A Survey of Blockchain-Enabled Hybrid Federated Learning for Trustworthy Intrusion Detection in IIoT Networks

Abstract

The Industrial Internet of Things (IIoT) has improved industrial automation and communication by connecting smart devices, sensors and machines, but the same connectivity has made industrial networks considerably easier to attack. Intrusion Detection Systems (IDSs) are central to protecting these environments, yet conventional centralised designs concentrate sensitive operational data in one place and introduce privacy risks, scalability limits and a single point of failure. Federated Learning (FL) addresses the privacy problem by training collaboratively without moving raw data, while blockchain technology supplies the trust, transparency and integrity guarantees that plain federated aggregation lacks. This research work reviews recent intrusion detection work based on machine learning, deep learning, federated learning and blockchain, with particular attention to studies evaluated on the ToN_IoT dataset. The analysis identifies model poisoning, weak trust management, class imbalance and resource constraints as the dominant unresolved issues. Guided by those gaps, a hybrid framework is proposed that combines PCA-based feature reduction, lightweight local IDS models, Federated Averaging (FedAvg) and Hyperledger Fabric-based trust validation, with the aim of delivering privacy-preserving, scalable and trustworthy intrusion detection for IIoT networks.