Li S.Z., Jain A.K. (eds.) Encyclopedia of Biometrics

Подождите немного. Документ загружается.

37 initiated the development of a similar multi-part

international standard in 2006. This ISO/IEC Project

29109, named ‘‘Conformance Testing Methodology for

Biometric Data Inter change Formats deﬁned in ISO/IEC

19794’’ is similarly structured, and also consists of P art 1:

Generalized Conformance Testing Methodology, and

multiple modality-speciﬁc P a rts, each dedicated to one

modality. At the time of preparation of this paper, Part 1

is being circulated to the National Bodies of JT C 1/SC 37

for a Committee Draft (CD) ballot, Parts 2 (Finger Min-

utiae), 4 (Finger Image), 5 (Face Image), 6 (Iris Image),

and 10 (Hand Geometry) are in the Working Draft

(WD) stage, and a number of other P arts are expected

to be presented at the July 2008 SC 37 meeting.

Conformance Testing Activities

Approval and publishing of the conformance testing

methodolog y standards alone does not ensure con-

formance of the Biometric products to the base stan-

dards. It is imperative that the p ublished testing

standards are adopted by the Biometric community,

including technolog y vendors, system integrators,

and end-users, and implemented in the form of con-

formance testing tools, processes, and programs.

Some of these efforts are already underway, although

at the time of publication of this paper there are very

few large-scale conformance testing and conformity

assessment/certiﬁcation programs for Biometric data

interchange forma ts.

The fact that a number of Biometric industry ven-

dors claim conformance of their products to national

and international data interchange format standards

suggests that at least some ﬁrst-party conformance

testing (vendor self-testing) is taking place. It is not

known whether the standardized conformance testing

methods and procedures are used for this testing.

There are indications that governments are interest-

ed in establishing second- or third-party conformance

testing programs. For example, the United States

Department of Defense described their Biometric Con-

formity Assessment Initiative in [1] that includes the

standards-based conformance testing and reporting of

Biometric products, although it is not known when this

program will be fully implemented.

Two large scale conformance testing programs

have been established ahead of the publication of the

necessary standards, and the methods used in these

programs have inﬂuenced the development of the

standards. In the US, the certiﬁcation for biometric

algorithms to be approved for use with personal

identity veriﬁcation (PIV ) ass ociated w ith Homeland

Security Presidential Directive 12 (HSPD-12) requires

that they be tested in a program called MINEX. This

testing ensures that biometric templates produced

by th e templat e generation algorithms are confor-

mant to a proﬁled version of INCITS 37 8:2004 –

Finger Minutiae Format de ﬁned speciﬁc ally for PIV

[2]. Similarly, template generation algorithms that

are part of biometric products to be used with the

Seafarers’ Identity Docu ments programme associated

with the International Labour Organization Con-

vention No. 185 [3] must be tested by a third party

laboratory and found to be conformant to a proﬁled

version of ISO/IEC 19794 -2:2005 – Finger Minutiae

Data.

Current and Anticipated Needs

It is reasonably well understood that the major needs in

implementations of the Biometric systems can be de-

scribed as interoperability of the systems on all levels

and ability to interchange the Biometric data. These

needs can be fulﬁlled, to a signiﬁcant extent, by stan-

dardization of all aspects of Biome tric technology, in-

cluding Biometric formats for data interchange. Such

standardization requires the following:

1. Robust base standards must exist and be commonly

accepted

2. Biometric technology mu st be implemented in

conformance with the base standards

3. End-users must be able to verify conformance of

the implementation to the standards

The last element by itself can be further decomposed in

to the following:

1. Standardized conformance testing methodologies

must exist and be commonly accepted

2. Conformance testing tools implementing the stan-

dardized methodologies must exist

3. Laboratories performing the conformance testing

must exist and be able to produce standardized test

results

200

Conformance Testing for Biometric Data Interchange Formats, Standardization of

4. A process of certiﬁcation of test results by an inde-

pendent authority must exist

As shown above, developmen t o f the conformance

testing methodolog y standards is only the ﬁrst neces-

sary step in establishing the conformance testing

programs that would be able to reliably test Bio-

metric products and provide reasonably conclusive

determination of conformance (or no nconformance)

of the products to the ba se standards. While pub-

lishing of the conformance testing methodology stan-

dards, currently under development, and expeditious

development of conformance test ing tools that im-

plement these standards is recognized as an immedi-

ate need, establishing of such full-scale conformity

assessment p rograms in the near future should be

anticipated.

