Attack detection and consensus control for multi-agent systems under distributed replay attacks

This paper investigates the attack detection and secure consensus problem of multi-agent systems under distributed replay attacks. Attacks on the communication links between agents can result in performance degradation and inability to achieve consensus. To address this challenge, first, a novel distributed replay attack model is proposed, in which duration and frequency constraints are imposed on the attack. Unlike traditional models, this proposed model accounts for a more complex and practical scenario where the adversary launches independent attacks on each communication channel, and the attack topology pattern is time-varying. Second, a detection scheme based on moving target defence and dynamic encryption-decryption is proposed to address the challenge of undetected attacks in specific scenarios. This scheme ensures a high detection probability without compromising system performance. Then, a fully distributed attack-detection-based secure controller is designed to mitigate the impacts of attacks and ensure secure consensus. The relationship between attack duration and system stability under different attack patterns is investigated, and a sufficient condition for attack duration that allows the system to reach consensus is derived. Finally, the effectiveness of the proposed method is verified through simulation experiments.