基于统计特征的JPEG图像隐写分析技术

STEGANALYSIS TECHNIQUE OF FEATURE-BASED STEGANALYSIS FOR JPEG IMAGES

介绍一种新的基于特征统计的JPEG图像隐写分析技术。利用选择的特征组,进行多元回归分析设计一个线性分类器,可以精确地判定图像中隐藏信息的存在性。这种方法可以检测各种JPEG图像信息嵌入系统,如J-Steg,OutGuess和F5算法。相对于现有的基于图像质量度量(IQM)的统计方法,新方法的检测准确率更高(对同一嵌入率,本方法准确性提高了至少10%)。

In this paper, we introduce a new feature-based steganalytic method for JPEG images. The detection method is a linear classifier which is built using multivariate regression on the selected features, This method can detect any embedding scheme, such as J-Steg, Out Guess and F5. Using this mothod the estimating precision is about 10% higher than the existent in the method of IQM ac-cording to the same embedding ratio.