Abstract: An improved blind steganalysis method for JPEG images is proposed in this paper.It decomposes the DCT coefficients of the JPEG images directly to the wavelet domain，and calculates a series of higher-order statistics of the wavelet coefficient of each subband.Calibrates the received higher-order statistics of the wavelet coefficients in order to reflect the change of image better，and takes the calibrated statistics as final features of images.It applies support vector machine as the classifier to judge if an image contains secret messages.The experiment results of Outguess and JP Hide & Seek show that this method gets a higher detection rate.

Key words: steganography, blind steganalysis, wavelet decomposition, Discrete Cosine Transform（DCT）, Support Vector Machine（SVM）

摘要： 提出了一种针对JPEG图像隐写的盲检测改进方法。直接把JPEG图像的DCT系数分解到小波域，计算出各个子带小波系数的一系列高阶统计特征。对得到小波高阶统计特征进行校准，以使统计特征能更好地反映信息嵌入以后的变化。校准后的统计特征作为图像最后的特征。应用基于核技巧的支持向量机SVM作为分类器，判断图像是否含有隐藏信息。针对两种典型的JPEG类嵌入方法Outguess和JP Hide & Seek进行了实验，结果表明该方法有效地实现了信息隐藏的盲检测分析。

关键词: 信息隐藏, 盲检测, 小波分解, 离散余弦变换, 支持向量机