Dynamic Residual Graph Attention Network for Network Intrusion Detection System
Song Wang, Zhenming Zhang, Wei Li, Chen Bo Yin, Yu Ma, Weiyao Xu · 2024
With the rapid development of the industrial internet, it greatly promotes the advancement of social productivity. In the context of frequent network attacks, the security issues have become increasingly important. Network intrusion detection system(NIDS) is an important component of industrial internet security defense, which detects and prevents attack traffic by monitoring the status of network traffic. Many researchers have conducted research in this field, such as traditional machine learning methods. However, these methods still have some practical problems and many challenges remain unresolved. In this paper, we focus on the original network traffic and present an intrusion detection method based on the dynamic residual graph attention network (DRGAT). We introduce residual calculation and an improved dynamic attention mechanism to form a more robust graph-based NIDS, which can represent and learn traffic information, and detect various attack traffic with high accuracy. Experiments on two public datasets show that the proposed method outperforms the most advanced methods in multiple performance indicators, providing impetus for further research.