This is currently one of the principal and most vibrant areas of research in the Group, for which we have a recognised national and international reputation.
Our interests in biometrics – the identification of individuals from the detection and evaluation of their physiological or behavioural characteristics – are very wide, and encompass all aspects of biometric processing from individual biometric modalities (sources of identification data such as facial features, iris patterns, voice, etc) to systems integration issues, reliability and countermeasures to attack, and wider questions about system testing, operating standards, and so on.
We have a particular interest in developing enhanced processing algorithms for face recognition, for handwriting analysis (both from general writing style and through signature analysis), for fingerprint-based authentication and for identification through voice recognition. Much of our current work is studying multimodal biometrics to integrate identity-related evidence from two or more of these (or other) modalities, to enhance both reliability and flexibility in practical applications. We have built up a large multimodal database, and have carried out work to develop techniques to optimise empirically (using genetic algorithms) the integration of identification data from different modalities. We are currently beginning work to develop an approach which builds a general theoretical framework within which our more empirical techniques can be analysed and optimised.
The burgeoning complexity of (especially) multi-modal systems, particularly in typical application scenarios where different levels of access authorisation to data or locations interact with different levels of data sensitivity or entry authorisation, can create significant problems for implementation and system management. We have addressed this type of problem by developing an agent-based framework which can intelligently negotiate authorisation conditions and resolve conflicts. We are increasingly interested in the interaction of the human user with biometric systems, and have begun to develop techniques which can, on the one hand, assist the user to provide stable and high quality biometric data to minimise false rejection errors while, on the other hand, we are seeking new insights into the nature of human behaviour at a biometric system interface to maximise interaction effectiveness. Characterising user behaviour is critically important, both to maximise the effectiveness and robustness of biometric authentication, but also to ensure “inclusivity” for all users, without regard to physical, cultural or other factors which might affect the public acceptability of biometric technologies.. Our attention is now turning to important questions of protecting against attack of biometric systems, something which is becoming an increasingly important issue as introduction of the technology becomes more widespread. We are focusing our work on studying the extent to which synthetic biometric data (e.g. voice synthesis, the construction of artificial facial images, etc) can “spoof” current devices, and we are developing new approaches for the detection of such system attacks.
Our work in biometrics has been widely recognised. We have received external support through EPSRC and the DTI, from EU Research Programmes, and from individual companies for this work. Recent funded projects have included studies in signature verification, multi-modal systems, chip-based sensor integration, secure document processing, and biometrics-mediated encryption. We are part of the prestigious EU-funded BIOSECURE European Network of Excellence in Biometrics, and play an important role in DTI International Technology Watch Missions, in Committees developing biometrics standards, and in a range of national and international initiatives in the biometrics and security field.