An IoT DDoS attack detection model based on graph neural network-based temporal feature fusion

The application of artificial intelligence in cybersecurity has driven the intelligent and refined development of IoT attack detection technologies. Addressing the limitations of traditional methods in capturing attack propagation topology and temporal dynamics, this study constructs a DDoS attack recognition model based on temporal feature fusion using graph neural networks. The model takes node association information and traffic temporal features as inputs, employing multi-layer graph convolutions and gated recurrent structures to achieve feature aggregation and dynamic fusion. Experimental results demonstrate that the model achieves over 98% recognition accuracy on the CICDDoS2019 and Bot-IoT datasets, with significantly reduced detection latency, high robustness, and real-time performance.