Introduction to Biometrics

1
Introduction to Biometrics

• Dr. Bhavani Thuraisingham
• The University of Texas at Dallas
• Lecture 5
• Issues on Designing Biometric Systems
• September 7, 2005

2
Outline

• Biometric Terms
• Biometric Processes
• Accuracy of Biometric Systems

3
Biometric Terms

• Automated Use
• Computers / machines used to verify or determine
  identity
• Physiological/Behavioral Characteristic
• Physiological Identification based on physical
  properties such as finger scan, iris scan, - - -
• Behavioral e.g., identification based on
  gestures
• Identity
• A person may have multiple identities such as
  finger scan and face scan
• Biometric
• E.g., face scan is a biometric
• Biometric system
• Integrated hardware and software to perform
  verification and identification

4
Biometric Terms Verification and Identification

• Verification
• User claims an identity for biometric comparison
• User then provides biometric data
• System tries to match the users biometric with
  the large number of biometric data in the
  database
• Determines whether there is a match or a no match
• Network security utilizes this process
• Identification
• User does not claim an identity, but gives
  biometric data
• System searches the database to see if the
  biometric provided is stored in the database
• Positive or negative identification
• Prevents from enrolling twice for claims
• Used to enter buildings

5
Biometric Terms Logical vs. Physical Access

• Physical Access Systems
• Monitor, restrict and grant access to a
  particular area
• E.g., time reporting, access to safe, etc.
• Logical access systems
• Restrict or grant access to information systems
• E.g., popular for B2B and B2C systems

6
Biometric Process

• User enrolls in a system and provides biometric
  data
• Data is converted into a template
• Later on user provides biometric data for
  verification or identification
• The latter biometric data is converted into a
  template
• The verification/identification template is
  compared with the enrollment template
• The result of the match is specified as a
  confidence level
• The confidence level is compared to the threshold
  level
• If the confidence score exceeds the threshold,
  then there is a match
• If not, there is no match

7
Example Template
a1ij
b1ij - - - - - - - - - - -
r1ij
Tij
a2ij
b2ij - - - - - - - - - - -
r2ij
anij
bnij - - - - - - - - - - -
rnij
Tij is the jth synthetic template created by the
attacking system for user i
8
Biometric Process Examplehttp//www.foodserve.co
m/Biometrics20Defined.pdf

• Step 1 Finger is scanned and viewed by the
  MorphoTouch (Sagem Morpho Inc.) access unit at
  the point of entry.
• Step 2 In applications for children (under the
  age of 18) the image is standardized and resized
  before processing.
• Step 3 System develops a grid of intersection
  points from the swirls and arcs of the scanned
  finger.
• Step 4 The image is discarded from the record
  and is no longer available to the system or any
  operator. Only a Template remains that
  indicates the intersection points.
• Step 5 What MorphoTouch stores and recognizes
  for each individual is a set of numbers that can
  only be interpreted as a template.
• Comment The system only remembers and processes
  numbers for each individual, just like a social
  security number. The advantages with a biometric
  approach is that the number cannot be duplicated,
  lost or stolen, and, uniqueness is defined by the
  individual.

9
Enrollment and Template Creation

• Enrollment
• This is the process by which the users biometric
  data is acquired
• Templates are created
• Presentation
• User presents biometric data using hardware such
  as scanning systems, voice recorders, etc.
• Biometric data
• Unprocessed image or recording
• Feature extraction
• Locate and encode distinctive characteristics
  from biometric data

10
Data Types and Associated Biometric Technologies

• Finger scan Fingerprint Image
• Voice scan Voice recording
• Face scan Facial image
• Iris scan Iris image
• Retina scan Retina image
• Hand scan Image of hand
• Signature scan Image of signature
• Keystroke scan Recording of character types

11
Templates

• Templates are NOT compressions of biometric data
  they are constructed from distinctive features
  extracted
• Cannot reconstruct the biometric data from
  templates
• Same biometric data supplied by a user at
  different times may results in different
  templates
• When the biometric algorithm is applied to these
  templates, it will recognize them as the same
  biometric data
• Templates may consist of strings of characters
  and numeric values
• Vendor systems are heterogeneous standards are
  used for common templates and for
  interoperability

