Attack - Resilient H 2 , H ∞ , and ` 1 State Estimator

Due to its distributed nature, a cyber-physical system is vulnerable to various faults, including sensory integrity attacks. Such faults need to be accounted for in the design of a state estimator. In this paper, we consider sparse sensor faults, in which a small unknown group of sensors can be compromised. We first show a necessary condition that allows the state to be estimated in the presence of sparse sensor faults. We then propose an estimator that is resilient to such faults assuming the necessary condition alone. The proposed estimator requires fewer sensors than existing estimators and avoids potential losses in performance due to delays. Moreover, the proposed estimator’s worst-case estimation error in the two/infinity/infinity norm is given for the two/two/infinity norm bounded input disturbance. Alternatively, this worst-case bound can be considered an extension of robust control to that takes into account a sparse-unbounded input. This extension holds potential interest for a broader audience.