Operational Measures and Accuracies of ROC Curve on Large Fingerprint Data Sets

Published

May 8, 2008

Author(s)

Jin Chu Wu

Abstract

At any point on a receiver operating characteristic (ROC) curve in combination with two distributions of genuine scores and impostor scores, there are three related variables: the true accept rate (TAR) of the genuine scores, the false accept rate (FAR) of the impostor scores, and the threshold. Any one of these three variables determines the other two variables. The measures and accuracies of TAR and threshold while FAR is specified and the measures and accuracies of TAR and FAR once threshold is fixed are all investigated. In addition, the measures and accuracies of the equal error rate (EER) and the corresponding threshold are also explored. From the operational perspective, these exhaust all possible scenarios. The nonparametric two-sample bootstrap based on our previous studies of bootstrap variability on large fingerprint data sets is employed to compute the measurement accuracies. Four high-accuracy and two low-accuracy fingerprint-image matching algorithms invoking different types of scoring systems are taken as examples.

Citation

NIST Interagency/Internal Report (NISTIR) - 7495

Report Number

7495

NIST Pub Series

NIST Interagency/Internal Report (NISTIR)

Pub Type

NIST Pubs

Download Paper

https://doi.org/10.6028/NIST.IR.7495

Keywords

Receiver operating characteristic (ROC) curve, Fingerprint matching, Nonparametric two-sample bootstrap, Standard errors, Confidence interval, Equal error rate

Citation

, J. (2008), Operational Measures and Accuracies of ROC Curve on Large Fingerprint Data Sets, NIST Interagency/Internal Report (NISTIR), National Institute of Standards and Technology, Gaithersburg, MD, [online], https://doi.org/10.6028/NIST.IR.7495 (Accessed October 2, 2025)