基于二重扰动SVM集成分类的JPEG图像隐写检测

JPEG Image Steganalysis Method Based on the SVM Ensemble Using Double Disturbance

针对目前大部分JPEG隐写分析方法主要采用单一支持向量机分类器,可能产生泛化能力恶化、分类精度难以提高等问题,提出了一种新的基于二重扰动SVM集成分类的JPEG图像隐写检测方法.利用多项式核函数和高斯核函数构造混合核函数模型,并用主成分分析法进行特征变换,去除冗余信息,最后在变换后的特征空间上进行模型参数和特征的二重扰动产生成员分类器,并用多数投票法对它们进行组合.实验结果表明,与传统方法相比,具有更高的JPEG隐写图像检测率.

In view of the current steganalysis method of JPEG image mainly uses a single sup- port vector machine （ SVM）, existing the problem that generalization ability deteriorate and classifi- cation accuracy is difficult to improve, a new JPEG image steganalysis method is proposed based on the SVM ensemble using double disturbance. The method constructs a mixed kernel function using polynomial function and gaussian kernel function, and removes the redundan feature using principal component analysis method for feature transform. Finally member classifier is generated by distur- bing feature space and model parameters, and the finial decision is made by the majority voting pro- cedure. Experimental results show that this method effectively improves the detection rate of stega- nalysis in JPEG images.