BS ISO IEC 19795 — Information Technology — Biometric Performance Testing and Reporting Series

🖐️ BS ISO/IEC 19795 is the British Standard family published by BSI adopting the ISO/IEC 19795 series on biometric performance testing and reporting. The series comprises six documents and was developed with participation from BSI Technical Committee IST/44 (Biometrics). It provides internationally harmonized principles, methods and reporting requirements for evaluating the performance of biometric recognition systems (fingerprint, face, iris, voice, etc.): from principles and framework, technology and scenario evaluation methodologies, and interoperability performance testing, to application-specific performance (such as access control), template update and operational performance evaluation. The series is the core standard family for biometric product selection, system acceptance, algorithm evaluation and regulatory compliance, and is widely applied in border control, public security, financial payment and identity management.

1. Series Overview and Positioning 📌

Evaluating the performance of biometric recognition systems raises unique challenges: error rates (such as false acceptance rate FAR and false rejection rate FRR) are small but never zero; sample correlation, population distribution and acquisition conditions have a large impact on results; and performance evolves over time with user interaction and template updates. The value of the ISO/IEC 19795 series is that it systematically defines the evaluation metrics (FAR, FRR, identification rate, error rates, etc.) and the evaluation methodology (technology evaluation vs scenario evaluation), and specifies test protocols, sample sizes, confidence intervals and a uniform reporting format, so that performance data from different organizations and algorithms are comparable and traceable.

2. Series Structure and Technical Content 📚

The six documents of the BS ISO/IEC 19795 series (of varying edition years) form a complete performance-evaluation system:

The internal logic of the series: Part 1 establishes the metrics and framework; Part 2 distinguishes “technology evaluation” (laboratory, algorithm-level) from “scenario evaluation” (terminal, user-level); Part 4 deals with interoperability (exchanging templates between different sensors/algorithms); Parts 5 to 7 cover advanced topics of access control, field operation and template update.

3. Key Technical Requirements 🔬

3.1 Principles and framework (Part 1)

Part 1 specifies the core concepts of biometric performance evaluation: it defines false acceptance (recognizing an impostor as the genuine user), false rejection (not recognizing a genuine user), and performance metrics for the two operating modes of identification and verification; it defines the difference between technology evaluation (algorithm/technology evaluation) and scenario evaluation; and it gives basic requirements for sample size, confidence intervals and report format. The part also standardizes the subject-sample relationships and data partitioning (genuine and impostor trial pairs).

3.2 Technology and scenario evaluation methodologies (Part 2)

Part 2 specifies the procedures of the two evaluation types in detail: technology evaluation uses standardized biometric datasets (including genuine and synthetic samples) and computes error-rate curves (such as DET curves) and the equal error rate (EER) under controlled conditions; scenario evaluation simulates real usage (user interaction, acquisition failures, enrolment and verification sequences) and reports system-level performance (such as FTA, FTER, FAR, FRR). The part stresses test protocols, sample representativeness, acquisition repeatability and statistical treatment.

3.3 Interoperability performance testing (Part 4)

Part 4 addresses the interoperability of biometric templates or characteristics: performance degrades when a template generated by one device/algorithm is verified by another device/algorithm. The part specifies the goals, coverage and metrics of interoperability testing (such as interoperability error rates), and “sufficiency” assessment (whether a template set is sufficient for a particular application). It underpins interoperability validation in multi-vendor environments (e.g., biometric passports, cross-system authentication).

4. Engineering Practice 🛠️

• 🔍 Product selection and acceptance: compare scenario-evaluation performance of different products based on the target application (1:1 verification or 1:N identification) and target error rates.
• 🔧 Algorithm R&D evaluation: build standardized benchmark datasets per Parts 1/2 and unify evaluation pipelines to track the benefits of algorithm iterations.
• 📊 Compliance and certification: organize third-party performance testing per the series to provide evidence for regulated applications such as border control and finance.
• 📋 Operations and template management: conduct operational performance monitoring and template-update evaluation per Parts 6/7 to maintain long-term recognition quality.

5. Frequently Asked Questions ❓