A Image Steganography Scheme Based on Auxiliary Classifier Generative Adversarial Networks

Abstract:
In order to solve the problem that the current generative steganography based on generative adversarial networks (GAN) couldn’t control the content of the generated image, this paper proposed an image steganography scheme based on auxiliary classifier generative adversarial networks (ACGAN). Firstly, ACGAN was trained by random noise and tag information to get the generator controlled by tag information. Secondly, the extractor network was trained by the generated image to get the trained message extractor. The images with different contents could be generated after training according to the specific environment for steganography, which could effectively avoid the exposure of the communication process and enhance the security of the communication system.