Enhancing IIoT Cyber-Attack Detection: An Improved MobileNets Model with Adaptive Recursive Feature Elimination

Abstract

In the era of advanced Industrial Internet of Things (IIoT) cyber-attacks, the need for improved detection models is crucial. This research presents an enhanced MobileNets model specifically designed for advanced IIoT cyber-attack detection. To achieve higher efficiency and accuracy, an adaptive recursive feature elimination (ARFE) strategy is proposed for effective feature selection. Through iterative elimination of less relevant features, the predictive performance of the model is optimized. To ensure robustness and generalizability, the proposed approach is trained and validated on six diverse, real-world IIoT datasets: UNSW-NB15, CICIDS2017, RPL-NIDDS17, N_BaIoT, NSL-KDD, and BoT-IoT. The evaluation of the proposed model on these datasets demonstrates its effectiveness in detecting cyber-attacks in various IIoT environments. The findings of this research contribute to the advancement of cyber-attack detection techniques in the context of IIoT, paving the way for enhanced security in industrial systems.