Can a “poor” verification system be a “good” identification system? A preliminary study

The matching accuracy of a biometric system is typically quantified through measures such as the False Match Rate (FMR), False Non-match Rate (FNMR), Equal Error Rate (EER), Receiver Operating Characteristic (ROC) curve and Cumulative Match Characteristic (CMC) curve. In this work, we analyze the relationship between the ROC and CMC curves, which are two measures commonly used to describe the performance of verification and identification systems, respectively. We establish that it is possible for a biometric system to exhibit “good” verification performance and “poor” identification performance (and vice versa) by demonstrating the conditions required to produce such outcomes. Experimental analysis using synthetically generated match scores confirms our hypothesis that the ROC or CMC alone cannot completely characterize biometric system performance.