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

Подождите немного. Документ загружается.

clocks that imprint the time & date onto a time card.

To decrease data entry errors and administrative

overhead, compan ies have migrated to electronic TA

systems that use an ID card or ID number to identify

each employee. However, card-based and ID-based

systems leave open the loophole of buddy-punching,

where employees punch-in or punch-out for one an-

other to be paid for hours they did not work. This

loophole is being closed by biometric TA systems that

ensure all employees are paid for exactly the hours they

are physically at work.

References

1. Rowan L.: Workforce Management for SMB’s: Gaining Control

of Your Most Expensive Resource, IDC, February 2007

2. http://www.dol.gov/esa/whd/regs/compliance/whdfs22.pdf

3. http://www.dol.gov/compliance/guide/minwage.htm

4. http://recognitionsystems.ingersollrand.com/ﬁles/600.pdf

5. http://www.timeandtech.com/Docs/ttG-tt3-Datasheet.pdf

6. Rand, I.: Case Study #51 – HandPunch...McDonald’s is Lovin’

It!, February 16th, 2007. http://recognitionsystems.ingersoll-

rand.com/cs/cs.php?id=35

7. Simmons, N.J.: GeoSlavery: Big Brother. Black Star News, Septem-

ber 5th, 2007. http://www.blackstarnews.com/?c¼125&pa¼3660

8. Dir. Steven Spielberg.: Minority Report. Based on short story by

Philip K. Dick. Perf. Tom Cruise. Twentieth Century Fox and

Dreamworks SKG, 2002

9. Clarke, G.T.: Biometrics in the workplace – where to draw the

line? Personal Information Protection Act Conference 2006,

April 26 and 27, 2006. Calgary, AB. http://www.governmente

vents.ca/pipa2006/presentations/a1_gary_clarke_paper.pdf

10. O’Dell, R.: Buddy-punching of timecards is OK, union backers

tell council. Arizona Daily Star (Tucson, AZ), September 7, 2006.

http://www.azstarnet.com/metro/145557.php

Time and Attendance Terminal

A time and attendance terminal is a device that is more

sophisticated than a time clock. It uses electronic data

capture record to record a clock-in or clock-out transac-

tion when the employee presents an authorized creden-

tial to the clock. The credential may be an ID number,

proximity card, smartcard, biometrics etc. There are

three main time and attendance terminals, card or prox-

imity-based, ﬁngerprint, and hand geometry.

▶ Time and Attendance

Time Clock

Time clock is a mechanical or electronic device that is

used to track employee hours. When interacting with

the device, employees either ‘‘punch-in’’ or ‘‘clock-in’’

when registering the start of the time, and ‘‘punch-out’’

or ‘‘clock-out’’ when leaving work or the assigned task.

▶ Time and attendance

Time Series

A time series is a sequence of data points, measured

typically at successive times, spaced at (often uniform)

time intervals. Learning on time series data attempts to

understand the relationship between data and time,

e.g., making forecasts. The order of the data points

along the time axis is an important factor to be con-

sidered in the learning methods.

▶ Incremental Learning

Tongue-Print Recognition

A biometric technology for automatically identifying

or verifying a person using information of tongues.

As the tongue can be protruded from the body for

inspection, the shape and texture information can be

acquired from its images as ‘‘tongue-print’’ for the rec-

ognition process. Unlike face and ﬁngerprint, the tongue

is an internal organ and well protected in the mouth, so

it is basically immune to forgery. This is advantageous

for protecting user’s privacy and security.

▶ Biometrics, Overv iew

1340

Time and Attendance Terminal

Tooth Biometrics

▶ Dental Biometrics

Top and Secondary Choices

Every biometric matcher employed for user recogni-

tion compares the presented biometric with the stored

templates and generates the matching scores corres-

ponding to each of the possible user identities . The

top choice refers to the user identity corresponding to

user template generating the highest matching score.

The secondary choices refer to the remaining choices of

the possible user identities corresponding to the tem-

plates that do not generate the highest matching score.

