41
Bridging The Gap: From Biometrics to Forensics Anil K. Jain Michigan State University http://biometrics.cse.msu.edu February 2, 2015 1

Bridging The Gap: From Biometrics to Forensics

  • Upload
    others

  • View
    4

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Bridging The Gap: From Biometrics to Forensics

Bridging The Gap: From Biometrics to Forensics

Anil K. Jain Michigan State University

http://biometrics.cse.msu.edu

February 2, 2015

1

Page 2: Bridging The Gap: From Biometrics to Forensics

Outline

2

• Questions about identity

• What is biometrics?

• Why biometric recognition?

• Automatic biometric recognition

• Challenges

• Bridging the gap

Page 3: Bridging The Gap: From Biometrics to Forensics

Bridging The Gap • Uncooperative subjects

• Incomplete and poor quality information

• Fusion of information

• Capacity (distinctive) of body traits

• Persistence of body traits

• Probabilistic models (confidence, reject option) – Data to learn the models?

Page 4: Bridging The Gap: From Biometrics to Forensics

• Is Steve authorized to access the website?

• Is Cathy the owner of the bank account?

• Is the visa associated with John?

• Is Mary authorized to enter the building?

We rely on credentials: documents & secrets

Questions About Identity

Page 5: Bridging The Gap: From Biometrics to Forensics

• Stolen passports are routinely used for illegal activity. Two Iranians boarded flight MH370 using passports stolen in Thailand

• 160,050 U.K. passports were either lost or stolen abroad (2008- 2013) • INTERPOL has data on 40M lost/stolen travel documents

What Happens to Lost and Stolen Passports? (BBC News Magazine, 28 March 2014, http://www.bbc.co.uk/news/magazine-26783486)

http://press.homeoffce.gov.uk/press-releases/passport-warning?version=1; http://www.pbs.org/wgbh/pages/frontline/shows/trail/etc/fake.html

Page 6: Bridging The Gap: From Biometrics to Forensics

Driver Licenses Beta test of face-based scrubbing of 13.5M records in Michigan DMV database shows photos of different subjects in some records!

Courtesy: Pete Langenfeld, MSP

Page 7: Bridging The Gap: From Biometrics to Forensics

Too Many Passwords!

Most common passwords: 123456, password, qwerty

Page 8: Bridging The Gap: From Biometrics to Forensics

Bank ATM

Input to ATM:

Card

4-digit PIN

8

• Machine does not know who is inputting the card and PIN • Estimated no. of unauthorized transactions in 2012: 31M with $6B value

https://www.frbservices.org/files/communications/pdf/research/2013_payments_study_summary.pdf

Page 9: Bridging The Gap: From Biometrics to Forensics

Biometric Recognition

Automated recognition of individuals based on their behavioral and biological characteristics [ISO/IEC JTC1 2382-37:2012]

– Recognition of individuals

– Use of automated methods (no human in loop)

– Context (application)

Page 10: Bridging The Gap: From Biometrics to Forensics

Bank ATM With Biometrics

Biometric ATM in Lanzhou

Finger Vein recognition

module

10

Three-factor authentication: Card, PIN and Finger Vein

Page 11: Bridging The Gap: From Biometrics to Forensics

Origin of the Term Biometrics • First use of Biometry attributed to William Whewell, FRS

(1831); used for “calculations on lives” or demography; analysis of biological data using mathematical & statistical methods

• Morris (1875): “… a word whose precise derivation illustrates its intended meaning- from the Greek words Bios, life; Metron, a measure”

• Pollack (1981): “What makes each person unique?” Introduced biometrics in the context of access control

• Stigler, “The Problematic Unity of Biometries”, BIOMETRIC5, S6 , Sept. 2000 • Pollack, “Technology: Recognizing the Real You”, New York Times, Sept 24, 1981

Page 12: Bridging The Gap: From Biometrics to Forensics

Bertillon System (1882)

H.T. F. Rhodes, Alphonse Bertillon: Father of Scientific Detection, Harrap, 1956

Page 13: Bridging The Gap: From Biometrics to Forensics

Fingerprints

“Perhaps the most beautiful and

characteristic of all superficial marks (on

human body) are the small furrows with the

intervening ridges and their pores that are

disposed in a singularly complex yet even

order on the under surfaces of the hands and

feet.” Francis Galton, Nature, June 28, 1888

Page 14: Bridging The Gap: From Biometrics to Forensics

Biometric Traits

14 Not all of these used in forensics

Page 15: Bridging The Gap: From Biometrics to Forensics

Biometric Recognition System

Feature Extractor

Template Database

Recognition

Enrollment

Matcher

(similarity score)

