Biometrics Viktor MINKIN

OutlineOutline Outline Introduction Biometric systems Biometric characteristics Fingerprints Unimodal systems Multi-modal systems Problems Links History and future

Introduction Biometrics [harmonized] Automated recognition of persons based on their biological or/and behavioral characteristics. Automated measurement of biological or/and behavioral characteristics of person for medical, security or psychological purposes.

Introduction Terms and definitions Template Capture Comparison Database Enrollment Matching Token User

Introduction Identification of a person –Verification/Verify Comparing one to one “Am I who I claim I am” –Identification Comparing one to many “Who am I”

Introduction Application Passport control Access to secured areas Surveillance ATMs Computer logins E-commerce Medicine Psychology

Introduction Traditional means of automatic identification (before biometrics) –Knowledge-based Use “something that you know” Examples: password, PIN –Token-based Use “something that you have” Examples: credit card, smart card, keys

Introduction Problems with traditional approaches –Token may be lost, stolen or forgotten –PIN may be forgotten or guessed by the imposters (25% of people seem to write their PIN on their ATM card) Estimates of annual identity fraud damages per year: –$1 billion in welfare disbursements –$1 billion in credit card transactions –$1 billion in fraudulent cellular phone use –$3 billion in ATM withdrawals

Introduction The traditional approaches are unable to differentiate between an authorized person and an imposter  Use biometrics which relies on “who you are” or “what you do”

Biometric Systems Requirements for an ideal biometric –Universality Each person should have the characteristic –Uniqueness No two persons should be the same in terms of the characteristic –Permanence The characteristic should not change

Biometric Systems Issues in a real biometric system –Performance Identification accuracy, speed, robustness, resource requirements –Acceptability Extend to which people are willing to accept a particular biometric identifier –Faked protection How easy is it to fool the system by fraudulent methods

Biometric Systems Identification accuracy FAR = false acceptance rate FRR = false rejection rate EER = equal error rate TER = total error rate = FAR + FRR FER= false enrollment rate

Biometric Systems Receiver operating characteristics (ROC) False Rejection Rate False Acceptance Rate Equal Error Rate

Biometric Systems FAR/FRR and comparison threshold

Biometric Characteristics Static (biological) parameters Fingerprints Face Iris Hand geometry / vein Retinal pattern Facial thermogram Lip information DNA

Biometric Characteristics Dynamic (behavior) biometric parameters Signature Voice Motion Pulse

Biometric Characteristics Market Shares

Biometric Characteristics Market development

Fingerprints Accurate Comparatively cheap hardware Questionable acceptance

Fingerprints Optical technology Light reflects from the surface of the prism where the finger is not in contact with it, while it penetrates the surface of the prism where the finger touches the surface of the prism. The resulting image goes through a lens into a video camera. Light source Finger Video Camera (CCD) LensPrism

Fingerprints Capacity technology

Fingerprints Fiber optic technology

Fingerprints Fingerprint types Arches Loops Whorl Bridge Dot Ridge Ending Bifurcation Enclosure Minutia types

Fingerprints Core & Deltas

Fingerprints Fingerprint minutiae

Fingerprints Image transformation Source FFT Flow field Directional Directional Directional image 1 image 2 irregularity Code Smoothing Binarization Skeleton Skeleton Minutiae formation cleaning search

Fingerprints Comparative testing

Fingerprints Fingerprint information

Unimodal Systems Facial ID Illumination Head pose Occlusion

Unimodal Systems Hand Vein Questionable accuracy Hand geometry

Unimodal Systems Retinal Pattern Highest accuracy Even more intrusive than iris recognition

Unimodal Systems Facial Thermo image and VibraImage Non-intrusive Lie detection View-dependent Emotion control Depends heavily on Criminals detector human factors, Medical monitoring body temperature Psychology testing

Multi-modal Systems Why multimodal [multiple] person identification? –Quest for non-intrusive identification methods No special purpose hardware needed Works potentially at greater distances –“Traditional” arguments for going multimodal: Increasing performance Increasing robustness –Mono-modal recognition techniques are likely to reach in a close future a saturation in performance.

Multi-modal Systems: Fusion “Early integration” or “sensor fusion” Integration is performed on the feature level Classification is done on the combined feature vector Features Modality 1 Classifier Features Modality 2 Features Modality n-1 Features Modality n Identity

Multi-modal Systems BioFinger 3 - Elsys includes BiCard, VibraImage, 3D-Elsys is biological and behavioral identification system

Multi-modal Systems The World population in 2000 was about M. people. The biometric document (ID card) market is more than $ There are 3 different ID card technologies: 1. Card with additional memory (chip, CD,..) 2. Card with 2d-bar code 3. BiCard (3D-Elsys)

Problems Errors rate Misunderstanding of real advantages and problems Incomplete true about biometric systems

Links International Biometric Group - NIST - Literature – Patents -

Biometrics evolution 19 century- not automated identification 20 century- biometric identification 21 century- emotion recognition and detection

Viktor Minkin Biometrics Thank you! 2004