If the difference between the highest matching score

and the second highest matching score is large, there is

high probability that the top choice represents the

correct user identity. However, if this difference is

small, the top choice may not represent the correct

user identity and secondary choices become important

in generating the decision s.

▶ Fusion, Rank-Level

Total Transaction Time

▶ Operational Times

Touch Tablet

▶ Digitizing Tablet

Touch-Screen

A touch-screen is an electronic display that locates and

captures the contact of objects within the display area.

Touch-screens can be divided in two main ty pes: those

oriented to ﬁnger-input and the on es oriented to

styles. Touch-screens of stylus oriented hand-held

devices are based on a resistive principle and are not

oriented to operation with the user ﬁngertips. Two

separated metallic layers are connected when the screen

is pressed. The position of the contact point can be

accurately detected, bu t only when the stylus is in

contact with the surface, contrary to pen tablets.

▶ Signature Databases and Evaluation

Tracing

The following of adjacent parallel ridges over a certain

length in the vicinity of the supposed event and by

calculating whether the count of ridges in between

increases or decreases, thus deﬁning the beginning or

ending of a ridge in between.

▶ Fingerprint Matching, Manual

Training

A process used to determine the values of the classiﬁer

parameters with the help of the prototypes of the data

to be recognized. The data used during this step con-

stitute the training set.

▶ Support Vector Machine

Training Data, Sufficiency

A training data set is sufﬁcient for a learning task if all the

knowledge required for correct future predictions is

Training Data, Sufficiency

1341

contained in the data set. In practice, it is hard to judge

whether a training set is sufﬁcient for a given task.

Generally, if the distribution estimated from the training

data covers all possible examples that may appear in the

task, then the training data is sufﬁcient for the task.

▶ Incremental Learning

Training Signature

Signature used in the enrollment phase.

The training signatures are used for enrollment of a

reference template, reference model parameter estima-

tion, decision threshold estimation, and the like.

▶ Signature Recognition

Transfer Learning

▶ Incremental Learning

Transformation

The transformation refers to the process of normalizing

the output (score) for a matcher to a desired range. The

range of output matching scores generated from the

different biometric matchers can vary signiﬁcantly.

This variation can be attributed to the different distance

criteria used to generate matching scores or the differ-

ent biometric features employed by different matchers.

▶ Fusion, Rank-Level

Transmission-Based Touchless

Finger Imaging (TTFI)

TTFI refers to tou chless ﬁngerprint sensor.

▶ Fingerprint, Palmprint, Handprint and Soleprint

Sensor

Transportable Asset Protection

ANTHONY P. RUSSO

Atrua Technologies, Inc, New York, NY, USA

Synonym

Asset protection

Definition

Transportable Asset Protection is a means by which

personal user secrets and privileges, stored in digital

form on a portable device such as a smart card or a

mobile handset, are secured from unauthorized access

and/or use. The assets, being carried on one’s person,

are highly prone to theft or loss, making the need for

security that much greater. The unique challenge posed

by Transportable Asset Protection is to provide ade-

quate security and performance using the lightweight

processing resources available on the mobile device in a

portable and interoperable way.

Introduction

The nature of a transportable asset is that it is carried

from place to place and therefore prone to theft or loss.

Once lost or stolen, the vessel (or vault) holding the

asset is subject to relentless physical attack by a hacker,

possibly for an indeﬁnite period of time. As our society

grows more mobile, our transported assets become

more valuable and our thieves more sophisticated,

hence the need for greater security and privacy is

increasing ra pidly. Transportable asset protection in

the modern sense was ushered in with the invention

of smart cards. Incorporating biometric authentica-

tion into such systems is a relatively new application

and research areas designed to address the unique

needs of mobile users for whom the standard methods

of authentication are either too cumbersome or not

secure enough.

A transportable asset protection system that uses

biometrics will incorporate at least one mobile com-

puting device or host capable of storing the asset.

Examples of mobile devices include smar t cards,

SIM cards, PDAs, mobile handsets, FLASH storage

tokens, notebook computers and other such relatively

1342

Training Signature

small lightweig ht items that can store data and/or

access services. Except for the notebook computer,

these devices are all examples of embedded systems,

and therefore include some limited – typically very

limited – dynamic memor y and compute power.

