Global face biometrics pass rate 95 percent; Sumsub sees room for improvement

The average global pass rate for remote identity checks with face biometrics is 95 percent, according to new research from Sumsub, well ahead of the 83 percent pass rate for ‘selfie with ID’ checks. Sumsub analyzed data it collected globally in 2021 to offer a snapshot into how well mobile onboarding and facial authentication processes are working.

The findings yield some insights that Sumsub Founder Jacob Sever tells Biometric Update in an exclusive email interview can help businesses make material improvements to their onboarding and authentication practices.

“Not all the companies are giving enough information to their end-users,” writes Sever. “Businesses may provide some education and tell their customers why they need to pass verification and how to do it successfully. For instance, sharing an article or a video with tips on passing facial biometric checks would help.”

The company found significant differences in process durations for selfie biometrics between end-user countries, with biometric liveness checks of users in the UK, Japan, Canada, Bahrain, Hong Kong, Singapore and the EU usually taking less than 35 seconds, compared to 45 seconds in the U.S., and an average of 1.5 minutes in Nigeria, Venezuela, Benin, Angola and Israel. Sumsub speculates that the difference may be due to familiarity with the process and associated movements in different countries.

Sever says in the email that Sumsub’s research shows that its face biometrics show even accuracy between different demographic groups, and that internet connection speed is another likely reason for regional disparities in how long the processes take.

Different operating systems also significantly effect the amount of time needed to complete face biometrics checks. The process usually takes over 50 seconds for Linux, Windows and Android users, compared to less than 39 seconds for MacOS and iOS users. This may also be due to the familiarity with the face biometric modality encouraged by Apple.

“To reduce processing times, we recommend adding as many tips as possible to the face authentication procedure, guiding a person through the process,” Sever explains. “For example, Sumsub provides built-in tips on the Web SDK. In addition, businesses could allow applicants to pass facial biometric check in their own language.”

Sumsub’s advice on informing end-users comes largely from the observation that 94 percent of failed face biometrics checks are caused by improper conditions. Poor video quality or failure to perform the required movements made up half of those failures. Most of the remainder were made up of the subject’s face being too far away, and therefore small, or too close, and therefore too large in the frame, but in an additional 5 percent, the subject’s face was not completely visible.

Almost 6 percent of errors were found to be due to possible digital manipulation. Of those, bypass or presentation attacks such as screenshots, photo manipulation, masks or deepfakes represented 70 percent. The rest were roughly split between the use of different devices or people, and forced verification and other fraud attempts.

Sumsub launched a face biometrics and liveness detection service for small and medium-sized businesses in December, and was recently selected by social crypto exchange BingX to help it with regulatory compliance obligations.

Article Topics

accuracy  |  biometric liveness detection  |  biometrics  |  face biometrics  |  facial authentication  |  identity verification  |  onboarding  |  Sumsub  |  user experience