In this paper, a novel strategy of Secure Steganograpy based on GenerativeAdversarial Networks is proposed to generate suitable and secure covers forsteganography. The proposed architecture has one generative network, and twodiscriminative networks. The generative network mainly evaluates the visualquality of the generated images for steganography, and the discriminativenetworks are utilized to assess their suitableness for information hiding.Different from the existing work which adopts Deep Convolutional GenerativeAdversarial Networks, we utilize another form of generative adversarialnetworks. By using this new form of generative adversarial networks,significant improvements are made on the convergence speed, the trainingstability and the image quality. Furthermore, a sophisticated steganalysisnetwork is reconstructed for the discriminative network, and the network canbetter evaluate the performance of the generated images. Numerous experimentsare conducted on the publicly available datasets to demonstrate theeffectiveness and robustness of the proposed method.
展开▼
