Intrusion Detection for Internet of Things Networks using Attention Mechanism and BiGRU

Abstract

The rapid growth of Internet of Things (IoT) devices has led to increased vulnerability to serious security threats within IoT networks. As such, it is crucial to employ network intrusion detection techniques to monitor these networks. This paper presents an intrusion detection model, SEW-MBiGD, which integrates data processing and fusion neural networks to address data imbalance and insufficient feature learning in existing models. Firstly, to balance the dataset and mitigate the influence of edge data, the model employs Synthetic Minority Over-sampling Technique (SMOTE) and Edited Nearest Neighbors (ENN) algorithms for data preprocessing, while also utilizing Wasserstein Generative Adversarial Networks (WGAN) to generate minority class data. The proposed intrusion detection model is based on Bidirectional Gated Recurrent Unit (BiGRU) and multi-head self-attention (MHSA) mechanisms, which effectively manage long sequence data and capture correlations between data and global features. Experimental results demonstrate the efficacy of the proposed SEW-MBiGD model outperforming baseline models in achieving a balanced dataset and classification tasks.