Gaps in Standards Development

The development of the conformance testing method-

ology standards in national and international stan-

dards development bodies is progressing quite

rapidly, and it is not unreasonable to expect comple-

tion of majority of these development projects within

the next 24 months. There are, however, cer tain gaps in

the existing projects that will need to be addressed at

some point in the future, for the testing methodologies

to remain useful. These gaps can be divided into three

categories:

1. Completeness of the standard. Currently, the con-

formance testing methodology standards don’t

provide (and probably will never provide) full,

absolute coverage of all requirements of the base

standards. For example, Type B and Level 3 testing

are currently out of scope of the existing Parts

of the conformance testing methodology stan-

dards. The motivation behind this is based on

practical reasons, and on the fact that certain

requirements can not be tested in a reasonable

manner; nonetheless the conformance testing cov-

erage is not 100% conclusive. It is expected that

additional test cases/assertions will be developed as

the conformance testing methodologies mature,

but it is unlikely that the desirable full coverage

will ever be reached.

2. Coverage of modalities. Currently, even if most

of the existing modalities’ conformance testing stan-

dardization is planned, many of the Parts have not

been initiated even as preliminary drafts. For some

relatively new modalities, such as DNA or Voice, it is

not even clear how conformance testing should

be performed. It is fully expected that eventually

conformance testing methodologies will be devel-

oped for all modalities, but at the present time this

is a signiﬁcant gap.

3. The testing methodologies are almost always ‘‘behind’’

the base standards. The base standards, however

robust and mature, are always undergoing changes,

amendments and revisions. These changes, some-

times signiﬁcant, may not be immediately be

reﬂected in the corresponding conformance testing

standard, and the time gap between the base stan-

dard change and the conformance testing method -

ology standard corresponding change may be

signiﬁcant – from several months to several years.

Summary

The increased need for interoperability of Biometric

systems, especially their ability to interchange and

share biometric data records has driven the demand

for standardization of nearly every aspect of the Bio-

metric technology. One of the primary elements of this

standardization effort has been development of the

Biometric Data Interchange Format Standa rds and

corresponding conformance testing methodologies

that ensure fulﬁllment by the biometric implementa-

tions of the requirements speciﬁed in the base

standards.

References

1. Woodward, J., Cava, S.: DoD biometric conformity assessment

initiative. Defense Standardization Program J., April/June 2005

2. Wilson, C., Grother, P., Chandramouli, R.: Biometric Data Spec-

iﬁcation for Personal Identity Veriﬁcation, NIST Special Publi-

cation 800-76-1 (2007)

3. Seafarers’ Identity Documents Convention (Revised), Interna-

tional Labour Organization Convention No. 185, International

Labour Organization (2003)

Conformance Testing for Biometric Data Interchange Formats, Standardization of

201

Conformity

▶ Conformance Testing for Biome tric Data Inter-

change Formats, Standardization of

Contact Microphones

Contact microphones ar e special microphones that

transduce not sound, but vibrations in solid bodies into

electrical signals. These can be used, for instance, to

capture speech dir ectly from the throat’s surface, which

is an interesting possibility in very noisy environments.

▶ Voice Device

Contact-Based

It refers to a device for which it is needed to touch the

sensing area to image of the ridge-valley pattern.

▶ Fingerprint, Palmprint, Handprint and Soleprint

Sensor

Contactless

Manner of biometric recognition in which it is not

necessary for the person to touch the sensor. Some

people are reluctant to touch a publicly used device

for various reasons, and some contact-type sensors

tend to become dirty through use. There are high

expectations that the contactless-type authentication

method can be used in public situations and will

prove valuable in places that require a high level of

hygiene such as hospitals.

▶ Palm Vein

Contextual Biases

Contextual biases are external inﬂuences on a percep-

tual decision due to outside factors such as details

from a case.

▶ Latent Fingerprint Experts

Continuous Classification

▶ Fingerprint Indexing

Contour Detection

Contour detection is the coarse edge detection tech-

nique that extracts the outer boundary of an object,

without extracting inner edges.

▶ Hand Geometry

Contrast

In general terms, contrast is a measure of the difference

between two objects, concepts, or other entities. In

photography, optics, and image acquisition contrast

usually refers to the difference in hue, saturation, or

intensity between two portions of an image.

