School of Engineering and Digital Arts

Signature Biometric Systems: Usability Assessment


Biometric recognition for human identification and verification (using modalities such as face, iris and signature) are increasingly being deployed to improve security for a range of application domains including travel and national identification, physical access security and on-line management. The human signature modality is a key player in the biometrics market, with particularly strong uptake in financial, legal and retail security. This Overseas Travel Grant aims to facilitate an investigation into the usability of biometric signature devices through collaboration between the University of Kent, UK and Purdue University, US.

While considerable research has been conducted into the assessment of the algorithmic performance of biometric technologies, a range of new issues concerning the measurement of usability factors have emerged which require urgent attention following large scale deployment of public biometric systems. A vital component in implementing a biometric system is to ensure that it performs adequately to meet the security requirements of the application. Conventionally this has been measured using a number of error rate metrics to provide an indication of computer-based performance. These statistics do not, however, quantify other errors contributing to the overall performance, specifically i) relating to the environment within which the system is used and ii) how the test subject interacts with the system. As biometrics become more widespread testing methodologies and their associated evaluation metrics need to be re-examined. The work funded by the grant has enabled the development of a new biometrics user interaction model (HBSI - Human Behavioral Interaction Model) for the signature modality. The model allows developers to assess how and why an error occurs by establishing metrics for performance issues relating, separately, to user interaction errors, hardware/sensor failure to detect samples and misclassification of sample provenance. All of these metrics complement conventional performance statistics to enable a deeper understanding of the human and system factors relating to overall performance and show clear performance considerations specific to different signature technologies. This has wide implications for the performance assessment of signature devices. Importantly we have also theoretically assessed our model on other behavioral biometrics and complex interaction systems which use biometrics alongside other ID reading devices.

Team Members

School of Engineering and Digital Arts, Jennison Building, University of Kent, Canterbury, Kent, CT2 7NT

Enquiries: contact us

Last Updated: 23/01/2017