Play CSI with the IEEE SPS IFS TC

Digital images are everywhere. However, the way people deal with digital images may raise several legal and ethical issues that need to be addressed. Before legal action can be taken based on an image, whose content truthfulness may be questionable, it is necessary to provide some forensic evidence about the image itself.
In the last years, the field of Digital Image Forensics has emerged with several tools and algorithms presented to help researchers and forensics experts to assess the authenticity of digital documents. However, the lack of a clear benchmark and an established comparison protocol resulted in the absence of evaluation of such algorithms under real-world conditions.
In this context, the IEEE Information Forensics and Security Technical Committee (IFS-TC) launched the First IFS-TC Image Forensics Challenge, a detection and localization forensics contest, open to students, researchers and practitioners interested in role-playing a forensic analyst.

Aims

With this challenge, the IFS-TC has three major objectives in mind:

• Providing the community an open data set and protocol to evaluate the latest image forensic techniques;
• Benchmarking the current state-of-the-art techniques with respect to their ability to detect image forgeries (content manipulation);
• Establishing a standard protocol as a common comparison ground truth for new techniques.

Rules

The First IFS-TC Image Forensics Challenge will happen in two phases.
Phase 1 (tampering detection): All competing challengers submit a binary answer for each test image (0 for "pristine/non-manipulated" and 1 for "fake").
Phase 2 (tampering localization): All competing challengers submit a binary map file for each test image (white for "pristine/non-manipulated", black for "fake" pixels). The objective here is to evaluate how good are current techniques at localizing possible manipulations given a manipulated image. Only test fake images will be evaluated.

Challenge Corpus

Several original images of various scenes, both indoor and outdoor, were taken with different digital cameras. The corpus is then partitioned into images referred to as "pristine" (non-manipulated) on one hand, and "forged" or "fakes" on the other hand. "Forged" images are obtained using a set of different manipulation techniques such as copy/pasting and splicing with
different degrees of photorealism.
A training set will be made available with the associated ground truth so that the challengers can tune their algorithms. The challenge will then be run with a disjoint testing set for evaluation.

Awards

The Winner and Runner-Up of each phase will receive certificates and financial grants (at least $1k) to attend WIFS’13 in Guangzhou, China.

Email contact: ifc@ic.unicamp.br