An asset can be any digital item of value to a user.

Asset types fall into three broad, somewhat overlap-

ping categories: (1) user secrets, (2) physical or logical

access to places or services and (3) rights. Secret data is

most commonly in the form of username and pass-

word data that can be used to access large amounts of

conﬁdential information, such as corporate databases

or encrypted documents. In more secure systems the

secret may be the private key of a digital certiﬁcate,

used to cryptographically encode and sign data. Accor-

dingly, these types of assets have a hard-to-measure

value to all involved parties that depends on the con-

ﬁdentiality of the secret.

Examples of privileged access include access to

wireless or corporate networks, use of services such as

internet access, online banking, or mobile-commerce

transactions. While passwords and keys are sometimes

used to access these services, many such system s

require a secure hardware token (e.g., smart card) for

multi-factor authen tication. These kinds of assets usu-

ally have a directly measurable monetary value to the

user and/or the service provider.

Finally, the last category of assets is the right to use

a device or a digital ﬁle. Devices with restricted use are

myriad, including automobiles, photocopy machines,

cell phones and ﬁrearms. However, digital rig hts are

also becoming increasingly important: the right to

listen to a downloaded song or watch a movie or

open a document. Digital Rights Management and

transportable asset protection are likely to become

increasingly coupled in the future. In most of these

cases the user has a reason to prevent third parties from

obtaining the asset; in other cases it is the serv ice

provider that the vested interest.

From an implementatio n perspective, of parti-

cular interest – due both to their importance and

popularity – are those systems composed of a host

(e.g., mobile handset) and a hardware token or other

▶ secure element (SE) used to store secrets and/or

provide access to services. The biometric sensor is

attached to the host, which usually offers little or no

physical protection from hackers and may provide

dubious overall security. In contrast, the secure ele-

ment – physically or wirelessly connected to the host

– is typically a closed, tamper-resistant, well-trusted,

standardized and usually certiﬁed system with a very

lightweight microcontroller and some FLASH memory

or EEPROM storage. Smart cards and SIM cards are

the most prevalent examples of SEs that are completely

separate entities from the host.

▶ Near Field Commu-

nication (NFC) controller chips, used for short ra nge

mobile transactions, are an example of a device that

can be permanently integrated into the host hardware

itself. The Trusted Platform Module (TPM) is the

secure element insid e the high-end notebook compu-

ters that incorporate them.

A well-designed secure element is a single-chip

entity designed to exacting speciﬁcations to achieve a

‘‘black box’’ nature. In the locked state, therefore, an SE

is fairly impenetrable. However, the vast majority of

these devices are unlocked by entering a 4-digit Per-

sonal Identiﬁcation Number or PIN, which in itself is

not terribly secure, since a PIN entry can be observed

by onlookers and does not tie the unlocking process to

a person as would biometric authentication. Further-

more, when cryptography is employed in a system,

such as digitally signing a document, the expectation

is for a very high level of security and trust. However,

those having experience with such systems know that

the weakest link is often in protecting the private key,

something that most existing secure elements rely

upon and that a simple PIN can provide. Without a

biometric match, the system-wide security of a 256-

bit cryptographic key is reduced from 2

256

to 10

4

per break-in attempt. Hence the need for biome trics is

clear, not only for increased security but also

convenience.

Despite the drawbacks of the PIN, in the discussion

that follows the host is considered to be a non -secure

entity while the secure element is maximally trusted.

Therefore, the more processing done on the secure

element, the safer the system. As will become clear,

the most secure platforms are the ones in which the

host and the secure elements are the same physical

device.

System Architectures

Implementations of transpor table asset protection sys-

tems are varied. The most tr ivial, least secure architec-

ture employs only a host, with no secure element

connected to the system. Security risks abound, the

Transportable Asset Protection

1343

most important of which is the lack of any secure

element to protect the assets. While the host itself

might employ security features such as user locking,

those are easily circumvented by a thief who steals the

phone and accesses the ﬁlesystem through alternate

hardware means. In short, this architecture is too inse-

cure to constitute a useful asset protection system.

