Temporal and Topological Enhanced Graph Neural Networks for Traffic Anomaly Detection

With the rapid advancement of network technologies, ensuring the security of network communications has become increasingly critical. Network traffic anomaly detection plays a pivotal role in identifying irregularities that threaten the security, reliability, and stability of cyberspace services. Recently, deep learning-based approaches, particularly those utilizing Graph Neural Networks (GNNs), have gained attention due to their powerful representation learning capabilities. However, these methods are limited by the receptive field of GNNs and their ability to capture temporal feature dependencies, leaving room for performance improvement. To address these limitations, we propose a novel GNN with a pre-characterization mechanism using PageRank to enhance the receptive field and improve accuracy without over-smoothing. Additionally, we incorporate a temporal attention module to capture potential temporal dependencies in the data. Our experimental results demonstrate that our method achieves a detection accuracy of 98.3%, representing a performance boost of approximately 3% compared to existing approaches.