A new security product can fingerprint hardware to identify source devices. BCS, The Chartered Institute for IT's Brian Runciman MBCS spoke to Chang-Tsun Li of Functional Technologies, one of the finalists in the BCS IT Awards 2011, and Ahmad Ryad Soobhany of Forensic Pathways about their approach to an interesting security problem.

UK IT Industry Awards 2011 LogoAnalysis of images is a well embedded skill, but Functional Technologies forensic image analyser (FIA) in collaboration with Forensic Pathways is capable of using the enhanced sensor pattern noise (SPN) extracted from images to identify specific source devices, verify content integrity and blindly classify images into groups such that each group corresponds to an unknown device.

The novelty of this product lies in the sensor pattern noise enhancer, which is currently the only method capable of preventing scene details from distorting the SPN and facilitating the forensic applications effectively.

The SPN enhancer gives the FIA several innovative functionalities. For example, source device identification. Due to manufacturing imperfection, semi-conductor sensors of digital imaging devices, such as cameras, scanners and camcorders, leave unique and minute sensor pattern noise in the images. Like human fingerprints, these are unique and can thus identify the source device. Dr. Chang-Tsun Li has developed a SPN enhancer (a novel mathematic model) that can effectively attenuate the interference of scene details.

‘In traditional source device identification, the investigator usually has a collection of devices or a database of reference SPNs, each representing one device,’ says Chang-Tsun Li. ‘However, the more challenging task of source camera linking is about establishing whether two images are taken by the same camera or not without the availability of the camera's reference SPN.

'This can only be carried out based on one SPN from each image. If any SPN is severely contaminated by scene details, the chance of reaching a correct conclusion cannot be expected to be high. The SPN Enhancer is capable of solving this problem effectively.’

An even more challenging task addressed by this innovation is blind image classification, which aims to classify a large set of images in the absence of the imaging cameras. Given a large number of images, classification based on the SPNs from images of full size is computationally prohibitive. This entails the need for SPNs from smaller blocks cropped from the images. However, cropping reduces the number of SPN components, making the SPN less discriminative, especially for images with strong details.

The system also allows content integrity verification: Given an image claimed to have been taken by a specific camera and to be free from tampering, one way to verify its integrity is to partition the image into small blocks and compare their SPNs to the corresponding blocks the camera's reference SNP. Low SPN similarity indicates high likelihood of tampering. The smaller the blocks, the better localised tampering (such as object addition/deletion) can be detected.

Success to date

The FIA has been deployed by the FPL and used by a number of police forces in the UK, France and Australia. A new Eurostar project aims to develop an integrated multimedia forensic system that incorporates the Forensic Image Analyser and a new video identification technology (developed by Videntifier) for combating crime and IPR infringement.

Functional Technologies has successfully secured a feasibility study grant from the Technology Strategy Board in 2011 to conduct further investigation into to the feasibility of reducing demosaicing distortion in order to further increase the commercial value of the SPNE and FIA.

Functional Technologies was a finalist in the BCS IT Awards Innovation and Entrepreneurship section.

BCS UK IT Industry Awards
Wikipedia: Demosaicing