An Intelligent Hybrid Deep Learning Framework for Intrusion Detection in 5G-Based IIoT Networks

Abstract

The rapid development of 5G-based Industrial IoT (IIoT) networks has increased their vulnerability to cyberattacks, especially intrusion attempts and botnets. To address the need for an efficient and accurate real-time intrusion detection system, this paper presents an intelligent hybrid deep learning framework based on a combination of deep neural networks (DNNs) and random forests (RF). For the purpose of analyzing high-dimensional characteristics of IIoT data, the suggested system makes use of sophisticated data preprocessing, feature selection via the use of SelectKBest, class imbalance correction through the use of SMOTE, and deep learning methods for hidden-layer expression. The model was evaluated on a state-of-the-art and realistic IIoT cybersecurity dataset, the Edge-IIoTset, which includes 61 highly correlated features spanning 14 attack types. Experimental results show that the proposed hybrid model is more effective than traditional machine learning techniques (DT, RF, SVM, KNN) and a single DNN model. This model meets the real-time requirements of IIoT applications with 99.20% accuracy, no false alarms, and extremely low inference latency of 0.03 to 0.05 milliseconds per sample. This approach makes 5G IIoT networks more secure, faster, and more adaptable. The results help ensure the security of 5G IIoT networks for use in critical areas such as smart factories, cyber-physical systems, power grids, and industrial automation.