Edge-based Analysis for Network Intrusion Detection using a GNN Approach

Anomaly detection is an essential task in detecting fraud and identifying faults in various areas, including network security. Network Intrusion Detection Systems (NIDS) have traditionally been used to detect known threats; however, machine learning models such as Graph Neural Networks (GNN) are gaining popularity with the increasing complexity of cyberattacks. This paper presents an implementation of a NIDS based on a GNN machine learning model for anomaly detection in network traffic, focusing on the edges of the input graphs. Moreover, the modules incorporated into the system are detailed. Finally, the results and analysis obtained from the experiment are presented, demonstrating the effectiveness of the proposed GNN-based anomaly detection approach.