Design And Implementation Of Covert Communication Technology Based On Generative Adversarial Network

In the context of the interconnection of all things,covert communication using traffic masquerading and channel invisibility has become a trend for malware to upgrade the network's adversarial capabilities,some new covert communication technologies that are resistant to analysis and imagery have emerged.Covert communication game framework based on the prisoner model,in the prior art solutions,there are common problems with modification of covert carriers and "statistical marks",that is,the covert algorithm realizes a clear statistical distribution.In order to predict and verify the future technical direction of covert communication,and provide forward-looking support for the management of covert communication,this paper discusses the nature of covert communication and the new paradigm of the game,there is no need to modify the covert carrier,the covert carrier is directly "synthesized" by using Generative Adversarial Networks,the corresponding covert content is extracted and restored,the feasibility and effectiveness are verified through relevant experiments.This paper is to explore new covert communication models and technologies,the main research results are as follows:Firstly,Analyze the shortcomings of existing covert communication technologies,and address the common problems of covert carriers being modified and "statistical marks",a new research idea is proposed to solve these problem by using the adversarial game theory of Generative Adversarial Networks,which will pave the way for the subsequent study of new network covert communication schemes.Secondly,with the goal of solving covert carriers being involved in modification and "statistical marks",analyzing the characteristic attributes of mainstream Generative Adversarial Networks,and screening out Information Maximizing Generative Adversarial Networks with more ideal theoretical concealment.Through preprocessing,the association mapping mechanism between the secret information and the InfoGAN covert information code vector Vc is established,through the Vc inputing generator G,the covert information is seamlessly synthesized into the secret network traffic data samples,and then a set of covert communication framework based on InfoGAN is proposed.Thirdly,in order to improve the accuracy of extracting the secret information by the receiver,this paper combines the network structural characteristics of the DCGAN with the network principles of InfoGAN,and designs InfoGAN whose generator network structure is opposite to the auxiliary discriminator Q,and then changes the output layer of Q to implement a refine network model R that can extract covert information vectors from traffic data samples.Fourthly,combining covert communication scheme based on InfoGAN,InfoGAN network model,generator network model and refine network model,a new covert communication system based on Generative Adversarial Networks is designed,and the overall design scheme of the system and the modular details of key functional modules are given.Finally,this paper carried out related experiments based on the system function and performance,and the experimental results show that the generator and refine network models proposed in this paper have engineering effectiveness,the sender can realize the seamless synthesis of covert information,the synthesized secret network traffic has network activity and concealment,and the receiver can correctly extract and restore the covert content.Comparative tests show that,while maintaining a certain transmission efficiency,the concealing ability of the system in this paper has obvious advantages over traditional covert communication tools.