Edge-Deployed Clinical Monitoring and Intrusion Detection Secured by Hyperledger Blockchain Decentralized Storage in Medical IoT

Abstract

Medical Internet of Things (MIoT) devices are vital to the functioning of intensive care units, allowing patient vital signs to be transmitted in real time but they are also a double-edged sword: network-level attacks can lead to the integrity of the data being compromised and delayed or failed clinical action can occur if a critical condition is not detected. This paper addresses both risks with a single edge-deployed framework called BlockEdge-Health. A Histogram Gradient Boosting (HGB) classifier trained on MIMIC-IV vital signs identifies critical patient states at the edge, while a K-Nearest Neighbors (KNN) intrusion detector trained on NSL-KDD guards the network layer. Every validated decision is committed to a Hyperledger Fabric permissioned ledger through a smart contract, and bulk medical files are stored on IPFS with cryptographic hash anchors held off-chain. On MIMIC-IV, the clinical module achieved 97.8% test accuracy and a cross-validated ROC-AUC of 0.998. The KNN detector reached 97.6% accuracy across ten cross-validation folds (97.82% precision, 97.97% recall, 97.89% F1, all with standard deviations below 1%). The Fabric network ran at 98 transactions per second under baseline load and around 145 TPS at steady state, with average latency dropping from 0.48 s during initialization to 0.31 s after peer synchronization. IPFS hash verification held at 100% accuracy, and storage overhead was 12.8 MB per 1,000 transactions, which is substantially less than the 18 to 20 MB reported by Ethereum-based comparison systems. The results show that a carefully designed edge architecture can serve both the clinical and cybersecurity demands of an ICU without cloud dependence and without trading one property against the other.