ABDNN-IDS: Attention-Based Deep Neural Networks for Intrusion Detection in Industrial IoT

The increasing trend of the Industrial Internet of Things (IIoT) within industrial environments magnifies the risk of security breaches and vulnerabilities. Maintaining confidentiality is a pivotal requirement for effectively establishing the IIoT environment. To promptly detect malicious endeavors, integrating an intrusion detection system (IDS) becomes imperative for continuously monitoring IIoT activities. The sophisticated automated IDSs are built upon the foundation of machine learning (ML) and deep learning (DL). However, these algorithms encounter challenges related to heavily imbalanced training data and the need for accurate predictions in a short timeframe. This paper introduces an attention-based deep neural network (ABDNN) designed to tackle these challenges for intrusion detection within the IIoT environment. The attention mechanism plays a pivotal role in determining the significance of each attribute in the input data. Subsequently, the deep neural network (DNN) comes into play, leveraging the previously determined attribute importance to predict network behaviors. This process yields the advantage of predicting network behaviors more efficiently in less time. The performance of the proposed ABDNN model was evaluated using the X-IIoTID dataset. To validate its effectiveness, a comparison was made between the performance of the proposed model and that of state-of-the-art approaches. This comparative analysis serves to validate the superior performance of the proposed ABDNN model.