Robust coverless image steganography based on human pose estimation

Coverless image steganography has garnered significant attention in the field of information hiding due to its robust anti-stego analysis security, as it eliminates the need for modifying the carrier image. With the proliferation of social platforms, a plethora of images containing abundant human information have emerged on the Internet, rendering them an ideal natural image carrier database. In this paper, we propose a pose estimation network model and devise a coverless image steganography algorithm based on pose estimation to conceal confidential information by mapping it onto human pose data. Since all selected carrier images are human-centered and statistically insensitive, and the high-level semantic information shows excellent stability against attacks, the concealment and robustness are enhanced. Experimental results on three public datasets demonstrate that the algorithm achieves an average robustness of up to 92% against multiple attacks.