Organizer: CASED, CROSSING and EC SPRIDE
Multimedia forensics (MF) is a new discipline aiming at collecting evidence about the past history of multimedia documents, including the identification of the source of the document, the distinction between computer generated and natural images, the detection of traces left by the application of certain processing tools like resampling or JPEG compression, the modification of the semantic content of the document through cut and paste or copy-move operations.Early works did not consider the presence of an adversary aiming at impeding the forensic analysis; as a result most techniques do not work properly if some simple countermeasures (collectively referred to as anti-forensic techniques) are taken. In an attempt to re-establish the validity of forensic analysis, researchers has started building new tools to detect the traces left by anti-forensic algorithms.
It is evident that any attempt to improve the forensic analysis will be accompanied by a dual effort to devise more powerful anti-forensic techniques that leave less and less evidence into the forged documents. While this is an unavoidable and possibly virtuous loop that will finally lead to more powerful forensic and anti-forensic tools, the need to investigate the ultimate limits of forensic analysis clearly exists. It is the goal of this talk to review some recent results that by relying on a bunch of disciplines including game-theory, hypothesis testing and information theory have started laying the basis for a rigorous theory of MF in which the presence of an adversary is explicitly taken into account.