摘要
通常基于统计特征的隐写分析算法特征针对性强,而通用隐写分析算法适应性强。结合两者的优点,提出了一种针对F5数字隐写技术的盲检测算法。根据F5算法对载体图像统计特性带来的影响,提取了21个特征。在分类器设计方面,选用了对样本数量和质量依赖性小的支持向量机。最后对不同训练样本下算法的识别能力进行了实验仿真。实验结果表明,使用高嵌入率载密图像进行训练能达到很好的检测效果,在虚警率为4%的条件下,对25%载密图像的检测结果都达到了95%以上。
Generally, steganalysis based on statistical characteristic has a strong pertinency, and a blind steganalysis can adapt itself to the steganography. Hereby, a steganalysis for breaking F5 on the basis of these merits is given. The 21 features are extracte for detecting. SVM is used to classify, which weakly depends on the quantity and quanlity of training samples. The experimental results that are done on the different conditions show that the algorithm woks well. Especially for detecting images with embedding rate as low as 25%, and the detecting rate keeps more than 95% when keeping the false alarm rate less than 4%.
出处
《计算机工程与设计》
CSCD
北大核心
2009年第5期1048-1050,1059,共4页
Computer Engineering and Design
基金
国家自然科学基金项目(60473022)。
关键词
隐写分析
F5
支持向量机
盲检测
统计特征
steganalysis
F5
support vector machine
blind steganalysis
statistical characteristic