However even those implementations that use a

secure element can vary widely in effectiveness.

Figures 1, 2 and 3 show examples of transportable

asset protection system s that offer basic security

(Fig. 1), very good secu rity (Fig. 2) and maximal

security (Fig. 3). Referring to the basic security archi-

tecture, a secure element is used only to protect the

assets and nothing else. In this case, the biometric

sensor, template extraction algorithms, and matcher

all execute from the host CPU, and enrolled template

data is stored on the host’s ﬁlesystem. This imple-

mentation is the easiest to engineer because typically

the host has much more processing power and memo-

ry than a secure element. Assets are protected with

only a PIN, but putting aside that shortcoming for

the moment, a number of additiona l security risks

exist with such a system.

Transportable Asset Protection. Figure 1 Diagram describing a basic transportable asset protection system employing

a host and a Secure Element. The host is connected to a biometric sensor and performs all template extraction

tasks. The secure element is used only to store assets, and must be unlocked using a PIN. Once unlocked, the secrets

are accessible by the host.

Transportable Asset Protection. Figure 2 Diagram of a transportable asset protection system employing the MOSE

architecture. The host is connected to a biometric sensor and performs all template extraction tasks. Assets and enrolled

templates are saved on the secure element, which has the ability to perform a biometric match. The SE is unlocked via a

PIN or by sending it a live scan template that matches the stored enrolled template. Once unlocked, the secrets are

accessible by the host.

1344

Transportable Asset Protection

The ﬁrst security hole is that templates must be

stored on the ﬁlesystem. This means these templates

are readable by anyone operating the host device, repre-

senting both a security and a privacy issue, as the tem-

plates themselves could be considered an asset. While the

natural inclination is to encrypt this data, the problem is

that encryption requires a private key, and that key must

be stored someplace for use later during decryption.

Usually the key is hidden somewhere on the ﬁlesystem

or in the code itself, but this kind of ‘‘security by obscu-

rity’’ is a major risk nonetheless [1]. If the host does not

guard against tampering, then it would be quite easy

for a hacker to access the data in unencrypted form by

tapping the memory bus during the match process,

where it must be in plaintext form prior to use. Even

easier hacks involve modifying the results of the

matching algorithm to force a match even when one

does not exist. [2]

A variation on this system uses the secure ele-

ment to simply store the enrolled templates, thereby

removing the need to store them on the host and

offering increased protection if the host is lost or

stolen. In this case, the templates are stored safely on

the tamper-proof secure element, where presumably

only very advanced hacking techniques could unlock

it. This approach offers only a marginal increase in

security, however, if the host and the secure element

are both lost, which can often be the case (e.g., a

mobile phone + SIM card). The memory tapping

techniques are still effective as the secrets must leave

the card before they are matched, and altering the

match results is still possible because the match occurs

on the host. Of note is that a PIN must still be used

to unlock the secure element before a biometric match

can be performed, and though this can be considered a

security feature, it removes the user convenience that is

a major selling point of biometric implementations,

since PIN entry, especially on a small mobile device, is

often cumbersome.

A much more secure approach is shown in Fig. 2,

where the matching algorithm is executed inside the

secure element itself, thereby eliminating the need for

PIN entry entirely. This conﬁguration is commonly

referred to as

▶ Match-on-Card (MOC) when used

with a smart card or SIM card [3]. A more general

term is Match-on-Secure-Element (MOSE). In this

architecture, the sensor and templa te extraction are

still performed on the host, bu t during the veriﬁcation

process, the template is sent to a locked secure element,

which only opens upon a successful match against the

enrolled templates stored there. This approach has a

number of important advantage s. Firstly, the enrolled

template data, once stored, never leaves the secure ele-

ment for any reason; it is well-protected under all cir-

cumstances. Secondly, it is not possible to simply alter

the match results because the matcher executes in a safe,

black box environment. One other important aspect of

this implementation is that it allows for secure replace-

ment of the PIN with a biometric. This preserves both

the convenience and security largely responsible for the

market success of biometric systems.

