Racist face-recognition algorithm integrated into anti-cheating software

The Proctorio program, which monitors students to prevent cheating on exams, in half the cases fails to identify the faces of Black users. пишет Motherboard.

Satisan decided to study Proctorio’s facial-recognition methods and found that they operate identically to OpenCV — an open-source computer-vision algorithm.

He conducted several tests using OpenCV and a dataset designed to evaluate the accuracy of computer-vision algorithms on faces of different nationalities.

It turned out that the algorithm not only failed to recognise Black faces in more than half the cases, but also showed low accuracy in identifying people of any nationality — the highest hit rate was below 75%.

Proctorio does not disclose details of how its facial-recognition algorithms operate. However, a mention of OpenCV in the licenses section on the site indirectly confirms the use of this algorithm.

In response to questions from journalists, a company representative said that Satisan’s tests prove their software merely looks for faces, not identifying the people.

Earlier in April, researchers at the University of Cambridge created an online game Emojify, which demonstrates the capabilities and limitations of human emotion-recognition technologies.

In January 2021, Amnesty International called for a ban on facial recognition in public spaces. The organization argues that biometric identification of people using street-surveillance cameras is a severe intrusion into citizens’ privacy.

In early 2019, a Black American spent ten days in jail due to a facial-recognition error. The man intends to sue the local police over the baseless arrest.

Subscribe to ForkLog AI news on Telegram: ForkLog AI — all the news from the world of AI!