Fed-FeRe: An Enhancing Approach for Efficient Anomaly Detection in IoT Security

The proliferation of the Internet of Things (IoT) significantly enhances the complexity of device interactions within these networks, elevating susceptibility to cyber threats. Anomaly detection is crucial in defending IoT systems against these threats while preserving user privacy. In this study, we introduce a novel federated learning-based approach, named Fed-FeRe, which integrates federated learning with
a Chi-Square-based feature reduction technique and a Gated Recurrent Unit (GRU) model to address anomaly detection in IoT networks. This integration improves the accuracy of anomaly detection and reduces the computational burden and communication overhead in scenarios with non-independent and identically distributed (non-IID) data, typical in IoT environments. Our comprehensive evaluations demonstrate that our approach significantly enhances detection accuracy while reducing communication and computational demands, affirming the potential for real-world applications. This paper contributes to advancing machine learning techniques in enhancing IoT security, offering a robust, decentralised anomaly detection method that ensures data privacy across diverse IoT environments.