BLIND STEGANALYSIS USING WAVELET STATISTICS AND ANOVA

摘要：

Blind steganalysis detects hidden message without any knowledge about steganographic method.To implement blind steganalysis, statistical model based on high-order wavelet decomposition is built to capture statistical difference between cover images and stego images.However, not all wavelet statistics are able to reflect well statistical changes due to hidden message embedded.Analysis of Variance (ANOVA) is applied to test which wavelet statistics are more sensitive to hidden message.The statistics that are sensitive to hidden message are selected as images features.Using Support Vector Machine (SVM) as classifier, our experiment results show that ANOVA is effective and our method effectively improves accuracy.

关键词： Blind steganalysis Steganographic Wavelet statistic Analysis of variance Support vector machine

作者: SHUANG-HUAN ZHAN HONG-BIN ZHANG

作者单位: College of Computer Science, Beijing University of Technology, Beijing, China 100022

会议类型: 国际会议

会议名称: 2007 International Conference on Machine Learning and Cybernetics(IEEE第六届机器学习与控制论国际会议)

会议地点: 香港

会议语种:英文

页码: 2515-2519

在线出版日期: 2007-08-19（万方平台首次上网日期，不代表论文的发表时间）