A network traffic anomaly detection method based on packet series reconstruction learning

Network traffic anomaly detection has always been an important research topic in network security. In the intricate cyber landscape, cyberattacks are evolving into more sophisticated and diverse forms, especially with the advent of technologies like encryption protocols. Traditional anomaly detection methods have inherent limitations, such as heavy reliance on labeled samples, and have long been unable to adapt to the real network environment. To address these issues, we propose a network traffic anomaly detection method based on packet series reconstruction learning. Self-supervised refactoring learning is introduced to enhance the model’s ability to represent network traffic, reducing the dependence on labeled data. Simultaneously, in the process of packet encoding, we also incorporate a frequency domain enhancement module to enhance the temporal dependencies existing in the packet sequence. Experiments on the cryptographic dataset Datacon prove the effectiveness of our method, which achieves better results than the current mainstream methods.