The attacks on such systems tend to focus on the

host processing: both sensor data and extracted tem-

plates can be snooped and modiﬁed, allowing for

Transportable Asset Protection. Figure 3 Diagram of a transportable asset protection system employing a host that

is itself a secure element. All processing is performed on the secure host, making it the most trustworthy, albeit

the most costly design choice.

Transportable Asset Protection

1345

replay and denial-of-se rvice attacks. These issues are

addressable for the most part, as this class of hacks can

only be launched during the authentication process

and therefore will be useless if the device is stolen

afterward. Overall, this is the most cost-effective and

technologically feasible approach to transportable asset

protection.

The least cost-effective approach, but offering the

highest level of security, is one in which the sensor,

template extraction, and matching are all located on

the secure element (see Fig. 3). In this case it is not

possible for a hacker to snoop or alter any biometric

data or algorithms. The most promising attacks on

such a system involve fake-ﬁnger spooﬁng, for which

there are countermeasures available. However, such a

system is costly, because the microcontroller on the

secure element can no longer be lightweight: much

more memory and computing power would need to

be available to process the raw biometric data into a

template and perform the match. Other considerations

involve powering such a secure element and addressing

the mechanical mounting issues presented by placing a

sensor on a smart card or other miniature dev ice,

where it must meet ﬂexibility and height standards to

be universally adopted.

Role of Standards

As noted, the secure elements used to protect trans-

portable assets are t ypically standard-based tokens

such as smart cards and SIM cards. Being portable,

standards have evolved so that these tokens can inter-

operate with a variety of different host hardware, as

long as the token and the host support the same stan-

dard. For commercial deployments, it has become

increasingly difﬁcult to attract interest in proprietary

solutions and formats due to this expectation of

interoperability.

A good example to illustrate the need for standards

is a GSM-based mobile handset, which uses a SIM

card to verify the user and provide wireless network

access. Let’s assume this is a MOC system where the

biometric sensor and template extraction algorithms

are on the handset, and the removable SIM card has

embedded in it a biometric matcher where the en-

rolled templates are stored on the SIM card. Given

that wireless network operators require that a SIM

card be usable in any GSM handset, it is prudent to

ensure that handsets adhere to the same template

format standard that the SIM card does, or else the

biometric matching would not work. While it is pos-

sible and sometimes necessary to fall back to having

the user enter a PIN instead, this scenario is undesir-

able from the operators’ point-of-view as it weakens

the value of deploying biometric SIM cards to their

network subscribers.

The ISO-7816 standard [4] is an example of

how smart cards interact with their hosts at both an

electrical, physical, and communication level, and ISO-

7816-11 even allows for biometric Match-on-Card

commands. But as the above example illustrates, stan-

dard commands are not enough to ensure compati-

bility. In a MOSE architecture the host and secure

element must agree on the nat ure and format of bio-

metric templates to be used for enrollment and

unlocking . Although some widely-accepted standards

exist for face recognition [5] and other biometrics [6],

ﬁngerprints currently offer the most options in this

regard, and the only ones thus far designed to directly

address the requirements of a MOSE system [7].

Algorithmic Challenges and

Approaches

To fully appreciate the challenges involved with creating

a biometric matcher for use within a secure element, it is

necessary to understand the execution environment in

which it runs. The most common secure elements em-

ploy 8-bit or 16-bit microcontroller units (MCUs), hav-

ing available RAM from 500 bytes to 2KB, and 128KB or

less of ROM. Internal clock rates are typically below 30

MHz, often in the 7–15 MHz range. There is usually no

ﬂoating-point support, and sometimes no native hard-

ware support for signed integer math, as is the case with

an 8051-class MCU [8]. The computational power of

the Pentium-class CPU found on most PCs is roughly

3 orders of magnitude more than that of the MPU

found on most secure elements; available RAM is 6

orders of magnitude more. While the processing power

for secure elements does increase from year to year,

low-end security processors continue to dominate the

mass market due to its low cost. Therefore, it is impor-

tant to design it with these in mind.

