Optimal DoS attack policy against remote state estimation

Abstract

Security of Cyber-Physical System (CPS) has gained increasing attention in recent years. Most existing works mainly investigate the system performance given the attacking patterns. In this paper, we investigate how the attacker should design its DoS attacking policy so that the system performance can be deteriorated as much as possible. Specifically, we consider the scenario where a sensor sends its data to a remote estimator through a wireless channel for state estimation, while an attacker decides whether to jam the channel at each sampling time. For two typical system performance indexes, i.e., the expected average estimation error and the expected terminal estimation error, we construct optimal attack scheduling schemes, respectively. We also propose the optimal attack schedules to avoid the given intruder detection mechanism. Numerical examples are presented to demonstrate the effectiveness of our results.