DIGDH: A Novel Framework of Difference Image Grafting Deep Hiding for Image Data Hiding

Data hiding is the technique of embedding data into video or audio media. With development deep neural networks (DNN), quality images generated by novel methods based on DNN getting better. However, there still room for similarity between original and models which were trained existing frameworks to improve, it hard receiver distinguish whether container image from real sender. We propose a framework introducing key_img using over-fitting characteristic combined with difference grafting symmetrically, named (DIGDH). The can be used identify sender easily. experimental results show that without changing structures networks, proposed generate higher cover secret images. According analysis steganalysis tool StegExpose, model closer random distribution.