基于条件概率的MFP量化隐藏的隐写分析算法

MFP隐藏方法将小波变换后的中频系数对分为重要系数与不重要系数,用不同的算法嵌入隐藏信息,且对原图像的小波分解的各子带系数的直方图统计特性影响并不明显。首先用有限状态机分析了不重要系数对的嵌入效果,接着利用中心极限定理的原理考察了重要系数对量化后的误差分布,最后,用条件概率的意义,设计了一个简单的统计量作为隐写分析的根据。实验结果表明该统计量适用于中频系数对隐写方法的检测。

Steganalysis Algorithm Based on Conditional Probability for Images Created by Middle Frequency Pair Hiding Scheme

Middle frequency pair(MFP) embedding algorithm adopts different quantization measures according to the coefficients classified into significant coefficients or not. Statistical discrepancies of wavelet sub-band coefficients of a stego image created by quantizating wavelet transform middle frequency pair coefficients between its original cover image is not obvious. First,the finite state machine is used to analyze the embedded effect for Least Important Coefficients. Second,in view of the Center Limit Theorem,the error distribution of Significant coefficients are studied. At last,an important and simple statistics is designed to detect the stego images created by MFP according to the principle of conditional probability. Experimental results indicate that this statistics is applicable for the detection of the quantization of wavelet coefficients embedding algorithm.