A Temporal Deep Learning Framework for Multi-Class Intrusion Detection in Industrial IoT Networks

The rapid proliferation of the Industrial Internet of Things (IIoT) has significantly expanded the cyberattack surface of connected industrial systems, creating an urgent need for intelligent, reliable intrusion detection mechanisms to safeguard edge computing environments. This paper proposes a deep learning-based Intrusion Detection System (IDS) that integrates Long Short-Term Memory (LSTM) networks with the Synthetic Minority Over-sampling Technique (SMOTE) to address the challenges of temporal traffic modelling and severe class imbalance in multi-class intrusion detection. The proposed framework employs sequential feature learning to capture temporal dependencies in IIoT network traffic while applying SMOTE exclusively to the training data to improve the representation of minority attack classes and prevent data leakage. Experimental evaluation on the Edge-IIoTset dataset, comprising more than 15 traffic classes, demonstrates that the proposed model achieves training accuracy of 86.24% and validation accuracy of 74.60%, with balanced precision, recall, and F1-score of 0.76, 0.74, and 0.75, respectively. These findings demonstrate that combining temporal feature learning with effective data balancing provides a practical and scalable approach for enhancing multi-class intrusion detection and supporting real-time cybersecurity monitoring in Industrial IoT environments.