Abstract:

As for the deficiency of the existed steganalysis algorithm for color image detection, this article provides a steganalysis algorithm based on wavelet statistics to analyze the statistic differernce between carrier image and stego image. The method, transforming the color image into the YCbCr space and extracting the prediction subband coefficients’ probability density function(PDF) and the predictionerror image subband coefficients’ histogram characteristic function(CF) moments through the wavelet decomposition, makes use of the support vector machine(SVM) to achieve classification. The experimental results for the popular JPEG steganographic schemes, for example, F5, OutGuess, Jphide&Seek, demonstrates that the proposed scheme can achieve good detection accuracy.

Key words: steganalysis, predictionerror image, PDF moments, CF moments