An Novel Image Steganography Method via Deep Convolutional Generative Adversarial Networks

In this paper, we propose a novel image SWE method based on the deep convolutional generative adversarial networks (DCGANs). We map the secret information into a noise vector, and then use the trained generator neural network model to generate the carrier image based on the noise vector. During the process of image generation, there is no modification or embedding operation.本文提出一种基于深度卷积生成对抗网络（DCGANs）的新型图像SWE方法。我们将秘密信息映射至噪声向量，随后利用训练完成的生成器神经网络模型，基于该噪声向量生成载体图像。在图像生成的全流程中，未执行任何修改或嵌入操作。