Scribd
Upload
62 views31 pages

Introduction To Biometrics Introduction To Biometrics: Executive Education

This document provides an introduction to biometrics, including an overview of biometric usage scenarios, historical uses, modality categorizations, the generic biometric process, potential matching errors, and spoofing attacks and defenses. It discusses key concepts such as enrollment, capture, processing, extraction, template storage, comparison, and consequences of matches and non-matches. Thresholds are important for deciding matches, and there is a relationship between threshold and false accept and false reject rates. Statistical sampling is also discussed.

Uploaded by

kiran143a
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
62 views31 pages

Introduction To Biometrics Introduction To Biometrics: Executive Education

This document provides an introduction to biometrics, including an overview of biometric usage scenarios, historical uses, modality categorizations, the generic biometric process, potential matching errors, and spoofing attacks and defenses. It discusses key concepts such as enrollment, capture, processing, extraction, template storage, comparison, and consequences of matches and non-matches. Thresholds are important for deciding matches, and there is a relationship between threshold and false accept and false reject rates. Statistical sampling is also discussed.

Uploaded by

kiran143a
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 31

Introduction to Biometrics

Executive Education

Learning Objectives
Biometric Usage Scenarios Properties & Measures of Modalities The Biometric Process Comparison Difficulties Potential M t hi P t ti l Matching Errors E Spoofing and its realities and defenses

Usage Cases / Usage Scenarios


Verification Proving who you claim to be A one-to-one mapping Identification Sea c g o Searching for the pe so a b o e c e person biometric belongs to A one-to-many search y

Historical Uses
Verification Deter and detect fraud Access control Identification Id tifi ti Identify repeat offenders Identify perpetrator of a crime
4

Modality Categorizations
Physical vs. Behavioral Interactive vs. Passive Overt vs. Covert O t C t Close up vs. Distant
5

Seven Measures of a Modality


Universality Distinctiveness Permanence Collectability Performance ( c. speed & accuracy) e o a ce (inc. accu acy) Acceptability Resistance to Circumvention

An Experts Opinion
Distinctiv veness Unspoofa ability Performa ance Permane ence Collectab bility Acceptab bility Universa ality

Fingerprint Iris Retina Hand Geometry Palmprint Hand Vein Voice Face Face Therm. DNA

M H H M M M M H H H

H H H M H M L L H H

H H M M H M L M L H

M M L H M M M H H L

H H H M H M L L M H

M L L M M M H H H L

M H H M M H L L H L

Ada apted fro An Introduc om I ction to Biometric Recognit R tion by Jain.


7

Generic Biometric Process*

*Adapted and modified from a figure in Handbook of Biometrics, pg. 97

Enrollment
Enrollment is a critical process in a biometric system. bi ti t The basis of biometrics: a biometric sample linked t li k d to a known person k Special case

Capture
Obtaining the biometric sample: Scanning a finger Taking a picture of an iris Taking a picture of a fingerprint left at the scene of a crime Recording someone talking Etc. Also referred to as Acquire q
10

Process
Some reasons for pre-processing: Sub-optimal lighting conditions Eyelashes covering iris Dirty or dry fingers Person isnt looking straight ahead P i t l ki t i ht h d Contact lenses causing glare Many faces in one photo Background noise in a recording
11

Extract
Obtain properties of the sample that are suitable for computer storage and comparison. The result is referred to as a template. The form and size of a template is dependent on the modality and the approach being used.

12

Fingerprint Example (1)


Enrollment Print Minutiae Found

= Bifurcation = Termination
13

Store in a Database
Minimally, the template is stored. Optionally, the capture is also stored. What is a possible benefit of also storing the enrollment capture? How might template storage differ depending how it will be used? d di h ill b d?

14

Compare templates and decide

15

Compare
A biometric system can be thought of as a complex pattern recognition machine. Automating this step accurately is very difficult never an exact match. The method of comparison is biometric dependent. biometric-dependent.

16

Fingerprint Example (2)


Enrolled

Claimant Cl i t

17

Iris Example

18

Consequence of a Match
Comparison output: Yes / No, On / Off, Match / No-match Results: R lt Green light / Red light Unlock door / Keep door locked Login / Keep out
19

Potential Matching Errors


False Reject Also known as False Non-Match Also known as Type I Error Important measure: FRR False Accept a se ccep Also known as False Match Also known as Type II Error Type Error Important measure: FAR
20

The four possible outcomes


Genuine Accept False Accept Genuine Reject G i R j t False Reject
21

Which is the worst outcome? Why?

How Do We Decide?
How can computers do the equivalent comparison for complex patterns? The rough answer: statistics. Very different from passwords

22

Contrived Example
Assume the following are two templates (enrolled and claimant).
Enrolled biometric Claimant Comparison Result 11100101100110110010110101011010 11000111100110010001100101011011 00100010000000100011010000000001

Is this close enough to be considered a match?


23

An Aside on Scoring
In the previous example, if two templates of the same person are compared we can say that the genuine score is 7. If two templates of different p p are p people compared we can say the the p imposter score is 7. To restate: is 7 a score that should be considered a match?
24

How do we decide?
How do we decide whether the previous templates are a match? There is typically a value referred to as / a threshold above/below which it is considered a match. What is a good value for the example?

25

The Relationships
Threshold and FAR and FRR are dependent values. If an algorithm threshold is changed so that the FAR goes down, then the FRR will most likely go up. There is no perfect biometric that has a 0% FAR and 0% FRR. Consequence on a large database?
26

Statistical sampling
Database of 1st capture
Alice Bob B b Carol Dave Eve Frank Gina Harry Ingrid Jack

Database of 2nd capture


Alice Bob B b Carol Dave Eve Frank Gina Harry Ingrid Jack

Database of 3rd capture


Alice Bob B b Carol Dave Eve Frank Gina Harry Ingrid Jack
27

FAR vs. FRR


Another way to look at it:
Threshold Com mpariso ons Imposter Genuine

Score

28

Spoofing?
Example spoofing attacks: Fake finger(tip) Photo or video of a face Recording of a voice Example defenses E l d f Visual oversight (e.g., a Guard) Automated liveness detection Use of multiple modalities
29

Some Parting Lessons


Biometric systems are not perfect, so plan accordingly Despite lack of perfection, they are very useful in a lot of situations There i no best modality Th is b t d lit

30

The End

31

You might also like