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 image's features.Using Support Vector Machine (SVM) as classifier, our experiment results show that ANOVA is effective and our method effectively improves accuracy.