T-GAN: Transformer-based Generative Adversarial Network for Network Traffic Anomaly Detection

With the rapid development of network technologies, network traffic anomaly detection has become critical for ensuring information security and network stability. However, challenges persist in anomaly detection tasks, including the complex dynamic patterns inherent in network traffic data and the scarcity of labeled anomaly samples, which lead to low recall rates. This paper proposes a Transformer-based Generative Adversarial Network (T-GAN) anomaly detection model. The generator learns the distribution of normal traffic data, while the multi-head attention mechanism captures long-range dependencies in network traffic sequences to enhance detection efficiency. Experimental results demonstrate that the proposed model achieves near 100% recall with a tolerable false positive rate on the test dataset, validating its effectiveness and feasibility.