Distributed Detection of Covert Attacks for Interconnected Systems

The problem of covert attacks detection in a network of interconnected subsystems is addressed in this paper. Existing approaches in the area of covert attacks detection have been devoted to centralized systems and are mainly based on mismatch between the attacker's model and the actual plant. Instead, in this paper, we consider a large-scale system where the attacker has full knowledge on the subsystems models. By using the information received from neighboring subsystems and by exploiting the mismatch between a distributed Luenberger observer and a decentralized unknown input observer, we propose a local detection strategy allowing each subsystem to detect anomalies in its neighborhood. The effectiveness of the proposed strategy is shown in a numerical example.