A Bayesian Approach for Modeling SensorInfluence on Quality, Liveness and Match Score Values in Fingerprint Verification

A Bayesian Approach for Modeling SensorInfluence on Quality, Liveness and Match Score

Values in Fingerprint Verification

Ajita Rattani1, Norman Poh2 and Arun Ross1

1Dept. of Computer Science and Eng., Michigan State University, USA

http://www.cse.msu.edu/~rossarun/i-probe/2Dept. of Computing, University of Surrey, UK

Abstract

• Recently a number of studies in fingerprint verification have combined match scores with quality and liveness measure in order to thwart spoof attacks.

• However, these approaches do not explicitly account for the influence of the sensor on these variables.

• In this work, we propose a graphical model that account for the impact of the sensor on match scores, quality and liveness measures.

• Effectiveness of the proposed model has been assessed on the LivDet 2011 fingerprint database using Biometrika and Italdata sensors.

Outline

• Fingerprint Spoof Attacks • Fingerprint Liveness Detectors• Problem Statement and Contributions• Proposed Graphical Model• Experimental Validations• Conclusions

3

Spoof Attacks

• A spoofing attack occurs when an adversary replicate the biometric trait of another individual in order to circumvent the system.

• These attacks pose the most severe threat to biometric systems as

– they can be easily performed using simple techniques and commonly available materials and,

– do not require any knowledge of internal functioning of the system.

4

Contd..,

A forger counterfeits a biometric sample of a given user to gain unauthorized access.

5

Fingerprints and Spoof Attacks

• Studies have shown that fake fingerprints can be easily fabricated using commonly available materials like silicone, latex etc., [Matsumoto et al., 2002, Yambay et al., 2012].

• Fingerprint liveness detection algorithms have been proposed as a counter-measure against spoof attacks.

Matsumoto et al., Impact of artificial gummy fingers on fingerprint systems, Opt. Sec. Counterfait Deterrence Techniq. IV 4677, 275-289.Yambay et al., Livdet2011 – fingerprint liveness detection competition, ICB, 2012

6

Fake Fingerprint Fabrication Process

SiliconeLatexEcoflex

7

Fingerprint Liveness Detectors

• They aim to discriminate live biometric samples from the spoofed (fake) artefact.

• The algorithms for fingerprint liveness detection examine the textural, anatomical and/or physiological attributes of the finger.

• The output of these liveness detection algorithms is a single-valued numerical entity referred to as liveness measure.

8

Live / FakeLive / Fake

EcoFlex

Gelatin

Latex Silgum

Live

9

An example training based fingerprint liveness detector

Liveness Detector and Biometric System

• Liveness detection algorithms are not designed to operate in isolation; rather, they have to be integrated with the overall fingerprint recognition system.

• Accordingly, recent studies have combined match scores generated by a fingerprint matcher with liveness values, as well as image quality, in order to render a decision on the recognition process.

State of the Art

Reference Variables combined

Scheme Database

Marasco et al., BTAS,

2012

fingerprint match scores + liveness values

BayesianBelief Network

LivDet 2009

Chingovska et al.,

CVPRW, 2013

face match scores +

liveness values

Logistic Regression

Replay attack, 2011

Rattani et al., ICB, 2013

fingerprint match scores +

quality + liveness values

Density-basedfusion

framework

LivDet 2011

Problem Statement

• However, in the aforementioned schemes, the influence of the sensor on the 3 variables - match scores, liveness values and quality has not been considered.

• Such a consideration is essential for several reasons: (a) the quality of an image is impacted by the sensor used;

(b) most liveness measures are training-based and are impacted by the sensor that was used to collect live and spoof data;

(c) understanding sensor influence, can help in facilitating sensor interoperability for fingerprint matchers and liveness detectors.

Contributions

• Development of a graphical model for fusing match scores, liveness measures and quality values while accounting for sensor influence;

• Implementation of the proposed model using a Gaussian Mixture Model (GMM) based Bayesian classifier;

• Evaluation of the proposed model using fingerprint data from two different sensors in the LivDet 2011 fingerprint database.

Notations

• y = match score between pair of fingerprint images.

• q = quality of a fingerprint image

• l = liveness measure of a fingerprint image

• k = class label i.e., {C, I, S}

• d = fingerprint acquisition sensor

Graphical Model Conditional Probabilities

Bayesian Classifier

a) Model A(conventional classifier)

•

b) Model B(Rattani et al., ICB, 2013)

•

k

y

k l

yq

Contd..,

c) Proposed Model C

•

•

k d

y l q

Conjectures:1. Match scores, quality and liveness measures are sensor dependent

(d {y, q, l})

2. Further their exist no significant correlation between quality (q) and liveness

measures (l).

Experimental Validations

• Dataset: LivDet 20111

• Liveness measure: LBP + SVM2

• Quality: IQF freeware developed by MITRE3

• Match Scores: NIST Bozorth3 matcher4

• Performance metrics :

– OFAR : Overall False Acceptance Rate

– GAR: Genuine Acceptance Rate

– O-EER : Overall Equal Error Rate

1Yambay et al., Livdet2011 - fingerprint liveness detection competition, ICB, 20122Nikam and Aggarwal, Local binary pattern and wavelet-based spoof fingerprint detection,

IJB, 20083http://www.mitre.org/tech/mtf/4http://www.nist.gov/itl/iad/ig/nbis.cfm

1. Performance Evaluation of Model B under Cross-sensor Evaluation

k l

yq

2. Validation of the Conjectures for Model C

0 50 100 150 200 250 3000

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

0.1

Scores

Lik

elih

oo

d

Genuine-BiometrikaSpoof-BiometrikaImpostor- BiometrikaGenuine- ItaldataSpoof- ItaldataImpostor- Italdata

Biometrika Italdata20

30

40

50

60

70

80

Sensors

Qu

alit

y

Match scores and quality measures are sensor dependent

Contd..,

0 0.2 0.4 0.6 0.8 10

0.5

1

1.5

2

2.5

3

3.5

4

4.5

Liveness measure

Lik

elih

oo

d

BiometrikaItaldata

Liveness measures are sensor dependent

No significant correlation exist between quality and liveness measure ( =0.25)

3. Assessment of Model C in a Multi-sensor Environment

Conclusions

• Recent studies have addressed the security of fingerprint verification system by combining match scores with quality and liveness measure.

• We advance the state of the art by modeling sensor influence on them.

• This is realized through a graphical model that account for the impact of sensor on liveness, quality and match score values

Contd..,

• Experimental investigations on the LivDet11 fingerprint database indicate that

– Existing training-based fusion framework cannot operate effectively in a multi-sensor scenario

– The proposed graphical model can effectively operate in a multi-sensor environment

• The effectiveness of the proposed model can be further enhanced by improving the sensitivity of the underlying liveness detector

Thank you