Detection of Insider Attacks in Distributed Projected Subgradient Algorithms

The gossip-based distributed projected subgradient algorithms (DPS) are widely used to solve decentralized optimization problems in various multi-agent applications, while they generally vulnerable data injection attacks by internal malicious agents as each agent locally estimates its descent direction without an authorized supervision. In this work, we explore the application of artificial intelligence (AI) technologies detect attacks. We show that a general neural network is particularly suitable for detecting and localizing agents, can effectively nonlinear relationship underlying collected data. Moreover, propose adopt one state-of-the-art approaches federated learning, i.e., collaborative learning protocol, facilitate training models gossip exchanges. This advanced approach expected make our model more robust challenges with insufficient data, or mismatched test simulations, least-squared problem considered verify feasibility effectiveness AI-based methods. Simulation results demonstrate proposed methods beneficial improve performance over score-based methods, peer-to-peer indeed target issues.