基于校准DCT特征的JPEG图像隐写分析  

JPEG IMAGES STEGANALYSIS BASED ON CALIBRATED DCT FEATURE

作　　者：关俊波[1] 马春波[1] 敖珺[1] 

机构地区：[1]桂林电子科技大学信息与通信学院,广西桂林541004

出　　处：《计算机应用与软件》2014年第1期229-231,263,共4页Computer Applications and Software

基　　金：国家自然科学基金项目(61167006);广西省自然科学基金项目(2012GXNSFBA053173)

摘　　要：提出一种基于校准DCT特征和支持向量机的JPEG图像隐写分析方法。在JPEG图像的DCT域,采用联合概率密度计算DCT系数块内和块间相关性,联合低频AC系数统计特征,利用校准技术提取243维特征,然后使用支持向量机进行训练和分类。对三种常见的JPEG图像隐写算法(F5、Outguess、MB1)进行了二分类实验。实验结果表明:在嵌入率较低的情况,该算法的检测率优于其他算法。A JPEG image steganalysis algorithm based on calibrated DCT feature and support vector machine is proposed. In DCT domain of JPEG image, the joint probability density is used to calculate both the intra-block and inter-block correlations of DCT coefficient; combining the low-frequency AC coefficients statistical features, the calibration techniques is applied to extract the 243-dimensional features, and then the support vector machine is utilised for training and classification. Two classification experiments are carried out for three kinds of common JPEG image steganographic schemes （ FS, Outguess, MB1 ）. Experimental results demonstrate that the detection rate of the proposed algorithm is better than other algorithms under the condition of low embedding rate.

关 键 词：JPEG图像 联合概率密度 校准 SVM 隐写分析 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]