基于DCT域共生矩阵的JPEG图像隐写分析  

Steganalysis based on co-ocurrence matrix in DCT domain for JPEG images

作　　者：黄聪[1] 宣国荣[1] 高建炯[1] 施云庆 

机构地区：[1]同济大学计算机科学与技术系 [2]美国新泽西理工学院电气工程和计算机系

出　　处：《计算机应用》2006年第12期2863-2865,共3页journal of Computer Applications

基　　金：国家自然科学基金资助项目(90304017)

摘　　要：提出一种新的针对JPEG图像的通用隐写分析方法,该方法直接提取DCT系数,利用共生矩阵去挖掘出块中低频系数的相关性,最后形成120维特征并用SVM进行分类。针对4种公认的安全性较高的JPEG类嵌入方法F5、Outguess、MB(未去除分块特性MB1,去除分块特性MB2),对CorelDraw1096张图库进行了实验,结果表明,该方法的识别率和运算速度明显优于现有算法。A new steganalysis scheme based on co-occurrence matrix in DCT domain for JPEG images was proposed. A total of 120 dimensional feature vectors were derived from the co-occurrence matrix, which was calculated directly in DCT domain and was sensitive to the data embedding process. Then, SVM was used to classify the 120 dimensional feature vectors. The experimental results for 4 kinds of popular JPEG steganographic schemes （ F5, Outguess, Model based steganography with and without deblocking） have demonstrated that the proposed scheme outperforms the existing steganalysis techniques in both detection rate and speed.

关 键 词：隐写分析 JPEG图像 共生矩阵 SVM 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]