Steganalytic Scheme for JPEG Images Based on Multivariate Regression
You Xin-gang · Dianzi xuebao · 2009
A co-occurrence feature space is constructed,which can effectively reflect the changes of higher-order statistical characteristics caused by JPEG steganographic techniques.Supported by this feature space,a new scheme for designing steganalytic algorithms is presented based on multivariate regression model.With only JPEG stego images and no embedding details needed,this designing scheme is able to provide a good generality.Experimental results indicate that,our designing scheme can also produce remarkable steganalytic algorithm even for MB1 embedding mechanism,which has reached high performance of security.