FAR / FRR / EER
The three error metrics that describe a biometric system's accuracy trade-off — how often it wrongly accepts an impostor (FAR), wrongly rejects a legitimate user (FRR), and the threshold where the two rates meet (EER).
aka: False Acceptance Rate, False Rejection Rate, Equal Error Rate
FAR and FRR move in opposite directions as the matching threshold shifts, so a system can’t minimize both at once — a stricter threshold that rejects more impostors also rejects more legitimate users by mistake. EER, the single point where the two curves cross, is a convenient one-number comparison between systems, but real deployments almost never operate there: security-sensitive ones deliberately bias toward a lower FAR.