Cloud-CPS Security at the Edge: A Federated Learning Approach
Prema Kumar Veerapaneni, Akshay Nagpal, Koushik Kumar Ganeeb, Vidyasagar Parlapalli, Darshan Mohan Bidkar, Gaurav Mehta, Bindu Mohan Harve · 2025
The integration of Cyber-Physical Systems (CPS) with cloud computing—forming Cloud-CPS—has enabled intelligent, scalable, and autonomous system operations across domains such as smart grids, industrial automation, and healthcare. However, this convergence introduces critical security and privacy challenges, particularly when handling sensitive data at the edge. This paper presents a federated learning- based security framework that enables collaborative threat detection across distributed edge nodes without transmitting raw data. By leveraging a multi-tier edge-cloud architecture and lightweight anomaly detection models, the proposed system enhances security while preserving data privacy and reducing communication overhead. Experimental results on benchmark datasets demonstrate that our approach achieves high detection accuracy, low latency, and robustness to non-IID data, making it suitable for real-time CPS environments. This work establishes a foundation for privacy-preserving, scalable, and intelligent security in next-generation Cloud-CPS infrastructures.