Biometric Testing: aligning standards and practice

The performance of Biometric recognition systems
has improved of the past ten years. For application it is
necessary to know how well they perform in specific operational
circumstances. This requires performance evaluation based on
scientific principles.
The Data Management and Biometrics (DMB) group at the
University of Twente, frequently evaluates biometric algorithms
for third parties in the public and private sector. We have
encountered mismatches between the stakeholders’ interest and
what is being prescribed in standards and guidelines that apply
to biometric performance testing, and reporting of its results.
In this paper we show how standards and practice can be
better aligned. First we explore the problem through case studies,
and suggest solutions for the issues we find. Given these results
we propose a systematic approach to analysing and reducing
misalignment. Our goal is threefold:
• increase our own awareness whether biometric performance
evaluations are in line with applicable standards; i.e., learn
and further improve our testing skills
• Investigate whether the standards reflect the diversity in
forms and objectives evaluations have
• suggesting changes to standards so they align more easily
with practice and/or developing additional guidelines to
facilitate standardised practice in specific cases