation of the potential consequences of various attacks on the remote monitoring system.