Resilient Distributed State Estimation Under Stealthy Attack

Abstract

Network security issues are significant for cyber-physical systems while existing wireless networks are vulnerable to cyber-attacks. This paper proposes a novel detector for stealthy attacks in distributed consensus-based filtering, which has fewer objects to detect and thus can reduce the time consumption of the detector. Based on the proposed detector, we can analyze the estimated performance of the distributed estimator under a stealthy attack. The purpose of this study is to analyze the properties and performance of the estimator equipped with the proposed detector depending on the adversary's available resources. First, we study the sufficient condition for distributed state estimation to be resilient based on the assumption that the adversary has sufficient resources. With limited adversary resources, the optimal Kalman gain can be obtained by minimizing state estimation error covariance. Furthermore, a sufficient condition is presented to guarantee the convergence of the estimation error covariance in the estimator equipped with the proposed detector. Finally, a simulation example is provided to demonstrate the effect of the detector on the distributed estimator.