Researchs On Perturbed Quantization Based Steganography

Derived from the Greek literature, information hiding is an ancient technique for covert communication. In recent year, with the rapid development of Internet, huge amounts of information can be shared in every corners of the world, which makes this ancient technology have a new meaning and become a hotspot. Different from encryption, the purpose of steganography not only makes the message unintelligible but also hides the very presence of the communication by embedding secret message into innocuous-looking cover objects, which makes steganography more secure than encryption.In this paper, the main JPEG-based stananographic and steganalytic methods are analyed from the principle of information hiding. And a PQ steganalysis based on correlation of DCT coefficients and a JPEG compression process-oriented steganography are proposed respectively.Firstly, through analyzing J.Fridrich's Perturbed Quantization based staganography, the author found that it has significant impacts on both correlation of DCT coefficients and high-frequency of non-zero DCT coefficients. To classify the cover-image and stego-images more accurately, inter- and intra-block Markov features and LSB features are extracted. After a large number of tests, we found that the proposed algorithm can get better performances.Secondly, a JPEG compression process-oriented steganogtaphy are proposed by further research of PQ steganography. According to the principle of the statistical properties produced by PQ algorithm, the proposed method makes use of lossless images as carriers, which can better preserve the correlation of DCT coefficients. Moreover, the new method selects revisable DCT coefficients according to the differences of the DCT coefficients in each 8×8 DCT block, with which method impacts to the high frequency of DCT coefficients is reduced remarkably. And experiments confirmed that this method can better maintain the statistical properties of images and improve the security level significantly.Thirdly, from the analysis of Wet Paper Codes, we point out that there are some shortcomings on encoding complexities and embedding capacity. To make it more adapted to JPEG image steganography, an improved encoding method is proposed. In addition, the BER (Bit Error Ratio) is estimated, and a variety of solutions are given.