Secure State Estimation of Cyber-Physical Systems Under Switching Attacks

This paper deals with secure state estimation against switching mode and signal attacks on cyber-physical systems, possibly affected by adversarial extra fake measurement injection. A stochastic Bayesian approach is undertaken by exploiting Bernoulli and Poisson random sets for modeling the attack existence and, respectively, fake measurements, as well as multiple models for handling the various attack modes. A Gaussian mixture implementation of the resulting random set Bayesian filter is presented and a simulation case-study concerning an electrical power network is worked out in order to demonstrate the effectiveness of the proposed approach.