Abstract: In order to reduce the complexity of computation and improve the generalization of two binary-class support vector machines in images steganalysis, this paper brings forward a steganalysis method based on genetic algorithms and multiple one-class SVM. Genetic algorithm is applied to search and identify the potential informative features combinations for classification. The classification accuracy from the support vector machine classifier is used to determine the fitness in genetic algorithm. Experimental results show that the efficiency of detecting system is improved.

Key words: feature selection, genetic algorithm, multiple hypersphere, one-class SVM, image steganalysis

摘要： 提出一种基于遗传算法和多超球面一类支持向量机的隐秘图像检测方案。为了得到最能反映分类本质的特征从而有效实现分类识别，采用遗传算法进行图像特征选择，将支持向量机的分类效果作为适应度函数值返回，指导遗传算法搜索最优的特征选择方案。实验结果表明，与仅采用支持向量机分类而未进行特征选择的隐秘检测方案相比，该方案提高了隐秘图像检测的识别率。

关键词: 特征选择, 遗传算法, 多超球面, 一类支持向量机, 隐秘图像检测

CLC Number: 

• TP391