Research On The Generation Technology Of Encrypted Traffic Based On Generative Adversarial Nets

At present,the application of traffic encryption technology is more and more extensive.With the steady improvement in public awareness of network security,data protection awareness is also growing.For certain types of traffic,which is legally required for data encryption,there are already more than half of Internet traffic is through data encryption.As the proportion of encrypted traffic in the network is getting higher and higher,some new challenges are ushered in: on the one hand,some users use encryption techniques to hide the malicious traffic(such as worms,Trojan horses and viruses)can have huge negative on the network.Impact;on the other hand,for an attacker efficient decryption technology is complicated by a growing number of attackers turned to regard the target traffic analysis.Research on identification technology to deal with malicious traffic and traffic camouflage technology to deal with traffic analysis attacks has become a hot research direction in the field of network security.Traffic generation technology is one of the key technologies to study traffic identification and traffic camouflage,and it is of great significance to the defense methods to improve network security.Currently researchers at home and abroad made a lot more mature traffic generation model,but these little traffic model for application traffic,especially traffic encryption application.In this paper,for application-encrypted traffic,the use of generative adversarial network to generate traffic,this paper proposes an application encrypted traffic generation technology based on generative anti-network.For encrypted traffic identification,traditional traffic detection methods have been unable to effectively identify encrypted data..Compared with machine learning,deep learning can better reflect the essential characteristics of data.However,deep learning methods usually rely on a large number of labeled samples during training.The number of samples and the accuracy of the label directly affect the training results Recognition rate.Therefore,this paper proposes a DCGAN-based application encryption traffic generation model.Only a small number of application encryption traffic samples are needed to achieve sample expansion and generate a large number of traffic with real application encryption traffic characteristics.The generated traffic can be used for deep learning-based encryption Traffic identification.With regard to traffic camouflaging technology,many current defense methods against traffic analysis attacks lack dynamics.Once detected by an attacker,they completely lose their ability to evade.Therefore,this paper proposes a WGAN-based camouflaging traffic generation model,which can generate camouflaging traffic with target application characteristics.Even if it is detected by an attacker,it can be replaced with traffic with other application characteristics at any time,so as to avoid traffic flexibly Analyze the attack.