False acceptance rateFAR

How often the system wrongly says two different people match.

If a face check has a false acceptance rate of one in 10,000, then about one out of every 10,000 attempts by impostors would be wrongly accepted.

The proportion of impostor comparisons incorrectly accepted at a given threshold. An impostor comparison pairs a sample with a different person's record, and the threshold is the similarity score above which the system declares a match.

The rate depends on the threshold:

• A stricter threshold requires a higher similarity score, so fewer impostors get through and the false acceptance rate falls.
• A looser threshold accepts lower scores, so more impostors get through and the rate rises.

Test labs measure it by running a very large number of impostor comparisons and counting how many the system accepts.

NIST and many standards report the algorithm-level version as the false match rate (FMR), alongside the false non-match rate (FNMR). In everyday use, false acceptance rate (FAR) and false rejection rate (FRR) are the familiar names for the same trade-off.

False acceptance rate sits next to terms that are easy to mix up:

• False rejection rate (FRR) is the opposite error, where the right person is wrongly turned away.
• Equal error rate (EER) is the single point where FAR and FRR are equal.
• Face match is the comparison whose threshold sets the false acceptance rate.