Coverless Image Steganography Based on Multi-Object Mapping Rules

Existing coverless image steganography methods mainly focus on improving the hiding capacity and robustness against attacks. However, most of them ignore the number of images for constructing a complete index of feature sequences could be huge, thus making it challenging to apply to the datasets. In this study, we propose a coverless image steganography based on multi-object recognition, which builds an image sequence index based on objects to make the mapping rules more flexible. It enables a single image to generate multiple sequence combinations and effectively reduces the number of images required to construct a complete sequence index. The experimental results show that the proposed method can build a comprehensive sequence index using the existing datasets and be applied to the real data-hiding task. In addition, it can enhance the capacity while maintaining good robustness.