▶ Iris Device

▶ Photography for Face Image Data

202

Conformity

Convenience Sample

A convenience sample is a sample that uses individuals

or sample units that are readily available rather than

those that are selected to be representative or selected

via a probabilistic mechanism.

▶ Test Sample and Size

Convergence Feature Extraction

Convergence provides a more general description of

channels and wells than force ﬁeld feature extraction.

It takes the form of a mathem atical function in

which wells an d channels are revealed to b e p ea ks

and ridges, respectively, in the function value. This

function maps the force ﬁeld F(r) to a scalar ﬁeld C(r),

taking the force as input, and returning the additive

inverse of the divergence of the force direction, and is

deﬁned by,

CðrÞ¼div fðrÞ¼

lim

DA ! 0

fðrÞdl

¼rfðrÞ¼



ð1Þ

where fðrÞ¼

FðrÞ

is the force direction, DA is incre-

mental area, and dl is its boundary outward normal.

This function is real valued and takes negative values as

well as positive ones, where negative values correspond

to force direction divergence. Note that the function is

non-linear because it is based on force direction and

therefore must be calculated in the given order.

▶ Physical Analogies for Ear Recognition

Copula

A copula is a multivariate joint distribution that is

deﬁned on the n-dimensional unit cube [0; 1]

such

that every marginal distribution is uniform on the

interval [0; 1]. A copula can be used to capture the

dependencies or associations that exist between vari-

ables in a multivariate distribution. In the context of

score-level fusion, a copula may be used to describe the

correlation between multiple matchers.

▶ Fusion, Score-Level

Core

The topmost point on the innermost recurving ridge

of a ﬁngerprint. Generally, the core is placed upon

or within the innermost recurve of a loop as described

in the Standards document ISO/IEC 19794-2: Bio-

metric Data Interchange Formats – Part 2: Finger

Minutiae Data.

▶ Fingerprint Templates

Correct Index Power

The ratio of correctly retrieved ﬁngerprints over the

size of the database.

▶ Fingerprint Indexing

Correct Reject Power

The ratio of correctly rejected reference ﬁnge rprints

over the number of query images not having a

corresponding ﬁngerprint in the database.

▶ Fingerprint Indexing

Correct Reject Power

203

Correlation

The degree of relationship between two variables as

expressed using a single measure. The Pearson correla-

tion coefﬁcient is an example of one such measure. In

the context of multibiometrics, the correlation be-

tween the genuine (or impostor) match scores of two

biometric matchers can have a bearing on the perfor-

mance of the fusion scheme used to combine them.

▶ Multibiometrics

Correlation Map

Correlation map is a two dimensional array of correla-

tion values in the range [1, +1]. These correlation

values are obtained by computing the normalized local

correlation between two curvature feature maps. Each

pixel in the correlation map represents local

correlation between the corresponding pixels in the

curvature maps being matched.

▶ Palmprint Features

Correspondence

▶ Human Detection and Tracking

Cost Function

Tracking is an estimation process that computes the

position of a target based on optimization of a criterion

that relates the observations with the estimates . The

criterion is represented mathematically using the cost

function.

▶ Face Tracking

Countermeasures

Liveness Check is a validation that the biometric char-

acteristic is the true characteristic of the presenting

individual by the measurement of expected live fea-

tures such as pulse, temperature, humidity, movement

etc. as appropriate to the biometric characteristic.

Artifact detection is the detection of an artifact that

has been presented by measurement of speciﬁc prop-

erty of known artifacts (e.g., silicone rubber or gelatine

ﬁnger; photograph of face etc. depending on the bio-

metric characteristic). A point to be noted here is that

the liveness check and artifact detection are comple-

mentary approaches to countering the use of artifacts.

Biometric data encryption is a cr yptographic tech-

nique used to safeguard the conﬁdentiality of biomet-

ric data.

Biometric data signing is also a cryptographic tech-

nique used to safeguard the integrity of biometric data.

Cryptographic timestamps/session keys are crypto-

graphic techniques used to counter capture/replay attacks.

Supervised operation is a powerful countermeasure

against a range of threats that can occur when a subject

is interacting with a biometric system during enrol-

ment and veriﬁcation operations. It can be an effective

countermeasure to the use of artifacts, mimicry, and

physical attacks.

