Review of Coverless Image Steganography

Traditional image steganography techniques have revealed many drawbacks due to advancements in steganalysis and steganalysis detection technologies. With the continuous development and improvement of various generator-free models, more and more researchers are shifting their focus to the combination of image generation models and steganography techniques. Compared to traditional methods, generative steganography directly drives the secret information to generate images containing the hidden data, exhibiting better resistance to detection against existing statistical feature-based steganalysis. This paper describes and analyzes GAN-based steganography algorithms as well as the latest steganography algorithms based on diffusion models and deep generative models. At the end of the paper, current issues in generative steganography are discussed, along with future research and development directions.