Yoti improves liveness detection pass rates

Digital identity and age estimation company Yoti has released new figures on its liveness detection technology, showing success rate improvements in passive and active liveness over the past two years.

In May 2025, its MyFace passive liveness algorithms reached a 92 percent success rate at one attempt for mobile environments, 90 percent for desktop and 92 percent for blended environments. In comparison, March 2023 figures show that the success rates were 89, 78 and 89 percent, respectively.

The pass rates after three attempts also improved, reaching 97 percent for mobile, 94 percent for desktop and 97 percent for blended environments, up from 95, 87 and 93 percent, respectively.

Liveness detection is aimed at verifying whether the person in front of the camera is real instead of a photo or a video, also known as a Presentation Attack Detection (PAD).

Yoti has also published data on improvements in active liveness detection. Unlike passive liveness, which doesn’t require any action from the user and therefore reduces time and friction, active liveness asks the person going through a check to perform a series of movements with their head or hand in front of the camera.

At one attempt, the algorithm scored 76, 60 and 75 percent in mobile, desktop and blended environments, respectively. In 2023, those figures were 74, 56 and 72 percent.

At three attempts, MyFace active liveness detection reached an 87, 68, 85 percent success rate in mobile, desktop and blended environments, respectively, up from 83, 67 and 81 percent.

The differences between success rates in different environments are influenced by camera quality, image size and environmental factors. Mobile phones tend to have higher-quality cameras compared to laptops or self-checkout machines at grocery stores, the UK-based firm explains in the MyFace Liveness White Paper.

Yoti has been using its MyFace liveness detection feature in its facial age estimation product, which was tested by supermarket chains Asda, Co-op, Morrisons and Tesco in 2023. The technology was used to prevent underage customers from buying restricted items such as alcohol.

The trial has helped gain insight into the influence of factors such as sun glare, overhead lighting and camera positioning on the effectiveness of age estimation at self-checkout machines, says the company.

In 2023, Yoti MyFace achieved compliance with ISO PAD Level 2 compliance during testing by iBeta, with a 100 percent attack detection rate. The Level 2 test evaluates performance under more sophisticated attacks with latex face masks, 3D-printed artifacts or deepfakes.

Threat of attacks is rising

Yoti’s data also shows a significant rise in the number of attacks on age and identity verification checks, such as deepfake or injection attacks, in which fraudsters feed a fake video into the verification process.

The company saw around 1,000 attacks in February 2024, while in January this year, that number reached 6,000. The overall number of attacks has risen from 1.6 percent to 3.9 percent.

The company attributes this to the expansion of its services: Yoti now performs over 7 million checks per week. Since 2020, however, it has been developing a solution to prevent injection attacks called SICAP (Secure Image Capture).

“Our experience with providing verification solutions demands a capability of continual improvement due to a developing fraud environment,” says the company.

These conclusions are in line with those from the 2025 Face Liveness Market Analysis & Buyer’s Guide, published earlier this year by Biometric Update and Goode Intelligence. The research provides face liveness market forecasts and a list of pioneering liveness vendors, including Yoti.

Article Topics

accuracy  |  age estimation  |  biometric liveness detection  |  biometrics  |  face biometrics  |  passive facial liveness  |  Yoti  |  Yoti MyFace