iPB4 Biometric Device Range
–
Installation & User Guide
–
Rev 07 ©2017-2018 iPulse Systems
35
MEASURING PERFORMANC E ON BIOMETRIC SYSTE MS
Measuring the actual performance of a biometric system is a complex process affected by various factors,
including
–
but not limited to
–
environmental conditions, human interference and incorrect, sometimes
conflicting system settings.
Fingerprint-based biometric devices allow manual changes to the threshold at which a print is deemed an
accurate enough comparison to be considered a successful match. With a low threshold
–
for example,
only 10% of the minutiae po
ints need to be matched for a “successful”
identification
–
very few people will
experience trouble being identified by the system. The drawback to this, however, is an increased risk in
p
ossible “false” identifications.
Raising this threshold holds true for the opposite: with a higher number of minutiae points required for a
successful match, chances are increased that people will not be identified by the device. At the same time,
the risk of someone being identified as someone else drops quite significantly.
The ideal is a balance between these two extremes, a setting where the threshold for successful matches
is high enough to avoid false identifications, yet low enough to match people consistently and without
difficulty. These settings can be changed dynamically according to the requirements of the situation.
CORE METRICS
In general, three core metrics are used to determine the performance of a biometric system.
False Rejection Rate (FRR)
Also referred to as
False Non-Match Rate (FNMR)
, this indicates the probability that the system will fail to
detect a match between the input pattern and a matching stored template. It measures the percentage of
valid inputs that are incorrectly rejected.
False Acceptance Rate (FAR)
Also referred to as
False Matching Rate (FMR),
this indicates the probability that the system will incorrectly
match the input pattern to a stored non-matching template. It measures the percentage of invalid inputs
which are incorrectly accepted.
Equal Error Rate (EER)
Also referred to as
Crossover Error Rate (CER),
this indicates the rate at which both false acceptances (FAR)
and false rejections (FRR) errors are equal; i.e. for each false acceptance there will be one false rejection.
The EER is a quick way to compare different biometric systems
–
the system with the lowest EER score is
the most accurate.
CERTIFICATION AUTHOR ITIES
The first two metrics (
FRR
and
FAR
) are usually published by the
International Biometrics Group (IBG),
a
for-profit biometric industry organisation that performs these tests. There are, however, several
deficiencies and shortcomings in the IBG tests, the most crucial of which is that all tests are performed in
laboratory conditions with a relatively small sample database.
Summary of Contents for iPB4
Page 1: ......
