基于小波分解和方差分析的图像信息隐藏盲检测

Image-Based Blind Steganalysis Using Wavelet Statistics and Analysis of Variance

为了实现图像中隐藏信息的盲检测,建立高阶统计模型,提取高阶小波统计量捕获原始图像和隐藏图像之间的统计差异;方差分析用于检验所提取的小波统计量对隐藏信息的敏感程度。应用方差分析选取出对隐藏信息较敏感的小波统计量作为图像的特征向量元素,基于核技巧的支撑向量机(SVM)用作原始图像与隐藏图像之间的分类器,实现图像中隐藏信息的盲检测分析。实验结果及分析表明本文的方法能较有效地实现信息隐藏的盲检测分析。

Applying wavelet decomposition to build high-scale statistical model for capturing statistical difference between cover images and stego-images. However, not all wavelet statistics are able to reflect well statistical change due to hidden message embedded. By exploring analysis of variance, the statistics that are more sensitive to hidden message are chosen as features of images. Kernel-based support vector machine is chosen as classifier to implement blind steganalysis of images. Experiment results show that our method can reach a high testing rate to hidden message of images.