Security audit is a useful post-event analysis of

security log to check e.g., suspicio us events, integrity

of system conﬁguration, procedural compromises etc.

Performance audit is an ofﬂine check that the sys-

tem performs to a level that safeguards security, e.g.,

cross checkin g enrolment references against each other

to ensure that adequate separation between references

exists and there are no apparent cases of multiple

enrolments by a single individual.

▶ Biometric Security, Standardization

Counter Sign

1. A second sign to provide the proofs of approval

and/or receipt as on a previously signed document

such as a contract or a money order.

204

Correlation

2. Another sign, word, or signal used for replying the

sign from an anonymous or a hidden person.

▶ Signature Matching

Covariate

A covariate is a secondary variable or factor that can

affect the relationship between the dependent variable

and other independent variables of primary interest.

For instance, in biometrics view-point or illumination

is a secondary variable that can impact the relationship

between recognition ability, the dependent variable,

and identity, the independent variable. Covariate

needs to be controlled or monitored in a biometric

experiment.

▶ Evaluation of Gait Recognition

Covariate Studies

In multivariate statistical analysis the aim is to study

the independence and interdependence of two or more

random variables. In such a study, there could be other

confounding factors that affect the statistical analysis

of the variables of interest. The study of these con-

founding variables and desig n of appropriate theoreti-

cal and experimental set-ups in order to eliminate the

effect of these confounding random variables is called

covariate analysis.

For instance, in designing statistical methods for face

recognition, the statistical relationship between the iden-

tity variable and the face images obtained has to be

studied. The face images obtained are also dependent

on various confounding factors such as illumination,

pose, expression, and age of the subject, and the camera

internal parameters. These confounding variables are

considered covariates for the problem of image/video-

based face recognition. Similarly, in the case of gait-based

person identiﬁcation from videos, it would be interesting

to learn the statistical relationship between the subject’s

identity and the videos. In this case, there are several

environmental confounding variables such as the cloth-

ing of the subject, the shoe-type of the subject, the surface

of walking, the camera view and the presence of other

occluding objects enter a briefcase etc.

Two major approaches are available deal with to the

effect of covariates – enumeration and marginaliza-

tion. Enumeration refers to techniques where one

learns the statistical relationship between the random

variables of interest for every possible realization of the

confounding variable. This approach works well for

settings such as the presence/absence of a briefcase,

where there are very few distinct values that the con-

founding random variable can take. In marginalization

a joint probability density function of the relevant and

the confounding random variables is ﬁrst developed

and then marginalized (integrated) over the confound-

ing random variables in order to make an inference.

▶ Gait Biometrics, Overview

Craniofacial Reconstruction

▶ Skull, Forensic Evidence of

Craniofacial Superimposition

▶ Skull, Forensic Evidence of

Credential Hardening

The process of increasing the trust associated with

typical credential sets such as user names and pass-

words is credential hardening. This increased trust is

often achieved throug h a biometric augment as in

keystroke recognition. This is sometimes also referred

to as password hardening.

▶ Keystroke Recognition

Credential Hardening

205

Credentialing System

Credentialing System uses a physical credential, such as

a smartcard, as a means to aut henticate the identity of

a credential holder for purposes of author ization. It

includes the registration process as well as the

subsequent operational use of the credential. Registra-

tion may include enrollment, identity prooﬁng, back-

ground checking, card production, and issuance.

Possible uses are many, but generally include physical

and logical access control as well as beneﬁts eligibility/

redemption and other privilege or entitlement claims.

▶ Registered Traveler

Credit Check

▶ Background Checks

Crew Designs

▶ Test Sample and Size

Criminal History Check

▶ Background Checks

Criminal Law Enforcement

▶ Law Enforcement

Criminal Record Search

▶ Background Checks

Cross-Modality Face Biometrics

▶ Heterogeneous Face Biometrics

Cross-Validation

A popular approach to estimating how well the result

learned from a given training data set is going to

generalize on unseen new data. It partition s the train-

ing data set into k subsets with equal size, and then uses

the union of k¡1 subsets for training and the remaining

subset for performance evaluation. The ﬁnal estimate

is obtained by averaging after every subset has been

used for evaluation once. A popular setting of k is 10

and in this case it is called as 10-fold cross-validation;

another popular setting of k is the number of training

examples and in this case it is called as LOO (i.e.,

Leave-One-Out) test.

