Biometrics are increasingly deployed in domains ranging from social media authentication right up to border control. An important operational requirement for biometric systems is the supposed uniqueness and permanence of biometric records: physiological changes occurring between enrolment and verification are referred to as template ageing, and increase the likelihood of a misidentification. Its magnitude is hard to estimate, and the factors affecting it are relatively little studied.
This work proposes a measure of template ageing, called biometric permanence, and develops a methodology to estimate it in the presence of confounding factors. The measure is applied to a database of fingerprints obtained over a seven year period, using bootstrap resampling to obtain confidence intervals for the estimates of effect size. Fingerprint quality metrics are evaluated in terms of their ability to predict classification performance, and the subject-dependence of fingerprint quality is explored using the ideas of a biometric menagerie. Statistically significant demographic factors underlying biometric quality and template ageing are highlighted and discussed.
The results of this work may have implications for the procurement and administration of biometric systems: for example, in ensuring consistent performance across a broad population demographic, and in the choice of credential lifetime and re-enrolment policy.