12
Biometric Matching

• Part of the Biometric process Compares the user
  provided template with the enrolled templates
• Scoring
• Each vendor may use a different score for
  matching 1-10 or -1 to 1
• Scores also generated during enrollment depending
  on the quality of the biometric data
• User may have to provide different data if
  enrollment score is low
• Threshold is generated by system administrator
  and varies from system to system and application
  to application
• Decision depending on match/ nomatch
• 100 accuracy is generally not possible

13
Metrics for Accuracy in Biometrics Systems

• False Match Rate (FMR)
• False Nonmatch Rate (FNMR)
• Failure to Enroll Rate (FTE)
• Derived Metrics

14
False Match Rate

• System gives a false positive by matching a
  users biometric with another users enrollment
• Problem as an imposter can enter the system
• Occurs when two people have high degree of
  similarity
• Facial features, shape of face etc.
• Template match gives a score that is higher than
  the threshold
• If threshold is increased then false match rate
  is reduced, but False no match rate is increased
• False match rate may be used to eliminate the
  non-matches and then do further matching

15
False Nonmatch rate

• Users template is matched with the enrolled
  templates and an incorrect decision of nonmatch
  is made
• Consequence user is denied entry
• False nonmatch occurs for the following reasons
• Changes in users biometric data
• Changes in how a user presents biometric data
• Changes in environment in which data is presented
• Major focus has been on reducing false match rate
  and as a result there are higher false nonmatch
  rates

16
Example Changes in Biometric Data

• Finger Scan
• Mostly fingerprint remains the same
• Facial Scan
• Changes in facial hair, weight
• Voice scan
• Illness can affect voice
• Iris Scan
• Highly stable
• Hand scan
• Swelling can change shape

17
Example Changes in Presentation

• Different way of presenting enrollment and
  verification/identification data
• Different way of placing fingers and different
  facial expressions
• Volume of speech, change in tone etc.
• Changes also depend on the presentation systems
  used by different vendors

18
Example Changes in Environment

• False nonmatch rates can also occur when
  environment changes even though the biometric
  data and presentation remain the same
• Background lighting, noise in the background,
  temperature changes etc.
• Background noise may affect voice scan and
  lighting may affect facial scan
• Enrollment takes place in a well lit room while
  verification takes place in a dark room

19
Failure to Enroll Rate

• Biometric data for some users may not be clear
• E.g., fingerprinting
• i.e., users may not have sufficient distinctive
  biometric data
• Enrollment needs
• Need high quality enrollment such as two finger
  scans
• Many images for facial scans
• Enrollment process varies from vendor to vendor
• Examples
• Finger scan Low quality fingerprints
• Facial scan Poor lighting
• Iris scan glasses

20
Derived Metrics

• Derived metrics are obtained by analyzing other
  metrics such as FMR
• Equal error rate ERR
• Rate at which FMR is equal to FNMR
• Generally such a system is not effective
• Ability to verify rate ATV
• ATV (1-FTE)(1-FNMR)
• Idea is that if Failure to Enroll is high than
  False nonmatch rate is also high
• More valuable metric

21
Summary

• Verification vs Identification
• Biometric process
• Enrollment and creating templates
• Matching templates
• Determining if there is a match
• Accuracy metrics
• False Match
• False Nonmatch
• Failure to enroll
• Biometric systems are not 100 accurate

22
Suggestions for Paper I, Project

• Take one Biometric (such as finger scan, face
  scan) and carry out a survey
• Introduction
• Algorithms for Face scan and matching
• Analysis
• Summary and Directions
• Biometric Standards, Secure Biometrics, Possibly
  for Paper II
• Feature Extraction Methods
• Will have a guest lecture with demonstration on
  September 12, 2005
• Lei Wang, PhD student of Prof. Latifur Khan