Abstract: In this paper, a new steganalysis for Model-based(MB) steganography of JPEG images is proposed. MB lead the veins distributing of image tending to be accordant, but airspace gradient can measure veins of image commendably. So the new method use airspace gradient by transform of MB steganography, combined with calibration and re-embedding technology. Experimental results validate that this method is good at detecting MB1 and MB2 with low false rate.

Key words: difference histogram, calibration, re-embedding, Model-based(MB) steganography

摘要： 基于模型(MB)隐藏会使图像的纹理分布趋于一致，使其难以被检测。针对该问题，利用空域差分直方图能较好度量图像纹理的特性，提出一种新的图像MB隐藏检测算法。分析JPEG图像空域差分直方图在MB隐藏前后的变化情况，结合校准和重嵌入技术，实现MB隐藏的有效检测。实验结果表明，该算法具有较高的检测率和较低的虚警率。

关键词: 差分直方图, 校准, 重嵌入, 基于模型隐藏

CLC Number: 

• TP309