Canadian judge sends face biometrics use by govt in refugee claim to review

A judicial review of Canada’s use of facial recognition in evaluating immigration claims has been ordered by a federal judge in a case involving a pair of refugee claimants, Georgia Straight reports.

The Refugee Protection Commission found that the women were not Somalian citizens as they claimed in their application, but rather Kenyans who had submitted study permits under assumed names. This decision was based on biometric matches the agency obtained with Clearview AI’s app.

Attorneys for the two Somalian women who claim their refugee claims were wrongly rejected submitted the Gender Shades paper published in 2018 by in the Proceedings of Machine Learning Research.

That paper evaluated facial analysis software and found much higher rates of misclassification when the algorithms were applied to females with darker skin. It did not mention Clearview or test face biometric recognition, but rather gender and race classification.

Clearview CEO Hoan Ton-That has told Biometric Update that the company’s software has statistically minimal bias.

The RPD used its Global Case Management System to determine that the Kenyan women admitted on study permits did not attend the educational institutions they claimed to be enrolling in. Then- Minister of Public Safety and Emergency Preparedness Bill Blair applied to have the women’s refugee status vacated based on the finding that they had misrepresented their identities.

The match is contested by the plaintiffs, who allege that the actual Kenyan students are presumably still studying in Manitoba. They also say the evidence against them “likely came from questionable facial recognition software such as Clearview AI,” and should not be admitted.

Justice Avvy Yao-Yao Go ruled that the RPD wrongly relied on the Privacy Act to admit the biometric photo comparisons, and should not have exempted Blair from disclosing them. Go also ruled that the RPD ignored evidence counter to its conclusion, and that the facial similarities are inadequate to reject the claims.

Article Topics

accuracy  |  biometric bias  |  biometric matching  |  biometrics  |  Canada  |  facial recognition  |  immigration