Research & Development
Our latest advances
MATCHINGER
As members of the academic community leading for change, we introduce innovative approaches to better fulfil market demands. Our R&D initiatives tend to aid the market to get into end to end next best solutions. This makes us share MATCHINGER as a futuristic tool to be tested in business environment in order to answer market demands.
MATCHINGER is an image source attribution forensic tool that identifies whether a picture was taken with a specific digital camera device.
All digital imaging sensors intrinsically present a unique pixel-wise noise pattern due to tiny imperfections on the silicon wafer during the manufacturing process. This noise pattern, called PhotoResponse Non-Uniformity (PRNU), is passively embedded by the imaging sensor every time a picture is taken. Although the PRNU is undetected to the human eye in the resulting picture, it can be extracted and exploited as fingerprint of a digital camera device, like the scratches on a bullet to identify the gun.
Features
- Image source attribution, batch processing, and clustering
- Fingerprint extraction and quality measure
- Cropping, rotation, and JPEG compression are supported
- Camera brand and model distinguishability
MATCHINGER extracts the PRNU of a given camera device from a set of images. Once the PRNU is extracted, any image suspicious of coming from the given device can be tested. A quality measure of the extracted fingerprint is reported, and more images can be added to the extraction to improve the fingerprint quality. Batch of images can be tested against the previously extracted PRNU obtaining very low false alarm rates.
Moreover, MATCHINGER is able to classify a dataset of images by determining clusters of possible source camera devices. Thus, clusters are built meaning that images belonging to the same cluster have likely been taken with the same device.
The PRNU is robust along time and survives JPEG compression, as well as MATCHINGER supports image cropping and rotation. MATCHINGER is also able to distinguish camera devices even when devices are of the same model and/or manufacturer.
Finally, the design of MATCHINGER meets the standard practice required in forensic science, ensuring that the derived algorithm is explainable, which is essential to guarantee the credibility of MATCHINGER in any legal context and also to establish its limitations.
