基于D-S证据理论的融合图像隐写分析

Fusion image steganalysis based on Dempster-Shafer evidence theory

提出一种基于证据权的D-S证据理论的图像隐写分析方法.首先在空域,离散余弦变换(DCT)域和离散小波变换(DWT)域分别提取图像特征并各自进行预分类;然后对各域分类结果进行基本概率分配并进行证据权修正,利用D-S组合规则计算融合概率分配函数,形成最终的决策级融合分类结果.针对典型的隐写方法(如F5,JPHide,Jstego和YASS算法)进行检测,仿真结果显示,所提出的方法能显著提高单分类器的性能.

Based on evidence weight and Dempster-Shafer（D-S） evidence theory, an image steganalysis scheme is presented. The image is classified predictively by the characters exffacted from spatial, discrete cosine transform（DCT） and discrete wavelet transform（DWT） domain respectively. The basic probability assignments of varies classified results are assigned and modified by evidence weight. Then the fusion probability assignment function is computed by Dempster＇s combinational rule, and the last decision level fusion classify result is obtained. The detection works are presented to attack typical steganographical schemes such as F5, JPHide, Jstego and YASS. The simulation results show that the presented method can significantly improve the performance of single classifier.