A Classification Method For Network Intrusion Detection Based On Deep Generative Model
Yan Lu, Pengxiang Jiao · 2023
With the rapid development of Internet technology, network attacks are increasingly complicated and covert. This is manifested in the following point as for the current commonly-used network intrusion detection technologies: the classification of attack traffic in the network is not imbalanced compared with that of the normal traffic so that it is hard to effectively detect unknown attacks. To solve this problem, a brand-new variational auto-encoder algorithm and generative adversarial network are used to recognize network intrusion. Firstly, network data sets samples are expanded to generate diverse classes of samples and to balance the data classes of training samples; secondly, the potential feature space is used to generate new samples of network traffic to increase the polymorphism of training samples, hoping thereby to transform the problem of unknown attack recognition into the known recognition technology. Finally, the proposed scheme is verified and proved to be effective.