Face match1:1

Checking that a live selfie and the ID photo are the same person.

The system compares the face in the selfie with the face on the document and decides whether they belong to one person.

One-to-one biometric comparison producing a similarity score against a threshold. A threshold is the cutoff the service sets: scores above it count as a match, and scores below it do not.

A face match typically works like this:

1. The system finds the face in the selfie and in the ID photo.
2. It converts each face into a template, a mathematical summary of its features.
3. It compares the two templates and produces a similarity score.
4. It compares the score with the threshold and returns match or no match.

The threshold trades two errors against each other:

• False acceptance happens when the system wrongly says two different people match, which lets a fraudster through.
• False rejection happens when the system wrongly says the same person does not match, which turns away a genuine customer.

The National Institute of Standards and Technology (NIST) publishes independent accuracy tests of face recognition algorithms, which gives buyers an outside benchmark.

Face match sits next to terms that are easy to mix up:

• Face search (1:N) compares one face against many, instead of one against one.
• Liveness detection confirms the selfie came from a live person.
• Biometric template is the mathematical summary the comparison uses.