EE-GCN: A Graph Convolutional Network based Intrusion Detection Method for IIoT

Abstract

Intrusion detection in the Industrial Internet of Things (IIoT) is a challenge for the network security protection. Graph neural network (GNN) is employed to improve the network security by virtue of efficiently constructing a message passing function. However, existing intrusion detection methods based on GNN do not fully exploit the information of original data which results in the poor intrusion detection performance. In this paper, we propose an Exploiting Edge feature based on Graph Convolutional Network (EE-GCN), which can capture both the edge features of the network traffic link as well as the relationship between device nodes. In addition, we construct a two-layer GCN network to extract the edge features. Finally, two benchmark datasets (NF-BoT-IoT and NF-ToN-IoT) in Network Intrusion Detection System (NIDS) are used to evaluate the performance of the proposed method. The results show that the method proposed in this paper outperforms other methods.