A survey on Deep-Learning-based image steganography

With the development of Internet and multimedia development, digital image steganography is becoming more extensive in transmitting data with high capacity and security. Although there have been some relevant surveys, there is a lack of relevant surveys on Deep-Learning-based image steganography recently, especially for the developments and perspectives on generative steganography and reversible steganography. So this paper fills this gap and investigates recent developments in Deep-Learning-based image steganography from the viewpoints of strategy and network structures, which is summarized and categorized into ten strategies and five network models. In particular, this paper summarizes the relative merits and challenges of the leading techniques and discusses the research prospect of these fields for promoting the development of steganography.