Worst-Case Stealthy False-Data Injection Attacks on Remote State Estimation

This paper studies the problem of false-data injection attacks on remote state estimation. In contrast to existing work that presupposed linear attack models, the optimal information-based attack policy that can cause the maximum estimation quality degradation and deceive the interval chi(2) detector is obtained. The scenarios that attackers have different information sets from the remote estimator are studied in a unified framework. It is shown that, with the given information set, there does not exist another attack policy outperforms the proposed one. The result in this work reduces to the optimal innovation-based linear attack in a special case. The optimality of the information-based strategy is verified by theoretical analysis and numerical examples.