▶ Ensemble Learning

Cryptography

The science of transforming messages or data into in-

comprehensible formats for the purposes of conﬁdenti-

ality, integrity, authentication, or non-repudiation of

origin. Cryptographic systems have classically involved

two parties, a sender and a receiver whom w ish to

communicate a message secretly, although modern

uses include secure data storage as well as digital signa-

tures. Despite the intended use, cryptographic systems,

sometimes referred to as cryptosystems involve two

main blocks: encryption and decryption. Encryption is

the process of encoding data into an unreadable format

206

Credentialing System

(sometimes referred to as ciphertext) through the use of

a cryptographic key. Decryption is the process of decod-

ing encrypted data (ciphertext) back into a comprehen-

sible format through the use of a cryptographic key and

a complementary algorithm tied to encryption. Crypto-

graphic algorithms can be symmetric, requiring the

communication of a single secret key between sender

and receiver or they can be asymmetric relying on pub-

lic-private key pairs that do not explicitly require the

transmission between sender and receiver.

▶ Iris Digital Watermarking

Curse of Dimensionality

The demand for a large number of samples grows

exponentially with the dimensionality of the data (fea-

ture) space, and so is the difﬁculty to ﬁnd global

optima for the paramet er space, i.e., to describe the

data space. This phenomenon is known as the ‘‘Curse

of Dimensionality.’’ The fundamental reason for this

limitation is that high-dimensional functions have

the potential to be much more complicated than low-

dimensional ones, and that those complications are

harder to discern. A simple but effective way to allevi-

ate this problem is to reduce the number of dimensions

of the data by eliminating some coordinates that seem

irrelevant or extract and select salient and discrimina-

tory features for data representation.

▶ Biometric Algorithms

▶ Fusion, Feature-Level

▶ Multibiometrics

Curse of Misalignment

▶ Face Misalignment Problem

Cursive

In a fully cursive handwriting style, the writer is inclined

to connect all adjacent letters in a smooth way. What

makes it more complex is that different people may have

different habits to connect even the same character pair.

For example, some people like to connect the t-bar with

other letters, while others may connect the t-stem

to neighboring characters. Real-world handwriting is

often in a mixture of cursive and handprint styles.

▶ Signature Sample Synthesis

Custody Suite

Custody suites are areas of police stations in which

suspects arrested in connection with particular circum-

stances are held and questioned in the furtherance of

the enquiries by the police.

▶ Footwear Recognition

Cut Finger Problem

▶ Anti-Spooﬁng

▶ Fingerprint Fake Detection

▶ Liveness Detection

Cut Finger Problem

207

Dactyloscopist: Fingerprint

Examiner

▶ Fingerprint, Forensic Evidence of

Data Hiding

Data hiding techniques can be used to insert additional

information, namely the watermark, into a digital ob-

ject. The watermark can be used for a variety of appli-

cations ranging from copy protection, to data

authentication, ﬁngerprinting, broadcast monitoring,

multimedia indexing, content based retrieval applica-

tions, medical imaging applications, etc. Within the

framework of biometrics, robust data hiding techni-

ques can be used to embed codes or timestamps into

the template, in such a way that after the expiration

date the template is useless. Another perspective is to

hide the biometric templa te in a digital object to make

it invisible when either transmitted or stored.

▶ Conformance Testing for Biometric Data Inter-

change Formats, Standardization of

▶ Iris Template Protection

Data Interchange Format

▶ Common Biometric Exchange Formats Framework

Standardization

Data Protection

▶ Privacy Issues

Database Filtering

Filtering refers to limiting the number of entries in a

database to be searched, based on characteristics of the

interacting user. For example, if the user can be identi-

ﬁed as a middle-aged male, the search can be restricted

only to the subjec ts with this proﬁle enrolled in the

database. This greatly improves the speed or the search

efﬁciency of the biometric system. Filtering reduces the

probability of obtaining a wrong match, but this is

offset by the fact that the errors in ﬁltering also reduce

the probability of obtaining a correct match. Hence, in

general, ﬁltering drastically reduces the time required

for identiﬁcation but can degrade the recognition

accuracy.

▶ Soft Biometrics

Daubert Standard

The Daubert standard is a legal precedent set in 1993

by the Supreme Court of the United States regarding

the admissibility of expert witnesses’ testimony during

2009 Springer Science+Business Media, LLC