The most common performance metrics to rate the performance of a biometric factor, solution or application are theÂ False Acceptance RateÂ FAR and theÂ False Rejection RateÂ FRR.
When using a biometric application for the first time the user needs toÂ enrollÂ to the system. The system requests fingerprints, a voice recording or another biometric factor from the operator, this input is registered in the database as aÂ templateÂ which is linked internally to a user ID. The next time when the user wants to authenticate or identify himself, the biometric input is compared to the template(s) in the database by a matching algorithm which responds with acceptance (match) or rejection (no match).
TheÂ FARÂ or False Acceptance rate is the probability that the system incorrectly authorizes a non-authorized person, due to incorrectly matching the biometric input with a template. The FAR is normally expressed as a percentage, following the FAR definition this is theÂ percentage of invalid inputs which are incorrectly accepted..
TheÂ FRRÂ or False Rejection Rate is the probability that the system incorrectly rejects access to an authorized person, due to failing to match the biometric input with a template. The FRR is normally expressed as a percentage, following the FRR definition this is theÂ percentage of valid inputs which are incorrectly rejected.
You want both the FAR and the FRR to be low as possible. FAR and FRR are very much dependent on the biometric factor that is used and on the technical implementation of the biometric solution. Furthermore the FRR is strongly person dependent, a personal FRR can be determined for each individual.Â Also FRR might increase due to environmental conditions or incorrect use, for example when using dirty fingers on a fingerprint reader. Mostly the FRR lowers when a user gains more experience in how to use the biometric device or software.
FAR and FRR are key metrics for biometric solutions, some biometric devices or software even allow to tune them so that the system more quickly matches or rejects. Both FRR and FAR are important, but for most applications one of them is considered most important.
For example when biometrics are used for logical or physical access control, the objective of the application is to disallow access to unauthorized individuals under all circumstances. It is clear that a very low FAR is needed for such an application, even if it comes at the price of a higher FRR.
But when surveillance cameras are used to screen a crowd of people for missing children, the objective of the application is to identify any missing children that come up on the screen. When the identification of those children is automated using a face recognition software, this software has to be set up with a low FRR. As such a higher number of matches will be false positives, but these can be reviewed quickly by surveillance personnel.
When it comes to the quality of performance of biometrics the quality of the enrollment is critical.