Age Verification

The Identity Verification Toolkit allows online age verification of a person on a web or in mobile phone app. This uses age estimation from selfie algorithm and OCR of ID documents.

Certain businesses require age checks to be performed to establish if the user is permitted to enter a business or a website; or to purchase certain goods. Age check, passive liveness, and face matching can be combined to ensure compliance.

Steps in the process#

Selfie photo age check#

Identity Verification Toolkit can estimate age from the selfie image. The estimation algorithm is ranking as Nr.1 with the lowest mean absolute error in the NIST FATE AEV benchmark. It returns the estimated age in years. Passive liveness has to be evaluated on the provided photo as well to prevent presentation attacks.

Age Verification from Identity Document#

If the estimated age is close to the age limit permitted by law, scanning the identity document and extracting the date of birth can calculate the age precisely.

It is recommended to keep a certain tolerance for the estimation. For example, if the age limit is 18, a photo ID should be asked from everybody whose age was estimated under 26 if we want to have a false accept rate of one person per 500. See the tables below.

Threshold for Age Estimation to Perform Age Check#

The following characteristics have been measured with our testing dataset. Age is returned as an integer number representing years of age.

Age Estimation ACCURATE Algorithm (since DIS v1.50)#

Age Estimation FAST Algorithm (since DIS v1.0)#