Lightweight data security system for resource-constrained IIoT devices

The rapid expansion of the Industrial Internet of Things (IIoT) introduces significant security challenges, particularly for devices with limited computation, memory, and energy. This paper presents a lightweight federated learning(FL) intrusion detection framework integrated with a resource-efficient blockchain layer called PureChain. A trimmed 1D CNN–LSTM model is designed to minimize complexity while maintaining high detection performance, requiring only 41,762 parameters and a 0.17 MB footprint. Experiments on the CIC-Collections2023 dataset and an internally generated IIoT dataset demonstrate strong generalization with accuracies of 97.7% and 96.84%. Five-fold cross-validation confirms stable learning with low variance, and resource overhead measurements show 3.1-ms inference latency and 0.42-mJ energy cost, confirming suitability for constrained IIoT nodes. PureChain provides tamper-evident model validation with a gas cost of 65,618 gas and achieves a 98% contract security score following formal auditing. A structured threat model outlines adversaries, system assumptions, and mitigation mechanisms. Overall, the proposed architecture offers an efficient, secure, and deployable solution for next-generation IIoT intrusion detection systems.