Federated Learning-Driven Collaborative Protection of Privacy and Security in Distributed Autonomous Control Systems

Abstract

Distributed Autonomous Control Systems (DACs), the core infrastructure of modern industrial production, intelligent transportation and emergency response, face dual risks of data privacy leakage and malicious attacks due to their open communication and distributed collaboration. Traditional separate protection schemes cause resource conflicts and performance trade-offs, failing to meet high-reliability demands in safety-critical scenarios. This paper proposes the FL-PSCP framework driven by federated learning, integrating local differential privacy, secure multi-party computation, abnormal detection and trusted node authentication. It designs privacy-enhanced federated aggregation, federated multi-dimensional attack detection and dynamic trust evaluation. Experiments on smart grid control, multi-robot rescue and autonomous ship navigation show FL-PSCP raises attack detection rate by 19.3% on average, cuts privacy leakage risk by 78.5% and keeps average control error within 5%, providing an integrated privacy and security protection solution for DACs and boosting their safe application in complex environments.

Next article in issue