Research on Information Hiding Technology Based on Generative Adversarial Networks

Abstract

Aiming at the problems of the existing image steganography model that the network training is not easy to converge, the gradient explodes, and the quality of the generated samples is poor, this paper proposes an image adaptive steganography model SWGAN-GP based on an improved generative adversarial network. The scheme first uses the generative model G in the generative adversarial network to generate the original carrier information-driven by noise. Then the system extracts the secret information in the secret image through the discriminator and realizes the carrier-free information hiding with the help of the generative adversarial network. The experimental results and analysis show that the hiding method has the advantages of steganographic capacity, anti-steganalysis, and security good performance.