Conditional Diffusion Model for Image Steganography

To ensure robust, high-capacity, and secure communication, we propose a conditional diffusion model for coverless image steganography, called CDIS, which not only generates realistic stego images but also successfully extracts valid secret images in the case of distorted stego images. CDIS utilizes the DDIM inversion technique to achieve the transformation between images, ensuring the reversibility of steganography. Additionally, an image encoder is also used to learn the high-level semantics of the secret image and the stego image, thus speeding up the diffusion and generation process in a controlled manner. The data sampled by the latent semantics module follows the same probability distribution as the high-level semantics of real images, serving as a condition for generating the stego image, thus realizing steganography without embedding. We conduct comprehensive experiments to demonstrate the advantages of our proposed CDIS framework in terms of robustness and security.