为了提高对高精度原始图像LSB匹配数字隐写的检测能力,提出了基于直方图特性的LSB匹配隐写分析方法。根据高精度原始图像在LSB匹配数字隐写过程前后灰度直方图、差分直方图以及小波分解后的子带系数直方图的特性差异,提取了23维的特征向量,并使用支持向量机对其进行训练,建立区分载体图像与载密图像的分类器达到检测秘密信息的目的。实验结果表明,该方法在性能上要优于文献[5]的局部极值法。
In order to improve the detection of LSB matching steganography for high resolution raw images, a method of LSB matching steganalysis based on histogram properties is proposed.According to the different properties in histogram, difference histogram and subband coefficients histogram after wavelet decomposition between before and after the LSB matching, 23 features are proposed for SVM training, and a classifier based these features is used to distinguish cover and stego image.Experimental results show that, the proposed method is better than Zhang's local extremum method.