一种针对加性空域掩密算法的通用掩密分析技术

Universal steganalysis algorithm for additive noise modelable steganography in spacial domain

为了提高通用掩密分析技术的检测性能,针对基于加性噪声的空域掩密技术设计了一种通用掩密分析算法.通过分析LSB、+-k(k=1,3,5)和SM算法所引入的掩密噪声对图像差分直方图产生的影响,提出分别利用差分直方图分布的高阶统计矩和差分直方图曲线下特殊区域之间的面积比值作为分类特征的掩密分析特征选择算法.分别利用该算法和Farid的通用算法对由LSB、+-k(k=1,3,5)算法掩密后的图像进行检测,实验结果表明,对于不同的嵌入消息比例,该算法的掩密检测率较Farid的算法都有了大幅度的提高.

The detection performance of the universal steganalysis is improved with a new algorithm put forward for the additive noise modelable steganography in the spacial domain specially. The effects of stego noise introduced by LSB （least significant bits）, ＋ -k （k = 1, 3, 5）and SM （stochastic modulation） steganography algorithms on the image differential histograms are investigated. Then, feature selection for universal steganalysis is presented, in which the higher order statistics moments of the differential histogram distributions and the ratios between the two areas of special regions under the differential histogram curves are calculated as the features of the classifier. Experimental results show that it achieves better detection performance than Farid＇s universal algorithm for different proportions of embedded messages when it is applied to detect the stego images produced by LSB, ＋ - k （k = 1, 3, 5） algorithms.