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.
出处
《计算机应用与软件》
CSCD
北大核心
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)
