Graph Neural Networks for Intrusion Detection in Next-Generation 6G Networks: A Cybersecurity Perspective

As the future 6G networks are being developed, the number of connected devices and ultra-reliable low-latency communications is expected to increase further and bring new cybersecurity threats. The conventional IDS do not suffice the pertinence of the penetrative and constantly evolving 6G network which is hybrid in nature. This paper puts forward the GNN based intrusion detection system that analyze network traffic under the temporal graph structure in order to identify relational and structural patterns. In the context of the NSL-KDD dataset, the basic entities are represented as graph nodes and the interactions, as edges; the nodes bear attributes such as node size, session time, and type of communication protocol. Some of the tested architectures are GCN, GAT, and Temporal GNNs (TGAT, TGN). This showed that the proposed model was able to obtain good results with better performance when using TGAT, GAT and GCN. The classification accuracy achieved were namely 0.9975, 0.9960, and 0.9854 respectively. Besides, the system’s performance showed its heterogeneity across different attacks like DDoS, Botnet, and Phishing, and accuracy maintained above 0.98 all the time. Accordant with the findings above, it has been established that GNNs, when combined with temporal modeling, provide adequate real-time IDS for 6G networks.