Synchronization Detection and Recovery of Steganographic Messages with Adversarial Learning
As a means for secret communication, steganography aims at concealing a message within a medium such that the presence of the hidden message can hardly be detected. In computer vision tasks, adversarial training has be-come a competitive learning method to generate images. However, the gen-erative tasks are confronted with great challenge on synthesizing images. This paper studies the mechanism of applying adversarial learning for dis-criminative tasks to learn the steganographic algorithm. We show that through unsupervised adversarial training, the adversarial model can pro-duce robust steganographic solutions, which act like an encryption. Through supervised adversarial training, we can also train a robust ste-ganalyzer, which is utilized to discriminate whether an image contains se-cret information. Our model is composed of three modules, i.e. a generator, a discriminator and a steganalyzer. All the three members are trained simulta-neously. To formulate the algorithm, we use a game to represent the com-munication between the three parties. In this game, the generator and dis-criminator attempt to communicate with each other with secret messages hidden in an image. While the steganalyzer attempts to analyze whether there is a transmission of confidential information. Experimental results demonstrate the effectiveness of the proposed method on steganography tasks.
READ FULL TEXT