Improving Security and Privacy in Advanced Federated Learning Environments for Cyber-Physical Systems

Ensuring protection against a wide range of complex and advanced threats is crucial in the continually changing field of cyber-physical systems. This article examines the essential security and privacy elements in the context of novel service delivery models, explicitly focusing on federated learning as a fundamental paradigm. Our study aims to investigate the implementation of a synchronous reinforcement learning (RL) agent in a federated learning setting. Our purpose is to create resilient allocation strategies for cyber-physical systems. By engaging in synchronized training, RL agents establish efficient communication channels, enabling them to acquire a versatile policy that can effectively respond to many forms of assault. Our technique utilizes synchronous reinforcement learning in a federated learning context to create robust and adaptive systems prioritizing data protection in continuous data exchange settings. This study signifies a notable advancement in strengthening cyber-physical systems against developing threats, in line with the conference's focus on security and privacy for innovative service delivery models.