BPCER

BPCER stands for Bonafide Presentation Classification Error Rate, a performance metric used in biometric system evaluation, particularly for presentation attack detection (PAD). It measures how often a genuine or real user presentation is incorrectly classified as an attack.

Calculation and interpretation: BPCER is calculated as the number of bona fide (genuine) presentations that are misclassified as attacks divided by the total number of bona fide presentations. In plain terms, it is the probability that a legitimate user is blocked or falsely flagged by the PAD system. A lower BPCER indicates better usability, as fewer real users are interrupted by security checks.

Relation to APCER and ACER: PAD evaluations typically report both BPCER and APCER, the latter being the Attack Presentation Classification Error Rate (the proportion of attack presentations misclassified as bona fide). There is often a trade-off between BPCER and APCER controlled by the decision threshold of the classifier. Some evaluations also compute ACER (the average of APCER and BPCER) to provide an overall error measure.

Standards and context: BPCER is defined and used within the framework of PAD performance evaluation, including ISO/IEC 30107 standards, which outline metrics such as APCER, BPCER, and ACER for assessing vulnerability to presentation attacks across modalities like face, fingerprint, and iris recognition.

Example: If 100 bona fide presentations are tested and 4 are incorrectly labeled as attacks, BPCER = 4%. If 200 attack presentations are tested and 30 are misclassified as bona fide, APCER = 15%. These figures would inform the balance between security and user convenience for a given system.

Overall, BPCER is a key indicator of user experience in PAD, complementing APCER to reflect both protection against attacks and the practicality of normal use.