A high-performance intrusion detection model for IIoT networks using BBO-based feature selection and deep fusion

Industrial Internet of Things (IIoT) environments are increasingly exposed to sophisticated cyberattacks due to their heterogeneous architectures, large-scale deployments, and distributed communication mechanisms. Conventional intrusion detection systems (IDS) often exhibit limited effectiveness in accurately and efficiently detecting multi-class attacks, particularly under real-time operational constraints. To overcome these limitations, this paper presents a novel intrusion detection framework that integrates Biogeography-Based Optimization (BBO) for optimal feature selection with a deep learning–based fusion model comprising Convolutional Neural Networks (CNN), Deep Neural Networks (DNN), and Bidirectional Long Short-Term Memory (BiLSTM) networks. The proposed framework is evaluated using the Edge-IIoT dataset, which encompasses 15 distinct attack classes designed to emulate realistic IIoT attack scenarios. The BBO-based feature selection mechanism significantly reduces feature dimensionality while preserving critical discriminative information. Furthermore, the hybrid fusion model effectively exploits spatial, temporal, and high-level abstract representations of network traffic data. Experimental results demonstrate that the proposed approach achieves an accuracy of 99.50%, thereby outperforming existing state-of-the-art IDS models. In addition, an extensive ablation study and statistical validation using analysis of variance (ANOVA) confirm the robustness, reliability, and statistical significance of the proposed intrusion detection framework for multi-class attack detection in IIoT networks.