Yes: score > T

No: score < T

Preprocessor Preprocessor

T: threshold or system operating point

15

Verification: 1:1 matching; Identification: 1:N matching

Page 16: Bridging The Gap: From Biometrics to Forensics

Freq

ue

ncy

Similarity score

T = 0.18

Score: 0.10 False Reject

Score: 0.21 False Accept

Biometric Systems Make Errors

Page 17: Bridging The Gap: From Biometrics to Forensics

Designing Biometric Systems

Texas Health hospital system has 2,488 patients named Maria Garcia; 231 have the same date of birth

Coal miner identification for safety and T&A; “Anti-Black-Face” Technology

• Purpose, target population, system requirements

Page 18: Bridging The Gap: From Biometrics to Forensics

Drivers of Biometric Technology • Security: Does the person have a prior criminal record?

• Convenience: No need to remember password

• Audit trail: Who accessed the bank vault?

• Fraud detection: Is the card holder the rightful owner?

• De-duplication: One person, one passport!

Enablers: Processor, storage, sensor technology

Page 19: Bridging The Gap: From Biometrics to Forensics

Michigan AFIS (1989)

Enrollment of 724K subjects; average of 400 tenprint search/day @15K comparisons/sec.

Processor & Storage Technology

Michigan AFIS (2014)

Courtesy, Scott Blanchard, MSP

Enrollment of 3.M; 2K tenprint search/day, 35 latent search/day, 5 latent palm search/day @25M comparisons/sec.

Page 20: Bridging The Gap: From Biometrics to Forensics

Biometric Sensing Technology

Identix 12.5cm x 18cm x 6cm; 1 lb.

Cost: ~$1,500 (1995)

Digital Persona (Optical)

Morpho “Finger on the Fly” (Contactless)

Lumidigm (Multispectral)

Authentec, Swipe sensor (Capacitive)

Goodix (Capacitive)

96x96 pixels (500PPI)

Page 21: Bridging The Gap: From Biometrics to Forensics

Game Changer

OBIM (formerly US VISIT) (2004)

Page 22: Bridging The Gap: From Biometrics to Forensics

Game Changer

Walt Disney Theme Park (2005)

Page 23: Bridging The Gap: From Biometrics to Forensics

FBI Next Generation Identification (2008)

http://www.fbi.gov/about-us/cjis/fingerprints_biometrics/ngi/ngi2/

Game Changer

First AFIS in 1980; IAFIS launched in 1999

Page 24: Bridging The Gap: From Biometrics to Forensics

Name

Parents

Gender

DoB

PoB

Address

1568 3647 4958 Basic demographic data and biometrics

stored centrally

UID = 1568 3647 4958

10 fingerprints, 2 irises & face image

Central UID database

India’s Aadhaar Program (2009)

https://portal.uidai.gov.in/uidwebportal/dashboard.do

Game Changer

Page 25: Bridging The Gap: From Biometrics to Forensics

World’s Largest Biometric Database

25

~730 million unique 12-digit ID numbers have been issued as of January 2015

Page 26: Bridging The Gap: From Biometrics to Forensics

Game Changer

Laser-Cut Sapphire Crystal

Stainless Steel Detection Ring

Capacitive Single-Touch Sensor

Tactile Switch

Apple Touch ID (2013)

