Selfie & Face Match
Available in SDK v1.0+
This capability verifies that the user is present, and compares the user to the ID photo provided in a previous step.
Configuration
You can configure the tolerance of facial analysis. There are three levels of precision: soft, regular, and hard. The "soft" level tolerates greater deviations from the ID photo. This allows better acceptance, fewer false rejections, but also more false acceptances. The "hard" level is the opposite: fewer false acceptances, but more false rejections. The "regular" level is a middle ground between the two.
The value configured here affects the overall probability score, which is used in the definition of rules (see next section).
Rules
There are two dimensions across which you can create outcomes for this step: how similar a face is to the one provided in a previous step, and what the liveness score is. If one of these scores falls under a value that you specify, the user will be either rejected, or sent to manual review.
The face match score is the probability that the person in the selfie is the same person on the ID document. The probability value changes depending on the calibration in the previous step. Therefore, setting the same threshold here will lead to different outcomes, depending on the original calibration used.
The liveness score is the probability that the person in the selfie is present.