Reliable Secured Consumer IIoT Framework With Multi Layer Attack Interpretation and Prevention

Sustainable development and evolution of Industry 5.0 paved the way for the commercial and technical enhancement of the Industrial Internet of Things (IIoT) sensors. They are challenged by various factors such as privacy, authentication, security, data processing, and sustainability requirements. The proposed work is developed based on the multi-class classification models of Artificial Intelligence (AI) to handle security attacks and vulnerabilities as Intrusion Detection System(IDS). The proposed model is built with various Machine Learning (ML) models such as Decision Tree, Random Forest, Multinomial Naive Bayes, and Gradient Boosting Classifier. The Gradient Boosting Algorithm, compared with the other models, provided the best accuracy of 0.94, the classification probability of which is used in the development of the Local Interpretable Model-Agnostic Explainer (LIME). The random forest model, which provided the next best accuracy of 0.93, was applied for the explanation of the global surrogacy with the SHapley Additive exPlanations (SHAP) Explainer. These two models interpret the feature relationships, weights, and influence on the target estimation in both the local and the global perspective. In comparison with the existing research, the proposed framework provides an increase of around 1% in accuracy, 1.2% in precision, 1.1% in recall, and 1.14% in f1-score. Explainable Artificial Intelligence (XAI) enhances trust and reliability in AI predictions by providing a clear explanation of how the model detects attacks, making it reliable to understand, trust, and apply the predictions in real-time IIoT environments.