Deep Learning Based Stacked Recurrent Neural Networks for Intrusion Detection in Industrial Control Systems Using Bio Inspired Meta Heuristics

An integral component of the transformational shift from traditional industries to intelligent manufacturing sectors is the Industrial Internet of Things (IIoT). The IIoT improves the productivity and functionality of industries by facilitating information mining, collection, and supervisory systems through the Internet-based connectivity of sensing gadgets and machinery equipment. The dynamic, large-scale, heterogeneous, and complex IIoT infrastructure makes identification of intrusions a guaranteed means of improving the IIoT’s efficient operation. In the IIoT context, anomaly detection techniques are frequently used to achieve security. The latest developments in deep learning (DL) models enable the creation of efficient anomaly detection methods. This study creates the DLAD model, a bio-inspired metaheuristic mechanism driven by DL techniques to identify and recognize intrusions in the context of IIoT. The objective of the suggested DLAD technique is to efficiently detect and categorize anomalies within the IIoT environment. In order to do this, Improved Crow Search Algorithm (ICSA) is derived using the DLAD model in order to extract a relevant subset of features. Furthermore, the identification and classification of abnormalities is carried out by the application of the Stacked Recurrent Neural Networks (SRNN) method. Ultimately, the Harris Hawks Optimizer (HHO) is employed to get the best possible tuning of the SRNN model’s parameters. Numerous simulations were run to ascertain the improved detection results of the suggested model on NSL-KDD dataset. The efficiency of the proposed model is 98.3% which is superior compared to the other prevailing works in the literature.