Resilient Filtering of Nonlinear Complex Dynamical Networks Under Randomly Occurring Faults and Hybrid Cyber-Attacks

This paper investigates the problem of recursive resilient filtering for discrete-time nonlinear dynamical networks with randomly switching topologies under randomly occurring faults and hybrid cyber-attacks. The attacks considered are composed of DoS attacks describing by binary random sequences and deception attacks, the amplitude of which obeys the Gaussian mixing distribution. A recursive resilient filter dependent on DoS attack sequences is designed to simultaneously estimate system states and occurred faults, where the desired gain is parameterized via the matrix iterative formula of filtering error covariance. Furthermore, the maximum posterior probability estimation is calculated to identify DoS attack sequences with the help of the classical Bayesian rule. In particular, a modified recursive filtering algorithm is established by replacing the corresponding parameters in the previously derived formulas about filter gains and covariance matrices. Finally, the effectiveness of the proposed recursive filtering algorithm is demonstrated by a simulation example.