Network traffic anomaly detection and security protection technology based on TDWGAN-TCT

The power network faces the challenge of low abnormal traffic and covert attacks. Traditional technologies are difficult to accurately identify anomalies and ensure data security. To solve this problem, a detection technique based on dual discriminator generative adversarial and temporal convolution converter is proposed. It utilizes a generator to synthesize samples, combines dual discriminator feedback to achieve data balance, and extracts temporal features through causal convolution and self attention mechanism to achieve classification. In addition, the research constructs a lightweight encryption mechanism based on Certified Public Key Cryptography (CL-PKC) and the Chinese remainder theorem. This mechanism utilizes physically unclonable features to achieve bidirectional authentication and achieves fast key negotiation through a system of congruence equations. The experiment is conducted on different network traffic datasets. In anomaly detection, the best accuracy is 96.1%, and in precision recall testing, the highest accuracy is 0.976, better than similar models. In security protection testing, the decryption time is only 581 ms, and the unicast negotiation latency is as low as 34.2 ms. In identity authentication testing, the accuracy is 98.6% when the number of identity forgery attacks is 1000,outperforming similar technologies. The proposed technology has good application effects. The research will provide technical support for accurate identification of anomalies and low-cost secure transmission.