Better measurement needed to asses recent advances in biometric bias mitigation

New research considers how to measure differences in false rejection rates (FRR) from biometric systems for various demographic groups, examining different methodologies for statistically testing these differences.

Professor Michael Schuckers from St. Lawrence University (SLU) presented the paper on bias in biometric authentication systems at the 2022 International Conference on Pattern Recognition (ICPR) in Montréal, Québec in August.

“With recent advances in deep learning obtaining hallmark accuracy rates for various computer vision applications, biometrics is a widely adopted technology for recognizing identities, surveillance, border control, and mobile user authentication,” reads the workshop description.

“However, over the last few years, the fairness of this automated biometric-based recognition and attribute classification methods have been questioned across demographic variations by media articles in the well-known press, academic, and industry research.”

The workshop argued that despite existing work in this field, there is an urgent need to explore the bias of existing biometric modalities and develop advanced methods for bias mitigation in existing biometric-based systems.

In particular, the research efforts analyze statistical modeling of the demographic bias of some biometric systems across race, age, and gender in different image spectrums, also assessing the impact of variables like lighting variations, make-up, and pose. The research also considers the trade-off between bias and accuracy of different mitigation techniques.

The workshop sought to increase awareness of demographic effects and recent advances, and provide common ground for discussion among academics, industry, and government.

Schuckers is lead author of the paper, which was written in collaboration with Clarkson University professors and the Center for Identification Technology Research (CITeR). It was funded by grants from both CITeR and the National Science Foundation.

The paper’s publication comes weeks after the U.S. General Services Administration (GSA) announced it was planning a research study to assess the impact of demographic factors on biometric and non-biometric proofing checks.

Article Topics

accuracy  |  biometric authentication  |  biometric bias  |  biometrics  |  biometrics research  |  Clarkson University