Cost Function Design For Image Steganography Under Embedding Distortion Minimization

This dissertation is funded by National Natural Science Foundation "Research on the steganography with controllable feature vector " under grant number 6113174.With the development of computer networks and multimedia tools and systems, new technologies not only bring people all kinds of conveniences regarding their daily life, but also lead to new forms of illegal behavior, including information disclosure, eavesdropping,etc. As a result, people demand higher better search for the information. Much attention has been paid to the image steganography and steganalysis due to the fact that they acted as an important research topic in the field of information security. The design of a security steganographic algorithm with low distortion and high payload has been the main goal of steganographic research. This dissertation will be addressing the issue of image steganography based on distortion-minimization. It focuses on the design of a cost function for modeling the distortion, which is based on the related theories and methods of distortionminimization. The main contributions of this thesis are as follows:(1) A new cost function combined with the strategy of clustering modification location is proposed. The proposed cost function is designed by using a high-pass filter and two low-pass filters. The high-pass filter is used to locate the less predictable parts in an image, and the two low-pass filters are used to make the low cost values more clustered. Experiments verify that the steganographic method with the proposed cost function makes the embedding changes more concentrated in texture regions, and has a better performance on resisting the state-ofthe-art steganalysis over prior works.(2) A new cost function combined with the strategy of clustering modification directions is proposed. The proposed cost function uses dynamic update and its whole process is divided into two stages. We get the initial costs from the updated stego image with partial secret information in the first stage. Then, in order to employ the strategy of clustering modification directions, we improve the costs by adjusting the initial costs according to the results of embedding from their neighborhood pixels. An advantage of the steganographic method with the proposed cost function is taking consideration to the correlation of modification directions.Experiments confirm that the proposed scheme makes the modification directions more clustered locally, and thus has a better performance on resisting some of the state-of-the-art steganalysis over prior steganographic methods.