Comments sought on working draft of ISO face biometrics fairness testing standard

The International Standards Organization has approved a working draft of the international standard for assessing facial recognition performance across different demographic groups.

Working Draft 4 of ISO 19795-10 sets out proposed methods for testing and measuring the differences in the performance of face biometric systems commonly referred to as bias.

Maryland Test Facility Principal Data Scientist John Howard announced the release of the working draft on LinkedIn, and invited organizations and labs working on fairness and face biometrics to read and comment on it.

“We hope this standard will inform (increasingly mandatory) testing efforts across the globe,” Howard writes. “This draft is not only significantly streamlined but also steadily approaching a finished product.”

“In recent years, many groups around the world have started to focus on the equitability of biometric systems, particularly face recognition,” Howard, editor of the 19795-10 standard, told Biometric Update. “To date, there has been no standard approach for organizations to follow when measuring differences in biometric system performance across demographic groups. The 19795-10 standard is the first effort to develop a consensus approach for measuring demographic differentials in biometric systems. Engaging international experts through the ISO process ensures that the standard uses only technically sound and well-vetted methodologies to support future demographic testing efforts.”

Howard’s colleague at the Maryland Test Facility, Technical Director Yevgeniy Sirotin, also serves as co-editor of the standard.

Those interested in providing feedback can get access to the Working Draft through their national standards bodies.

The standard in development is supported by a broad base of research, including a pair of papers recently published from experts at MdTF.

Comments are due by October 3, ahead of a release planned for 2023.

Article Topics

biometric testing  |  biometric-bias  |  biometrics  |  demographic fairness  |  face biometrics  |  ISO standards  |  Maryland Test Facility (MdTF)  |  standards