What is False Positive Identification Rate (FPIR)?
FPIR is the percentage of 1: N searches that yield an incorrect identity above threshold. This is obviously a bad thing. A high FPIR wastes analyst time and can provoke erroneous actions. Gallery size can make the problem worse – the larger the gallery, the more candidates to compare to, the more opportunities to be “confidently wrong.”
Control FPIR by pushing thresholds higher as gallery size increases, capping candidate lists, and instituting corroborating evidence before action is taken. Maintain strict liveness to keep spoofed probes from clogging the queue. If you’re using these results for a customer flow, embed the identification results within a larger identity verification process where high-stakes decisions require more than a face score to determine identity.
