Attack Detection and Distributed Filtering for State-Saturated Systems Under Deception Attack

In this article, a secure distributed filtering problem is investigated for a state-saturated system subject to random occurring nonlinearities and deception attack. To resist the hostile attack from a malicious attacker who can modify the data of the transmission module according to a certain probability, an attack detector with a dynamic threshold is designed for each sensor to decide whether to broadcast the current measurement innovation to neighbors. A distributed filtering algorithm is proposed to estimate system state, and a sufficient condition is established to guarantee mean-square boundedness of the estimation error. Moreover, a critical attack probability above which the steady-state estimation error covariance will exceed a preset value is found. In addition, a minimum attack magnitude is given to ensure that the attack signal is able to be completely detected. A numerical example is carried out to testify the validity of the proposed results.