Federated Learning Approach for Tracking Malicious Activities in Cyber-Physical Systems

Abstract

The fast rise of the Internet and advanced technologies causes an increase in network traffic, making network infrastructure increasingly complicated and varied. Mobile phones, wearable gadgets, and driverless cars are all instances of dispersed networks that create massive amounts of data every day. The processing capability of these devices has also increased steadily, necessitating the need to transport data, store data locally, and direct network calculations to edge devices. Intrusion detection systems are essential in guaranteeing the safety and confidentiality of such equipment. Deep Learning (DL) combined Intrusion Detection Systems (IDS) have gained prominence due to their excellent categorization accuracy. However, the requirement to store and communicate data to a centralized server may jeopardize privacy and security concerns. Federated learning (FL), on the other hand, fits in nicely as private information decentralized learning approach that does not transport data but instead trains algorithms locally and sends the parameters to a centralized server. This work targets to offer an extensive overview of the FL in intrusion detection systems.