Joint detection of full-frame linear filtering and JPEG compression in digital images

Image forensics aims at detecting clues regarding the history of a digital image. However, many forensic techniques present some limitations in real life scenarios, where the application of post-processing operators may alter the characteristic footprints exploited by forensic tools. In this paper we study the combination of JPEG compression and full-frame linear filtering, analyzing their impact on the statistical distribution of the Discrete Cosine Transform (DCT) coefficients of the image. We extract characteristic features from the DCT distributions and build an effective classifier able to jointly disclose the applied compression quality factor and the filter kernel. The proposed technique is computationally efficient and effective, as demonstrated by the extensive experimental analysis.