Funding Body: EU Seventh Framework Programme (FP7 Marie Curie Actions: Industry-Academia Partnerships and Pathways (IAPP))
Consortium: University of Warwick (UK), Forensic Pathways Ltd (UK), UniversitÃ degli Studi Roma TRE (Italy), XLAB d.o.o. (Slovenia)
Consortium Coordinator: Dr Chang-Tsun Li, University of Warwick
Project Duration: May 2010 - April 2014
Budget: EUR 1,559,325
Link to EU Archive: Click here
This project is about transferring knowledge of image and video forensics between the academia and industry. Building on the complementary expertise of the participants, the consortium aims to pursue two major lines of investigations: Device fingerprint based forensics and hidden data based forensics.
- The first line of investigation requires the formulation of a set of fingerprints left in the images/video by the hardware components or in-built signal processing algorithms of the imaging device and involves the study of its forensic applications. The applications we are aiming for include device identification, technology/licensing infringement detection, device linking, automatic media classification, tampering detection. To guarantee the value of these techniques, anti-anti-forensics measures will also be devised to detect/prevent removal or substitution of the fingerprint set.
- The second line of investigation is about the use and analysis of hidden data in the host image/video for authentication, content integrity verification, copyright protection (including ownership identification, proof of ownership, copy control, traitor tracing), and the use and detection of covert communications. Although data hiding is a relatively mature research area, their applicability in the real world is yet to be fully explored due to the fact that the security issues are often interleaved with multimedia processing issues and requires addressing. Therefore this line of research requires the formulation of specific security requirements and attack modelling for specific applications.
Overall, both lines of research require a wide variety of expertise such as multimedia signal processing, computer vision, pattern recognition, machine learning, and optimisation theory. Therefore it is expected that this project will also take the soundness and applicability of these theories to a new level and lay a solid foundation for wider lasting collaborations in the related areas.
In order to use digital images and video as evidence in legal cases it is important to be able to link the images/video in question to the source devices or other images/video which may have been taken by the same devices and to prove that they have not been tampered with. On the other hand, with the wide availability of user-friendly editing tools, a wrongdoer may well tamper with the content of a photo in order to mislead investigation or disguise evidence. As a result, authentication and content integrity verification have become acute challenges in many sectors, such as law enforcement, the insurance industry, security and the defence agencies, etc. To meet the needs of forensic authentication the analysis of a digital fingerprint left in the media by the devices has emerged as a new research discipline in the last few years. The state-of-the-art is to use sensor pattern noise caused by imperfections during the sensor manufacturing stage, lens aberration, camera response function and algorithms or parameters of camera components such as colour filter array, colour interpolation matrices and quantisation tables for acquisition device identification, integrity verification and authentication. However, all these methods use only one type of device fingerprint, which limits their applicability (e.g., some require that specific assumptions be satisfied while others are vulnerable to manipulations such as fingerprint removal or substitution.
The afore-mentioned new discipline of multimedia forensics can be called passive forensics because the investigator relies on what is left incidentally by the devices. By contrast, digital watermarking and steganography rely on the secret data/watermark embedded deliberately by humans for multimedia forensics or covert communications. As for digital watermarking, the state-of-the-art is to embed a small amount of data/watermark, usually invisible with some exceptions, into a host media so that it can be extracted later for forensic applications such as copyright protection, authentication, and integrity verification. For copyright protection, usually the watermark is made robust so that it can only be destroyed when the content it is intended to protect is also destroyed or seriously damaged. For applications such as authentication and integrity verification, the state-of-the-art is to make the watermark fragile so that when its host media is tampered with, the watermark is destroyed. Failing to extract the original watermark indicates that the host media is no longer trustworthy. Steganography is about hiding a secret message in the content of the host media in order to disguise the existence of communication. This is useful in covert communications and monitoring the distribution of stego-media (i.e., the media carrying hidden data) in order to infer criminal networks. However, the benefits steganography has to offer may as well be exploited for malicious purposes such as terrorism. Although these two closely related disciplines are relatively more mature than passive forensics, their real-world applicability has yet to be fully explored.
- To transfer existing image and video forensics techniques from Warwick and RomaTRE, to Forensic Pathways (FPL) and XLAB.
- To transfer commercialisation and innovation techniques from the participating companies to the two universities to improve the quality and relevance of their research.
- To formulate a set of device fingerprints for identifying acquisition devices, detecting licensing infringement, tampering detection, image/video classification and device linking.
- To devise a set of anti-anti-forensics measures to detect fingerprint removal or substitution
- To address security issues of digital watermarking to improve their applicability.
- To characterise steganographic schemes and the anomalies they introduce into the stego-media so as to design effective steganalysis systems for detecting covert communications and revealing criminal networks.