The Research Of Image Steganography Against Statistical Analysis

Steganography is the means of concealed communication, it takes an important role in military affairs, information and security of country. the basic demand for steganography is providing good security and large embedding capacity, and security is the key requirement in steganography. Security includes visual imperceptibility and statistical imperceptibility. Now Steganalysis is the main measure against steganography, it uses the statistical analysis means to detect the statistical abnormality of steganographic digital media, thereby to detect the presence of secret information, so the statistical imperceptibility is the main factor that steganography considers.The steganographic method based on least significant bit(LSB) is simple, providing large embedding capacity. The steganographic method based on human visual system(HVS) can providing larger embedding capacity and better visual imperceptibility, therefore they abtain the widespread application. But now multi-effective stegananalysis against them have already come forth. In order to improve steganographic security, according to these steganographic algorithms existing stastical abnormality and the corresponding steganalysis, this thesis proposes steganographic algorithms that can effectively resist statistical analysis , so have better security, the main achievement of this thesis as follows:(1) According to the security flaw in Bit-plane complexity segmentation(BPCS) steganography, i.e. two striking discontinuities in the complexity histogram unambiguously announce the presence of BPCS-based hidden information, this thesis proposed an improved algorithm, We set different threshold values at different bit planes, and construct a function to evaluate the block brightness, then combine region brightness where block locates and block complexity to select bit-plane blocks for embedding. Experimental results show that adopting the improved algorithm perserce the smoothness of complexity histogram, so effectively resist statistical analysis based on the complexity histogram, and the stego-image has better quality than BPCS-Steganography.(2) Side match(SM) steganography inevitably causes abnormality in the pixel-value difference histogram, so there are nusual steps in the histogram of pixel-value differences .This provides a clue for steganalysis to reveal the presence of a secret message. To enhance security, a modified scheme is proposed. Classify pixel-value differences |d| into a number of contiguous ranges with dynamic, which avoids occurrence of the above-mentioned steps in the pixel-value difference histogram. And adjust d to the value that is closest to the original among all the candidate values which can effectively reduce the distortion and achieves better imperceptible. Experimental results show the effectiveness of the proposed technique.(3) This thesis proposes a new steganographic algorithm that applying dynamical running coding into the least histogram abnormality(LHA). If the space used to insert secret infomation is larger than the number of secret bits, during the embedding process of LHA, dynamical running coding can use the residual space, and by calculating to select the cover bites to be changed in order to ruduce host alterations. Experimental results show that the algorithm improves the embedding efficiency, in the same conditions, the numbers of bit alterations introduced into a cover image is significantly less, thus improves the steganographic imperceptibility and enhances security against steganalysis.