Review on the Use of Federated Learning Models for the Security of Cyber-Physical Systems

Abstract

The field of critical infrastructure has undergone significant expansion over the past three decades, spurred by global economic liberalization and the pursuit of development, industrialization, and privatization by nations worldwide. This rapid growth has led to a proliferation of critical infrastructure across various sectors, necessitating decentralization efforts to manage the associated burdens effectively. With the advent of artificial intelligence and machine learning, computer scientists have sought innovative approaches to detect and respond to the evolving landscape of cyber threats. Despite efforts to subscribe to these changes, attackers continually devise new methods to evade detection, requiring constant vigilance and adaptation from cybersecurity professionals. Traditional centralized models of machine and deep learning demand substantial data and computational resources, making them susceptible to single-point failures. To address these challenges, scientists have introduced federated learning—a decentralized technique that minimizes computational costs while prioritizing data privacy and preservation. This review article delves into recent research and review papers concerning critical infrastructure security and federated learning, exploring various architectures, threats, vulnerabilities, and attack vectors. Through our analysis, we provide a comprehensive overview of federated learning, cyber-physical systems security, and the advantages of integrating federated learning into critical infrastructure environments. By synthesizing insights from diverse sources, our study contributes to a deeper understanding of federated learning's applications and implications in safeguarding critical infrastructures. We highlight the potential of federated learning to enhance cybersecurity measures while addressing the unique challenges posed by modern-day threats. As organizations and nations navigate the complexities of securing their critical assets, the adoption of federated learning emerges as a promising strategy to bolster resilience and protect against emerging cyber risks.