Steganography Based On Encoding And Image Statistics

Steganography, also known as covert communication, is an art of data hiding whose purpose is to hide the presence of secret information against eavesdroppers. A successful steganography system should perform as secure as possible. As far as image steganography is concerned, the cover image after embedding, i.e. the stego image, should have high visual quality and statistical anti-detection ability.To date, a vast amount of research has done in order to improve the security of steganography. In particular, a lot of effort has been put into increasing the embedding efficiency and various encoding schemes have come into being. In another way, many papers are published in attempt to preserve statistical features of cover images. Recently, adaptive steganography have drawn more and more attention of researchers.In this thesis, we take advantage of the existing encoding schemes in conjunction with our study on the impacts of embedding on image statistics and propose several new steganographic systems with high anti-detection performance. The main contributions of our study are as follows.(1) Steganography based on exchanging pixels and diamond encodingInspired by Steghide and diamond encoding, we develop a new steganographic method which not only preserves first-order statistics of the cover data, but also has high embedding efficiency. Pixels are allowed to be changed only by1or-1, thus decreasing the distortion of the cover image. Experiments demonstrate that the proposed method outperforms Steghide and diamond encoding in terms of embedding efficiency, visual quality and anti-detection ability.(2) Adaptive steganography based on variance of image histogramThe impact of embedding on the cover image depends on the distribution of the pixels. Our study shows that random noise makes no difference to the histogram of an image given that the histogram is homogeneously distributed. Following this principle, we propose a novel adaptive steganography based on variance of histogram, which has great anti-detection ability as shown in the experiments.(3) Histogram variance adaptive steganography using syndrome trellis codeAdaptive steganography based on variance of histogram is capable of selecting the embedding-insensitive areas of the cover image, while syndrome trellis code steganographic system provides very high embedding efficiency. Using STC encoding scheme in conjunction with the HV adaptive principle, we develop a novel steganography with higher security than each of the individual methods.