IoT Intrusion Detection Model based on CNN-GRU

Abstract

With the rapid development of IoT technology, security concerns surrounding IoT devices have gained attention. An intrusion detection system for IoT can quickly and accurately identify highly redundant data features in IoT traffic categories. To reduce data, feature redundancy during the identification process, this study proposes the use of Extreme Gradient Boosting (XGBoost) for feature selection to obtain an optimal feature subset. Additionally, to improve the accuracy of identifying malicious traffic in IoT devices, a fusion model combining Convolutional Neural Networks (CNN) and Gated Recurrent Units (GRU) for IoT intrusion detection is proposed. Finally, a comparative analysis experiment is conducted between CNN-GRU and CNN-LSTM, demonstrating that the proposed model achieves lower processing time while ensuring accuracy. Furthermore, the proposed method outperforms classical IoT intrusion detection algorithms in terms of precision and recall.