Challenge Competion for Image Forgery Detection (DEFALS)

The pervasiveness of communication networks has made our society has permeable to and dependent from information which is so easily disseminated. Digital images are a major tool of this information stream and are consequently a possible source of disinformation. Moreover, digital images can be easily edited and falsified by means of digutal technologies, without leaving obvious traces.

In most cases, in the image industry, the classic image editions such as cropping or contrast, luminosity and colorimetry edition are helpful and welcome. However, other kinds of image tempering can also be used as means of pressure and constraint in order to discredit and to misinform.

Nowadays, while creating a realistic falsified image is easy from the original picture, the detection of forgery is challenging in the absence of the original image. There is a need for the development of reliable and automated tools which will allow to check the integrity of suspect information, in cases where it may:

The DEFALS challenge competition aims to address this societal issue by:

The teams selected in response to this call will compete against each other in several stages, where their algorithm-based systems will be objectively evaluated, based on a mutual image data base, with shared criteria.

The main objective is to achieve an analysis system which allows an automated detection of tampering in images. Competitors will have to develop their systems against a test-corpus of images; in a second stage, they will have access to the evaluation-corpus and will have to provide their results to a third-party mandated by DGA. The data supplied in this challenge competition will show a great variety of shots (indoor and outdoor natural scenes, urban scenes and landscapes) with a large range of image quality (webcams, smartphones, compact and professional digital cameras).