Inference for performance evaluation of fingerprint identification systems based on a hierarchical random effects model
Sinjini Mitra · Communications in Statistics Case Studies Data Analysis and Applications · 2015
Biometric authentication is widely employed today, from immigration and border control to ensuring security on mobile devices. Just as it is important to devise efficient biometric systems, it is equally or even more important to evaluate their performance. This article introduces a novel method of performance evaluation, particularly to assess the “scalability” of fingerprint identification systems using a Bayesian inference technique based on hierarchical random effects models. We also extend this model to predict false alarm probabilities on “watch-lists.” Both of these aspects are important for national security applications of such systems. We illustrate our approach using three fingerprint databases.