Transparent Ensemble Deep Learning for Intrusion Detection in Industrial Internet of Things

The growing complexity and scale of Industrial Internet of Things (IIoT) networks have made them increasingly vulnerable to cyber-attacks, underscoring the need for reliable and precise Intrusion Detection Systems (IDS). However, many existing IDS solutions face challenges such as high false-positive rates, limited detection accuracy, and a lack of transparency in their decision-making processes, which hinders their practical application. This work presents an advanced IDS based on ensemble deep learning specifically designed for IIoT networks to overcome these limitations. The suggested system integrates a number of deep learning models, such as Artificial Neural Networks (ANNs), Convolutional Neural Networks (CNNs), and LASSO, to harness their unique advantages and enhance detection capabilities. Moreover, the system integrates explainability methods, such as Random Forest, linear regression, and Local Interpretable Model-agnostic Explanations (LIME), to offer clear and interpretable perspectives on the process of making decisions. Intrusion Detection System (IDS) proposed in this study is trained and tested using the ToN_IoT dataset, a comprehensive IIoT dataset that encompasses diverse network traffic data and various attack scenarios. Experimental findings indicate that the ensemble model notably improves detection accuracy while lowering false-positive rates compared to traditional baseline models. The incorporation of Random Forest, linear regression, and LIME enhances the system's transparency by helping to elucidate the role of individual features in the forecasts of the model, thereby making the system more interpretable additionally reliable. This research advances the field by introducing a scalable, precise, and explainable IDS for IIoT networks, with strong potential to enhance cybersecurity in industrial settings.