Attack on communication links analysis and detection over the clustered networks

This research focuses on the subject of attack analysis, detection, and mitigation in a network of MASs with linear dynamics. This network is divided into multiple clusters that are disconnected from one another. A leader is an agent that exists in each cluster and has the ability to interact with other leaders through a strongly connected directed graph at discrete instances. Firstly, we analyze the effect of communication link attacks on the consensus algorithm. We show that the consensus control law may be disturbed by a single attack on a communication channel using principles from matrix theory and graph theory. Then, a bank of impulsive unknown input observers is established to ensure each agent is able to detect the communication link attack. Moreover, a modification of the consensus algorithm of MASs based on an estimated attack is also given. Finally, the results of our research are illustrated using simulations.