The Use of Generative-Adversarial Networks to Counter Steganalysis

Abstract

An approach using a generative adversarial network (GAN) is proposed to increase the robustness of the steganographic method against modern steganalyzers. This approach is based on the combined operation of a GAN, a pixel importance map, and the least significant bit (LSB) substitution method. The results of the experimental studies confirmed the effectiveness of the proposed approach.