A Performance Assessment Framework for Mobile Biometrics

This project aims to develop and explore a robust framework for assessing biometric systems on mobile platforms, where data is often collected in non-constrained, potentially challenging environments. The framework enables the performance assessment given a particular platform, biometric modality, usage environment, user base and required security level.

The ubiquity of mobile devices such as smartphones and tablets has increased access to Internet-based services across various scenarios and environments. Citizens use mobile platforms for an ever-expanding set of services and interactions, often transferring personal information, and conducting financial transactions. Accurate identity authentication for physical access to the device and service is, therefore, critical to ensure the security of the individual, information, and transaction.

Biometrics provides an established alternative to conventional authentication methods. Mobile devices offer considerable opportunities to utilise biometric data from an enhanced range of sensors alongside temporal information on the use of the device itself. For example, cameras and dedicated fingerprint devices can capture front-line physiological biometric samples (already used for device log-on applications and payment authorisation schemes such as Apple Pay) alongside voice capture using conventional microphones.

Understanding the performance of these biometric modalities is critical to assessing suitability for deployment. Providing a robust performance and security assessment given a set of deployment variables is critical to ensure appropriate security and accuracy. Conventional biometrics testing is typically performed in controlled, constrained environments that fail to encapsulate mobile systems' daily (and developing) use.

This thesis aims to develop an understanding of biometric performance on mobile devices. The impact of different mobile platforms, and the range of environmental conditions in use, on biometrics' accuracy, usability, security, and utility is poorly understood. This project will also examine the application and performance of mobile biometrics when in motion.