摘要： 针对隐写分析中JPEG图像二次压缩的检测问题，提出一种新的压缩检测算法。通过对JPEG一次压缩和二次压缩后的绝对值直方图进行比较，发现直流与交流系数差分绝对值、交流与交流系数差分绝对值在二次压缩后有明显变化。在此基础上提取特征，采用SVM进行分类。实验结果表明，在虚警率小于5%的情况下，该算法检测性能优于Pevny T和Fridrich J提出的算法(IEEE Trans. on Information Forensics and Security, 2007, No.2)。

关键词: 隐写分析, JPEG图像, 质量因子, 二次压缩

Abstract: This paper proposes a new algorithm for detection of double compression in JPEG image for applications in steganalysis, which based on the analysis of the influences on absolute histogram brought about by the double compression. It compares the histograms of single-compressed and double-compressed images and finds that both the absolute differences of DC and AC coefficients and the absolute differences of AC and AC coefficients change obviously after double compression. Support Vector Machine(SVM) classifiers are trained to detect an image whether double-compressed or not. Experimental results show that the proposed method is better than the algorithm proposed by Pevny T and Fridrich J(IEEE Trans. on Information Forensics and Security, 2007, No.2) while the false positive alarm is controlled under 5%.

Key words: steganalysis, JPEG image, quality factor, double compression

中图分类号: 

• TP391