Hung Liao, Ph.D.
Identification vs Verification
Identification: Who am I? One
Verification: Am I who I claim I am? One
Detection: Find out whether there is an
instance of a given type of object in an
Recognition: detection + identification
False Acceptance Rate (FAR) : the probability
that a biometric device will allow a ‘bad guy’ to
pass. Related to security.
False Rejection Rate (FRR):the probability that a
biometric device won't recognize a good guy.
Related to convenience.
The point where false accept and false reject
curves cross is called the "Equal Error Rate."
The Equal Error Rate provides a good indicator
of the unit's performance. The smaller the Equal
Error Rate, the better.
Validity of Test Data
Testing biometrics is difficult, because of the
extremely low error rates involved.
Some are based on theoretical models.
Some are obtained from actual field testing.
It's important to remember that error rates
are statistical: they are derived from a series
of transactions by a population of users.
What is a good biometric feature?
Easy (or not too difficult) to acquire
Low processing cost
scan biometrics is based on the
distinctive characteristics of the human
A fingerprint image is read from a capture
device, features are extracted from the
image, and a template is created.
If appropriate precautions are followed, what
results is a very accurate means of
Fingerprints vs Finger
Fingerprint images require 250kb per
finger for a high
Can be acquired using ink
procedure, optical or non
scan technology doesn't store the
full fingerprint image. It stores particular
data about the fingerprint in a much
smaller template, requiring from 250
AFIS (Automated Fingerprint Identification
commonly referred to as "AFIS
Systems" (a redundancy)
is a term applied
scan technology can be
used in AFIS on 100,000 person databases,
it is much more frequently used for one
one verification within 1
Can be classified
according to the
Feature Extraction Steps
interrupt the otherwise
smooth flow of ridges, are
the basis for most finger
False Rejection Rates (FRR), or the likelihood
that the system will not "recognize" an
enrolled user's finger
scan, in the vicinity of
False Acceptance Rates (FAR), or the
likelihood that the system will mistakenly
"recognize" the finger
scan of a user who is
not in the system, are frequently stated in the
vicinity of 0.001%.
The point at which the FAR and FRR meet is
the Equal Error Rate, frequently claimed to be
Iris recognition is based on visible (via regular
and/or infrared light) qualities of the iris.
A primary visible characteristic is the trabecular
meshwork (permanently formed by the 8th
month of gestation), a tissue which gives the
appearance of dividing the iris in a radial
Other visible characteristics include rings,
furrows, freckles, and the corona.
Iris Recognition Technology
Iris recognition technology converts the
visible characteristics discussed before into
a 512 byte IrisCode(tm), a template stored
for future verification attempts.
The odds of two different irises returning a
75% match (i.e. having a Hamming
Distance of 0.25): 1 in 10^16
Equal Error Rate (the point at which the
likelihood of a false accept and false reject
are the same): 1 in 1.2 million
The odds of 2 different irises returning
identical IrisCodes: 1 in 10^52
Established prior to birth and
remains intact through out the life.
For more details
Check Dr. John Daugman’s web page:
scan reads the top and sides of the
hands and fingers, using such metrics as the
height of the fingers, distance between joints,
and shape of the knuckles.
Although not the most accurate physiological
biometric, hand scan has proven to be an
ideal solution for low
applications where deterrence and
convenience are as much a consideration as
security and accuracy.
Pros and Cons
Ease of use
Injury to hands
Most natural because this is how
we human recognize other people.
Remains a difficult subject.
Primary Facial Scan Technologies
automatic face processing
The most widely utilized facial
Local Feature Analysis (LFA) utilizes
dozens of features from different
regions of the face, and also
incorporates the relative location of
The extracted (very small) features
are building blocks, and both the type
of blocks and their arrangement are
used to identify/verify.
Features from both faces
enrollment and verification face
vote on whether there is a match.
Neural networks employ an
algorithm to determine the
similarity of the unique global
features of live versus enrolled or
reference faces, using as much of
the facial image as possible.
Automatic Face Processing (AFP)
is a more rudimentary technology,
using distances and distance ratios
between easily acquired features
such as eyes, end of nose, and
corners of mouth.
Not as robust, but AFP may be
more effective in dimly lit, frontal
image capture situations.