作　　者： ; ; ;

机构地区： 中山大学信息科学与技术学院

出　　处： 《计算机学报》 2006年第9期1702-1710,共9页

摘　　要： 实现了针对由Su等人提出的JPEG2000 Lazy-mode隐写术的可靠检测.在理论和实验分析的基础上,文章揭示了由Lazy-mode隐写术生成的掩密图像,其子带代码块噪声方差序列的振荡特征异于非掩密含噪图像的子带代码块噪声方差序列.因此,此文隐写检测算法的关键在于针对这两种子带代码块噪声方差序列进行序列分析,提取它们内在的振荡特征差异.在序列分析中,通过引入Hilbert-Huang变换,对噪声方差序列进行经验模式分解,构建了基于Hilbert谱的特征向量.实验表明,基于该特征向量的支持向量机(SVM)分类器能以平均90.6%的准确率识别掩密图像.根据检索,目前尚未有对JPEG2000 Lazy-mode隐写术进行成功分析的报道,因此,该文具有重大意义. In this paper, the authors present a steganalytic method to attack JPEG2000 lazy-mode steganography proposed by Suet al. With theoretical analysis and experiments, the authors discover that the vibrations of the code-block noise variance sequences of stego images generated by lazy-mode steganography are different from those of images without hiding message. So the key issue of the proposed steganalytic method is the analysis of these two categories of code-block noise variance sequences, which can distill the differences of the vibrations of them. The Hilbert transform based characteristic vectors are constructed via empirical mode decomposition of the sequences. The SVM classifier is applied to classification. The experimental results have demonstrated the correct rate to detect the hiding message achieves 90.6%. According to the best knowledge, there have not been any reports about successful attack against the JPEG2000 lazy-mode steganography before.

关 键 词： 隐写分析 变换 经验模式分解 序列分析 支持向量机

领　　域： [自动化与计算机技术] [自动化与计算机技术]