Research on network traffic intrusion detection based on GAN

With the rapid growth of computer network devices around the world, network threats are increasing, and securing the network environment has become an urgent problem. However, the current network traffic dataset is generally characterized by unbalanced data distribution, in which certain anomalous traffic types have a relatively small number of samples, which causes the detection model to ignore traffic types with a small number of samples, thus reducing the overall accuracy of the model. To address this problem, this paper proposes an improved intrusion detection method based on generative adversarial networks (GAN). By designing a regularized bi-discriminator structure to improve the generation quality of the generator. In addition, a random mask enhancement module is integrated after the generator to supplement the detailed information of data features. Our designed GAN model is applied to Support Vector Machine (SVM) and Long Short-Term Memory Network (LSTM), the experiment results show that it significantly improves the detection accuracy.