Steganalysis Of Additive Noise Modelable Information Hiding On Images

The purpose of information hiding on images is to hide the very presence of communication by embedding messages into innocuous-looking cover images. As a safe means of keeping communications secret, it also can be exploited by criminals as a tool to endanger national security and social stability. Meanwhile,steganalysis,as a counterpart of image steganography, is becoming a new focus in the information security technical field.The study on steganalysis in this paper is to detect the existense of secret messages in cover images. The paper first introduces LSB steganography algorithms and corresponding steganalysis algorithms in detail. Then, the paper focuses on analyzing the system model of image steganography based on noise adding, thoroughly discusses the use of corresponding steganography algorithms and then proposes a new steganalysis algorithm based on bispectrum analysis. In view of the independence between cover-image signals and the hidden data, the new algorithm construct a model of relationship among adjacent pixel difference histogram of stego-image,cover-image and secret message. Having a estimation on the statistical character of secret message and modeling adjacent pixel difference histogram of the cover image by Generalized Gaussian Distribution, we use bispectrum curve to describe the effect of the embedding signal to the statistical character of cover-image, then design an efficient parameter to measure the change of the statistical character of the cover-image. Then, the proposed steganalysis algorithm is simulated and a typical image steganography method is tested, the testing results prove that the new algorithm has a good performance at high embedding rates.