Graph Embedding for Graph Neural Network in Intrusion Detection System
Dinh-Hau Tran, Minho Park
Abstract
Currently, with the rapid expansion of network systems, network security remains a critical concern. Intrusion Detection Systems (IDS) are widely employed to efficiently detect network attacks. Extensive research has focused on applying machine learning models to IDS. Among these models, Graph Neural Network (GNN) is attracting attention as a promising candidate. However, preprocessing network data for the GNN model still poses several challenges. Thus, in this study, we propose an innovative approach to preprocess network flow data before feeding it into the GNN model. Our method involves extracting relevant features from flow data to create nodes and edges for the GNN model. The simulation results indicate that our proposed method significantly enhances the performance of IDS in detecting network attacks.