A Hybrid Transformer-BiLSTM Framework with Metaheuristic Feature Optimization for Anomaly Detection in Industrial IoT Networks

Abstract

Industrial Internet of Things (IIoT) networks are primary constituents in several industries such as Manufacturing, Energy, Transportation and Process Control. These networks are targeted by cyber attacks, which cause internal security lapses to the country. Traditionally, the IIoT networks suffered by three limitations such as, inadequate feature selection, failure to adapt to traffic dependencies, inability to distinguish attack classes which appears to be closer and identical but in reality they are different attack classes, could not able to address class imbalance. These bottlenecks are addressed in this research work by proposing (i) hybrid feature optimizer(Hybrid Grey Snake—Cuckoo Search (HGS-CS) metaheuristic optimizer), (ii). Combined Multi Transformer and LSTM approach for detection of attack classes with much better accuracy and (iii) Multi Dataset, cross entropy which enables the migrations are easier and reduces the class imbalance. Three datasets were used in this research are, X-IIoTID, Edge-IIoTset, and WUSTL-IIoT-2021. Multiple performance parameters such as F1 Score, Accuracy, AUC-ROC, Matthews Correlation Coefficient (MCC) etc., were used to test the proposed system. The feature reduction of 54.2%, Detection of attack classes with near perfect accuracy and cross dataset generalization and ability to distinguish multiple attacks with good t-SNE scores are contributions of this research.