Research On High-Order Statistical Secure Steganography Algorithm

Steganography is an important branch of information hiding research, and its main purpose is covert communication. Steganography is to embed secret information which needs to be transmitted into'innocent'cover-object without causing suspicion of monitors, and send the stego-object to the receiver through public transmission channels, and the receiver correctly extract the secret information from the stego-object he received. On the contrary, steganalysis is to analyze vary kinds of'innocent'communication behavior as a communication monitor, and detect the existence of covert communication, even further to extract the secret information being transmitted. Steganography and steganalysis confront each other and co-develop. They are the interdisciplinary subjects of information security, multimedia signal processing, pattern recognition and so on as the very active and important component in the research of information security field. It is the core and difficulty problem to evaluate the security of steganography system and design the embedding algorithms with high-capacity, high-security and low distortion. This dissertation studies deeply the evaluation of high-order statistical security of steganography, the design of steganalysis algorithm, the design of high-order statistical secure steganography algorithm, and the application of cover-codes. The main contributions can be enumerated as follows:Firstly, for the evaluation of the high-order statistical security of steganography, a high-order Markov chain model for digital image steganography is proposed in order to characterize the correlative of the spatial adjacent pixels of cover-image. And the effect of several general image scanning methods for the model is compared. Based on the model a high-order statistical measure between the cover-image and stego-image is proposed. Moreover, the relationship of the measure and the traditionalε-secure criterion is analyzed.Secondly, the effect of four kinds of additive and multiplicative spread spectrum steganography in the cover-images'spatial and DCT domain on the image high-order Markov chain model is studied. Then a kind of steganalysis algorithm, against the spread spectrum image steganography is proposed which extracts image pixels'statistical features based on the empirical matrix of the high-order Markov chain model and uses support vector machine for classification.Thirdly, a LSB match steganography algorithm with second-order statistical security is proposed. It optimizes the embedding algorithm based on the statistical measure of image spatial pixels'Markov chain model. Furthermore, for large numbers of steganography, after analyzing the effect on the statistical distribution of the cover image spatial pixels'Markov chain model causing by LSB match embedding, a fast LSB match steganography algorithm with second-order statistical security is proposed based on the dynamic compensation idea.Fourthly, the image high-order Markov chain model is extended to the prediction-error domain of image pixels. In the prediction-error domain of cover image pixels, a second-order statistical secure optimized quantization steganography algorithm which based on the Markov statistical measure is proposed. Moreover, for large numbers of embedding, a fast compensated quantization embedding algorithm which maintains the second-order statistical distribution of the pixel prediction-error of cover-object is proposed.Finally, a self-adaptive algorithm for image steganography which reasonably uses a cluster of cyclic cover codes based on the local pixels complexity of cover image is proposed. Furthermore, a self-adaptive image steganography algorithm which takes account of the perceptual distortion and second-order statistical security is proposed. The algorithm combines the dynamic compensation method based on image high-order Markov chain model with the self-adaptive method based on cover codes.At last, the deficiencies in the dissertation are summarized, and some open issues in information hiding as well as the future work are presented.