The simplest overall approach is to design the bio-

metric template so that it is a statistical feature vector,

allowing matching to be done via a linear or non-linear

1346

Transportable Asset Protection

classiﬁer, which is more suitable for imple mentatio n

on low-end MCUs. Cer tain biome trics lend themselves

more easi ly to lig htwe ig ht matc hing algor ithms than

others. Iris match ing and DNA match ing are two

example s w here most of the comp utational effor t is

in the templa te extraction step, while the match ing

step is relatively straig htfor ward. Standard ﬁng erprint

match ing does not fall into this categor y – it t y pi cally

req uires both local and g lobal geometric informati on –

which is unfor tunate given its populari t y w ith and

suitabili t y fo r mobile platfor ms.

Nonetheless, most of the implementation issues

have been overcome, and ﬁngerprint-based MOSE sys-

tems are among the most popular, including Giesecke &

Devrient’s STARCOS 2.4 smart card Operating System

w ith On-card matcher [9]. The initi al FP-b ased MOSE

algor ithms, ﬁrst introduced in 1999, used proprietar y

templa tes [10, 11] for which much of the data was

preprocessed on the ho st to minimize the compu ta-

tional load on the secu re elem ent. Second-generatio n

algor ithms use a standards-based templa te format [7]

but, due to the adde d burden this puts on the secure

elem ent, accurac y tra deoffs have been necessar y to

achieve acceptable match speed. Current work under-

way by the aut hor promises a more comple x third-

generati on, standards-based minu tiae matc hing

algor ithm w ith mu ch smalle r accuracy degradati on

and suit abilit y for execu tion on most secure elem ents.

Of note is that pattern match ing techn iques for

ﬁnge rprints, w ith their relatively l arge templa te size

and image-processing-b ased algorithms , are harder

to imp lement and in many cases simply not practical,

leaving the aforementioned stat istical featu re-vector

approach as the best alte rnative to minu tia-based

metho ds. Unfor tunately, in additi on to accuracy

issues , proprietar y stat istical approaches suff er from

lack of a ny w id ely-accepted tem plate standards, drasti-

cally reducing their suit abilit y for the marketplace.

Recent Trends

The overall trend for biome trics – esp ecially

ﬁnge rprints – is rapidly grow ing in the cons umer mar-

ket, most notably in noteboo k compu ters, mob ile

phone s, door locks a nd USB storage tokens. Whi le

some of these applicati ons focus on the convenience

aspec t of bio metrics (e.g ., password replacement) al-

most all of them perform some sor t of asset protection

functio nalit y, w ith var y ing levels of secu rit y depending

on the des ign and applicati on.

Simil arly, MOSE system s for asset protection con-

tinue to gain in populari t y, and dep loyments are

expecte d to grow quite ra pidly in the nex t 2–5 years.

Mobile handset applica tions, which employ N FC

chips and/o r SI M cards, are driv ing the commercial

need . Atrua Technolo gies, Inc ., by which the aut hor is

employed, has demo nstrated the ﬁrst SIM card w ith

on-chi p ﬁngerprint matchin g in 2007 as w ell as the

ﬁrst NFC controller w ith on-chi p matchin g . It is

expecte d that variou s vendors w ill offer MOSE capabl e

system s in the near future and that mobile hand sets

w ith MOSE capabi lities w ill reach the mass market

worldw ide w ithin this timefr ame.

On the government side, the U.S. has recently put

together guidelines for t esting Personal Identit y Veriﬁca-

tion (P IV ) c ards [12 ] withﬁngerprint-basedmatch-on-

card capabilit y using FIPS 201 minutiae format stan-

dards. A s PIV cards have become the standard w ay o f

identifying U .S. Government employ ees, the added secu-

rity and con venienc e of match-on-card will see a wide

deployment if the planned technology trials are

successful.

