5. Detection and Localization of Internal Malicious Attacks in Distributed Computing Networks

Figure 2: Distributed IoT networks exhibit distinct KOL network characteristics when infiltrated by internal malicious nodes

Project Description: Multi-agent algorithms for distributed computing are crucial for solving high-dimensional data processing problems, and as technology transitions from autonomous systems to swarm intelligence simulations, these algorithms are expected to gain greater importance in robotic applications requiring information fusion. Although fault tolerance is built into most instances of distributed information processing algorithms (especially those that work in a peer-to-peer manner), these methods are threatened by data injection attacks because they rely on consensus constraints to ensure convergence with the network. Global optimum, representing inference or decision-making. It turns out that attackers can exploit the consensus mechanism to steer the network to unwanted end results. The purpose of this project is to provide guidelines on how to design defense of multi-agent algorithms and improve the network's immunity to various attacks.