A new steganalysis paradigm based on image retrieval of similar image-inherent statistical properties and outlier detection

Conventional steganalysis method generally encounters the problems of embedding algorithm mismatch (EAM) and cover source mismatch (CSM). These problems cause difficulties in the use of steganalysis in the real world. Learning from the idea of image pre-classified, this study presents a JPEG steganalysis paradigm combining the similarity retrieval of image-inherent statistical properties (IISP) and unsupervised outlier detection. First, cover images with similar IISP to the test image are searched from massive image database to establish an aided sample set. Outlier detection is then performed in a test set composed of the test image and its aided sample set to judge the type of the test image. Experimental results show that the proposed paradigm can effectively avoid EAM and CSM. It demonstrates better performance than the steganalysis strategy using a mixed image set as the training sample. The proposed method has high detection efficiency with the unsupervised outlier detection.