Page 27: Bridging The Gap: From Biometrics to Forensics

Apple Pay (2014) with NFC device; 2-factor authentication

Game Changer

Page 28: Bridging The Gap: From Biometrics to Forensics

State of the Art Performance

Users

Imag

ing

Co

nd

itio

ns

Cooperative Co

nst

rain

ed

100% TAR @ FAR=0.1%

99.4% TAR @ FAR=0.01%

97.8% TAR @ FAR=0.01%

FVC2002 FRGC, Exp. 1 IREX III

Un

con

stra

ined

IIITD alcoholic iris Altered fingerprint FERET

76% TAR @ FAR=0.1%

75% Rank-1 accuracy

97% TAR @ FAR=0.1%

Non-cooperative

FVC2004 MBGC CASIA.v4-distance

92.2% TAR @ FAR=0.01%

98% TAR @ FAR=0.1%

95% Rank-1 accuracy

80% TAR @ FAR=0.1%

72% Rank-1 accuracy

NIST SD27 LFW UBIRIS.v2

66.8% TAR @ FAR=10%

28

Page 29: Bridging The Gap: From Biometrics to Forensics

Bridging The Gap

Forensic Science Biometrics

• Latent fingerprint • Latent palmprint • Fibers • Explosive residue • Paint chips • DNA • Tire marks • Shoe prints • Bite marks • Handwriting • SMT

• 2D Face • 3D Face • Fingerprint • Iris • Speech • Signature • Gait • Ear • Palmprint • Keystroke

• Distinctiveness • Minimize human bias and sources of error • Probabilistic models of similarity score; degree of uncertainty • Fusion of evidence

Page 30: Bridging The Gap: From Biometrics to Forensics

Image Alignment

Feature Extraction

Classification Score

BF FP

Probabilistic Multi-mark Fusion

LEA 1 LEA 2

LEA 3

Cartridge cases bullet-striated toolmarks

EM

Probabilistic Matching of Ballistics Images Toolmark types: breech face impression (BF); firing pin impression (FP); ejector mark (EM)

Page 31: Bridging The Gap: From Biometrics to Forensics

Distinctiveness of a Biometric Trait

• Given a 10-digit PIN, no. of unique identities that can be resolved = 10 billion

• But, what can we say about a biometric trait?

• How many traits needed to recognize 7B individuals?

• Biometric trait vs. representation

31

Page 32: Bridging The Gap: From Biometrics to Forensics

Distinctiveness of Face Representation (Intra-person Variability)

Same woman arrested at different times

32

Page 33: Bridging The Gap: From Biometrics to Forensics

Does the fingerprint matching accuracy decrease with time lapse?

June 2001 January 2003 January 2005

August 2007 April 2010 July 2013

Persistence of Biometric Trait

S. Yoon and A. K. Jain, "Longitudinal Study of Fingerprint Recognition", MSU Technical Report, MSU-CSE-14-3, June 9, 2014.

Page 34: Bridging The Gap: From Biometrics to Forensics

Biometrics For Lifetime

Inked finger of an Afghan woman after voting in Bamiyan, Afghanistan

Fingerprint capture of a baby at a health clinic in Cotonou, Benin

Page 35: Bridging The Gap: From Biometrics to Forensics

FaceSketchID System

Retrieval rank: 5,432

Retrieval rank: 112 S. Klum, H. Han, B. Klare and A. K. Jain, "The FaceSketchID System: Matching Facial Composites to Mugshots", IEEE Trans. Information Forensics and Security, December 2014

Page 36: Bridging The Gap: From Biometrics to Forensics

A Case Study on the Boston Bomber (Gallery of one million mug shots)

*

Tamerlan Tsarnaev

Page 37: Bridging The Gap: From Biometrics to Forensics

Soft Biometrics: Scars, Mark & Tattoos

http://wtvr.com/2012/05/04/pictures-investigators-seek-shirtless-heavily-tattooed-suspect/ 37

Page 38: Bridging The Gap: From Biometrics to Forensics

J-E. Lee, W. Tong, R. Jin, and A. K. Jain, "Image Retrieval in Forensics: Tattoo Image Database Application", IEEE Multimedia, Vol. 19, 2012

Page 39: Bridging The Gap: From Biometrics to Forensics

Confidence in Latent Comparison

latent (8 minutiae)

Rolled mate

Similarity score = 6.54; retrieval rank = 1; score range: [0, 13.29]

Similarity Score (log(sCOTS+1) )

Pro

bab

ility

Page 40: Bridging The Gap: From Biometrics to Forensics

Attacks on Biometric Systems Natural Fingerprint Minutiae Minutiae Density Map

Altered Fingerprint

Page 41: Bridging The Gap: From Biometrics to Forensics

Summary • Biometrics deals with real-time, automatic

recognition of a person based on body traits

• Biometric has its origins in forensics; current challenges in biometrics have forensics flavor

• Synergy between forensic science (data and domain expertise) and biometrics (models, learning algorithms and computing) is needed to put forensics science on a firm footing