Summary
This white paper presents Yoti's facial age estimation technology, which provides a secure method for estimating an individual's age by analyzing facial features. The system operates without requiring physical document checks or human intervention, adhering to the 'privacy by design' principle as outlined in the UK GDPR. It ensures that no individual can be identified through the model, promoting data minimization by only utilizing facial images, which are deleted immediately after processing. The technology employs a neural network trained to estimate ages, demonstrating a mean absolute error of 1.36 years for ages 6 to 12 and 1.52 years for ages 13 to 19. The True Positive Rate for correctly identifying individuals under 23 years old is reported at 99.65% for the 13-17 age group and 98.91% for the 6-11 age group. The paper also discusses the ethical considerations and compliance with regulations, including the approval received from the German regulator for high-level age assurance.
