The watchful forensic analyst: Multi-clue information fusion with background knowledge

Image Forensics (IF) is a challenging research topic, that suffers from strong limitations when facing with real world applications. A possible way to cope with these limitations is to resort to data fusion, whereby the outputs of different forensic tools are used to reach a final decision about the analyzed image. Nevertheless, existing schemes do not take full advantage of all the information available to the analyst, like the knowledge of the dependence of the performance of forensic tools on side conditions. Specifically, in this paper we show how the performance of forensic tools varies according to a number of parameters, most of which are directly observable by the analyst. After showing some practical examples, we propose a method to cast this background information into two multi-clue information fusion frameworks, yielding a significant improvement of the overall performance at virtually no cost.