Industrial internet of things fortify: multi-domain feature learning framework with deepdetectnet++ for improved intrusion detection

Abstract

The Internet of Things (IoT) connects more devices that have low user intervention requirements and can converse with each other. Intelligent Transportation (IoT) is a fast-growing field in computer science, but it's also vulnerable to various types of assaults due to the increasingly dangerous nature of the Internet. To secure IoT networks, the DeepDetectNet++ framework is proposed to identify and detect intrusions in IIoT systems and strengthen the IIoT system's security. The need for this research is informed by the fact that Intelligent Transportation IoT systems are becoming more exposed to complex cyberattacks that endanger the core functions of the systems. The goal is to design and implement a new generation intrusion detection system, DeepDetectNet++, that combines hybrid optimization and improved deep learning algorithms to increase the accuracy, sensitivity, and effectiveness of IIoT attack identification and categorization. Two IIoT datasets are used and they are preprocessed using outliers detection and missing values handling techniques. Moreover, the feature extraction phase extracts temporal features, flow-based features, and frequency domain features. A hybrid optimization strategy such as the Hybrid Pelican and Dragonfly Optimization (HPDF) technique is employed in the feature selection to identify the most discriminative features. Finally, a DeepDetectNet++ model is proposed to improve SKA-ResNet's model and Spatiotemporal Self-Attention (STSA)-Based LSTNet component to enhance the detection and classification performance of the developed model. The experimental results of the designed technique are validated with existing models and the developed model gained an accuracy of 98.3%, sensitivity of 97.75%, and F-measure of 98.3%. The developed model detects and classifies IIoT attacks accurately and with high efficiency.