Abstract

The increasing number of smart devices in the Internet of Things (IoT) has led to a surge in cyber- attacks, posing significant security risks to personal and industrial systems. Traditional centralized intrusion detection systems (IDS) often face challenges related to privacy, scalability, and the heterogeneity of data generated by smart devices. This paper introduces a novel approach, Federated Learning (FL) for smart device security, to address these challenges. By utilizing FL, multiple smart devices can collaboratively learn a shared intrusion detection model without exposing their raw data, ensuring privacy while improving detection accuracy. This paper presents the architecture of a federated intrusion detection system (FIDS), explores various FL algorithms suitable for smart device security, and evaluates the performance of the proposed system using real-world IoT datasets. The results show that the federated approach achieves comparable or even superior performance in detecting intrusions while preserving data privacy and reducing communication overhead. This paper concludes that federated learning has the potential to revolutionize smart device security by enabling decentralized, scalable, and privacy-preserving intrusion detection systems.