Intelligent Cyber Risk Assessment Framework for Industrial IoT using Ensemble Learning and Explainable AI

Abstract

Industrial 4.0 is supported by the IIoT, which can enable increased automation, efficiency, and real-time monitoring of industrial control systems. Regardless of these benefits, the rise of connectivity makes the IIoT systems susceptible to some of the greatest cybersecurity threats, which undermine confidentiality, integrity, and availability. This paper presents a ML-based model of cyber risk assessment designed to identify a threat in IIoT proactively. Supervised models, such as Multi-Layer Perceptron, XGBoost, LightGBM, RF, LR, KNN, DT, SVM, FSVM,FXGBoost, and ensemble voting classifiers are fully evaluated. According to the experimental findings, the ensemble voting-based model is more effective as it reaches a higher accuracy of 99.3 percent, with a higher F1-score of 0.993, compared to the isolated learners. Using eAI methodologies, such as LIME and SHAP, to explain the impact of factors on the prediction improves transparency. The trained model is then deployed using the Flask web framework, enabling real-time risk assessments and interactions with users. The system categorizes the IIoT cyber risk levels as Very Low, Low, Medium, High and Very High which are immediately mitigated. The proposed solution provides an effective, interpretable and scalable treatment, and enhances the cybersecurity resilience of the IIoT.