A Federated LearningBased Intrusion Detection Framework for Blockchain-Enabled Internet of Things (IoT) Supply Chains

Integrating Internet of Things (IoT) networks in supply chain operations implies the extension of real-time data possibilities along with augmenting the security challenges of cyber-physical nature. This paper proposed an interdisciplinary simulation framework that incorporates decentralized learning-trained deep neural networks for distributed anomaly detection on the IoT devices with a permissioned blockchain layer that secures anomaly recording. Several different simulation scenarios using NS3 and a blockchain simulator under cyber-attack scenarios such as data exfiltration, distributed denial-of-service attack, and malicious device behavior have shown that machine-learning networks do well in threat detection with over 95% accuracy while significantly reducing false positives relative to baseline methods. The blockchain layer contributes to a tamper-proof audit capability and increases supply chain data visibility and traceability for events. The unique contribution is captured in proposing an innovative federated-machine learning framework for detecting anomalies, a blockchain-driven storage layer connecting alerts and trained models to the system, and a comprehensive simulation study revealing an empirical superiority of the proposed system. However, the integration of federated machine learning in blockchain technology may foster response to incidents, adding single-tenancy separation to make the IoT data stream more trustworthy. This study helped in laying the groundwork for the realization of secure and intelligent IoT supply chain network deployments in real-world environments, providing them with solid security and transparency in operations.