Community-Based Distributed Flight Cluster Malicious Node Detection Based on Neural Network

Abstract

A flight cluster refers to a group of unmanned aerial vehicles (UAVs) that can work together to perform complex tasks through specific organizational and control protocols. The flight state of each node of a flight cluster under distributed control depends on its flight state and neighbors. When a node in a flight cluster is attacked, it becomes a malicious node, which can use the distributed control protocol to launch a variety of attacks that cause the flight cluster to crash. To overcome such fatal weakness, we propose a peer-to-peer network model for the flight cluster networks based on the flight cluster characteristics that include the cluster control relationship topology network and the cluster communication link topology network. A malicious node detection scheme for flight clusters based on community partitioning is presented, which fully considers the flight cluster control topology and communication topology. Then, a novel malicious node detection model for the flight cluster based on neural network is given. Finally, the experiment is performed under the virtual cluster simulation environment. The results show that the scheme is low complexity, feasible, and effective in improving the operational efficiency of the detection model.