Robust Filtering of Distributed Cyber-Physical Systems with Cyber-Attack Detection

We consider the distributed filtering problem for a large-scale system with a set of spatially connected sub-systems subject to randomly occurring deception or/and denial-of-service (DoS) attacks during the data transmission among subsystems. In this work we introduce a cyber-attack detection (CAD) model for each measurement by introducing an indicator variable with a beta-Bernoulli prior. The state of interest as well as this indicator are then inferred by a mean-filed variational Bayesian method in an iterative manner. Simulation results carried out on a standard IEEE 39-bus power system demonstrate the superior performance of the proposed method in the presence of deception or/and DoS attacks.