CAJ | 학술논문

提出一种新的结合重压缩检测的JPEG图像多类隐写分析方法,实现一次压缩和重压缩图像中多种隐写算法的识别.首先基于DCT系数首位数分布规律,提出一种重压缩检测方法,然后从系数直方图、块内相关性、块间相关性和空域块效应中提取盲检测特征用于隐写分析,最后用支持向量机构造JPEG隐写算法多类检测器.实验结果表明,本文方法的重压缩检测性能明显优于已有方法,且对嵌入改变量的鲁棒性较强,隐写分析特征不仅维数较低而且具有更好的检测性能,构造的多类隐写分析器能较好地识别JPEG隐写算法.
제출일충신적결합중압축검측적JPEG도상다류은사분석방법,실현일차압축화중압축도상중다충은사산법적식별.수선기우DCT계수수위수분포규률,제출일충중압축검측방법,연후종계수직방도、괴내상관성、괴간상관성화공역괴효응중제취맹검측특정용우은사분석,최후용지지향량궤구조JPEG은사산법다류검측기.실험결과표명,본문방법적중압축검측성능명현우우이유방법,차대감입개변량적로봉성교강,은사분석특정불부유수교저이차구유경호적검측성능,구조적다류은사분석기능교호지식별JPEG은사산법.
@@@@A novel multi-class steganalyzer with recompression detection for JPEG images is proposed to identify steganographic methods from singly and doubly compressed stego images. Based on the statistical distribution of the first digits of DCT coefcients, a JPEG recompression detection method is proposed. Fea-tures for blind detection are extracted from the histogram, intrablock correlation, interblock correlation and spatial blockiness. Finally, a multi-class detector against current steganographic methods is constructed with support vector machine. The experimental results show that the proposed recompression detection scheme outperforms the existing methods significantly, and is robust to embedding changes. The low dimensional steganalytic features have better performance, and the multi-class steganalyzer can identify the current JPEG steganographic methods reliably.