Blockchain-Enabled Hierarchical Federated Learning Framework for Anomaly Detection in IoT Systems

The rapid expansion of the Internet of Things (IoT) across domains such as industrial automation, smart healthcare, and intelligent transportation has intensified security challenges, particularly in terms of detecting anomalies across large-scale, heterogeneous networks. To address these challenges, this study introduces a blockchain-enabled hierarchical federated learning (Block-HFL) approach that combines federated model aggregation with blockchain-based authentication and immutable storage. This approach has enhanced scalability, reduced communication latency, and ensured trustworthy model management while preserving data privacy. In comparison with existing hierarchical and non-hierarchical FL approaches, the proposed Block-HFL framework introduces an accuracy-based leader election mechanism that enhances fairness and improves global model convergence. Experimental evaluations on the Edge-IIoTset dataset show that Block-HFL consistently maintains detection accuracy above 94% as the number of clients increases from 4 to 16, outperforming baseline FL models under similar non-IID conditions. Moreover, blockchain integration ensures secure, transparent, and tamper-proof global model management with minimal computational cost, confirming that the proposed framework provides an efficient and trustworthy solution for distributed anomaly detection in IoT systems.