Abstract

The proliferation of Internet of Things (IoT) devices has led to a surge in network traffic and potential security vulnerabilities, making intrusion detection systems (IDS) crucial for protecting IoT infrastructure. However, traditional centralized IDS approaches face challenges in handling the volume, velocity, and variety of IoT data, while also raising privacy concerns due to the collection and storage of sensitive network information. This paper proposes a federated learning (FL) framework for enhanced intrusion detection in IoT networks. FL enables collaborative model training across decentralized IoT devices without directly sharing raw data, thus addressing privacy concerns. The proposed framework utilizes a deep learning model trained in a federated manner to detect various types of network intrusions. We evaluate the performance of our approach using a simulated IoT network environment and demonstrate its effectiveness in detecting intrusions while preserving data privacy and achieving scalability. The results show that the FL-based IDS achieves comparable or superior performance to centralized approaches, while significantly reducing the risk of data breaches and complying with data privacy regulations. Finally, we discuss the challenges and future directions for applying FL in IoT security.