基于相关性和SVD技术的盲隐写分析算法

Blind steganalysis algorithm based on correlation and SVD techniques

针对图像分块离散余弦变换(DCT)域加性噪声隐写和基于奇异值分解(SVD)技术,提出了一种新的盲隐写分析算法。分析研究了载体图像和掩秘图像统计特征,建立了能够全面反映DCT系数相关性的数学模型;采用SVD技术提取图像特征,构建特征向量和盲隐写分析判决函数。试验结果证明:该算法检测可靠率在90%以上,综合性能比一般的隐写分析方法有明显提高。

Aiming at the steganographic model with the additive noise in the block-DCT （Discrete Cosine Transform） domain, a new blind steganalysis algorithm based on the Singular Value Decomposition （SVD） techniques was proposed. A new mathematical model was constructed, accounting for inter and intra-block correlations of the DCT coefficients. Using the SVD technology, the features of the images were extracted, and the eigenvector was built. The discrimination function was presented for detection the secret messages of the digital images without knowledge of original cover images. The experimental results show that the reliability rate of the detection is 90%, and its general performance is superior to that of the general detection algorithm.