Forgery detection from printed images: a tool in crime scene analysis

.The preliminary analysis of the genuineness of a photo is become, in the time, the first step of any forensic examination that involves images, in case there is not a certainty of its intrinsic authenticity. Digital cameras have largely replaced film based devices, till some years ago, in some areas (countries) just images made from film negatives where considered fully reliable in Court. There was a widespread prejudicial thought regarding a digital image which, according to some people, it cannot ever been considered a legal proof, since its “inconsistent digital nature”. Great efforts have been made by the forensic science community on this field and now, after all this year, different approaches have been unveiled to discover and declare possible malicious frauds, thus to establish whereas an image is authentic or not or, at least, to assess a certain degree of probability of its “pureness”. Nowadays it’s an easy practice to manipulate digital images by using powerful photo editing tools. In order to alter the original meaning of the image, copy-move forgery is the one of the most common ways of manipulating the contents. With this technique a portion of the image is copied and pasted once or more times elsewhere into the same image to hide something or change the real meaning of it. Whenever a digital image (or a printed image) will be presented as an evidence into a Court, it should be followed the criteria to analyze the document with a forensic approach to determine if it contains traces of manipulation. Image forensics literature offers several examples of detectors for such manipulation and, among them, the most recent and effective ones are those based on Zernike moments and those based on Scale Invariant Feature Transform (SIFT). In particular, the capability of SIFT to discover correspondences among similar visual contents allows the forensic analysis to detect even very accurate and realistic copy-move forgeries. In some situation, however, instead of a digital document only its analog version may be available. It is interesting to ask whether it is possible to identify tampering from a printed picture rather than its digital counterpart. Scanned documents or recaptured printed documents by a digital camera are widely used in a number of different scenarios, from medical imaging, law enforcement to banking and daily consumer use. So, in this paper, the problem of identifying copy-move forgery from a printed picture is investigated. The copy-move manipulation is detected by proving the presence of copy-move patches in the scanned image by using the tool, named CADET (Cloned Area DETector), based on our previous methodology which has been adapted in a version tailored for printed image case (e.g. choice of the minimum number of matched keypoints, size of the input image, etc.) In this paper a real case of murder is presented, where an image of a crime scene, submitted as a printed documentary evidence, had been modified by the defense advisors to reject the theory of accusation given by the Prosecutor. The goal of this paper is to experimentally investigate the requirement set under which reliable copy-move forgery detection is possible on printed images, in that way the forgery test is the very first step of an appropriate operational check list manual.