摘要： 对自适应隐写的安全性问题进行分析，提出一种基于自然图像的高斯混合模型分析方法。在总嵌入强度相同的条件下，比较自适应和非自适应扩频隐写载密随机变量概率密度函数的特征函数，验证自适应扩频隐写的统计安全性高于等嵌入强度下非自适应扩频隐写。分析结果表明，该方法能为提升信息隐藏系统的抗统计分析性能提供理论依据。

关键词: 自适应隐写, 安全性分析, 高斯混合模型, 扩频隐写, KL散度

Abstract: Aiming at the security problem of adaptive steganography, the analysis method based on Gaussian mixture model(GMM) of real image is proposed. Compared the stego random characteristic function of probability density function between the adaptive Spread Spectrum Image Steganography(SSIS) and the general non adaptive one under the condition that the total embedding intensity is equal, it demonstrates that the security of adaptive SSIS is higher than that of non-adaptive schemes. Analysis result shows that the method provides theoretical evidence for using adaptive scheme to improve statistical imperceptibility of steganography.

Key words: adaptive steganography, security analysis, Gaussian Mixture Model(GMM), Spread Spectrum Image Steganography(SSIS), Kullback-Leibler(KL) divergence

中图分类号: 

• TP309