Related Entries

▶ Template Security

▶ Biometrics and Security, Standardization

▶ Common Biometric Exchange Formats Framework,

Standardization

▶ Fingerprint Matching, Automatic

References

1. Schneier, B.: Applied cryptography, Wiley, Inc., New York (1996)

2. Maltoni, D., Maio D., Jain, A.K., Prabhakar, S.: Handbook of

ﬁngerprint recognition, Springer, New York (2003)

3. Russo, A.P.: ‘‘Fingerprint-based authentication and smart cards:

Issues and trends’’, E-Payments 2000 Conference. http://www.

epf.net/PrevMtngs/Sep00/Sep00Meeting.html. Accessed Oct 5,

2000

4. ISO/IEC 7816-4: ‘‘Information technology–Identiﬁcation cards–

Integrated circuit(s) cards with contacts, Part 4: Interindustry

commands for interchange,’’ International Standard (1995)

5. ANSI-INCITS 385-2004: American National Standard for

Information Technology – Face Recognition Format for Data

Interchange (2004)

Transportable Asset Protection

1347

6. ANSI-INCITS 398-2005: American National Standard for Infor-

mation Technology – Common Biometric Exchange Formats

Framework (CBEFF) (2004)

7. ISO/IEC 19794-2: Information Technology—Biometric Data

Interchange Formats—Part 2: Finger Minutiae Data (2005)

8. MacKenzie, I.S., Phan, R.C.-W.: The 8051 Microcontroller, Pren-

tice Hall, (2006)

9. Giesecke & Devrient: STARCOS

2.4 Card Operating

System, Bio Version http://www.gdai.com/portal/page-_pageid=

42,70526&_dad=portal&_schema=PORTAL.htm. Accessed Dec

30, 2005

10. Russo, A.P.: Fingerprint matching algorithm for low-cost 8-bit

smart cards. RSA Conference Europe 2000 (2000)

11. Russo, A.P. (inventor): Method and system for ﬁngerprint tem-

plate matching. U.S. Patent Number 6681034 (1999)

12. National Institute of Standards and Technology: FIPS Pub 201-1

Personal Identity Veriﬁcation (PIV) of Federal Employees

and Contractors. Federal Information Processing Standards

Publication (2006)

Tread Pattern

The outsole or undersi de of a shoe can sometimes have

a complex pattern of ridges and various shapes which

provide traction for the wearer. For example, athletic

shoes may have a tread pattern composed of many

slanted ridges placed close together so as to give an

improved grip to the running surface, while hiking

shoes may have a thick and high proﬁled wedge-

shaped tread for easy walking in soft terrain. This

complex pattern is commonly referred to as a tread

pattern.

▶ Footwear Recognition

Trusted Biometric System

Biometric system with template protection, compared

to traditional biometric system, paying more attention

to security and privacy issues of the users.

▶ User Interface, System Design

Trusted Traveler

▶ Registered Traveler

1348

Tread Pattern

UBM

▶ Universal Background Models

ULW

▶ Universal Latent Workstation

Unauthorized Data Collection

In principle, all data collected must be backed by the

law, whether state legislation or contractual obliga-

tions. A data item may well be necessary for a particu-

lar purpose, but this does not mean that the purpose

itself is authorized, or that the agent (human or ma-

chine) collecting the data is empowered to do so.

▶ Privacy Issues

Unauthorized Data Disclosure

From a privacy perspective, the end-user owns his or

her own personal data, even if the data resides in an

organization’s computer system. Therefore, permission

should be sought from the end-user before data is

disclosed or shared. In practice, different jurisdictions

have different notions of data ownership, which may

result in disclosure without the end-user’s explicit con-

sent or even knowledge.

▶ Privacy Issues

Unification Framework

A uniﬁcation framework includes a collection of fusion

algorithms and it uses the evidences obtained from the

input biometric probe data to dynamically select the

optimal fusion algorithm. The selected fusion algo-

rithm is then used to compute the fused biometric

information. The uniﬁ cation or reconciliation should

satisfy most of the application requirements and yield

better recognition performance.

▶ Fusion, Sensor-Level

Universal Background Models

DOUGLAS REYNOLDS

MIT Lincoln Laboratory, Lexington, MA, USA

Synonyms

General model; Person-independent model; UBM;

World model

Definition

A Universal Background Model (UBM) is a model use d

in a biometric veriﬁcation system to represent general,

person-independent feature characteristics to be com-

pared against a model of person-speciﬁc feature char-

acteristics when making an accept or reject decision.

For example, in a speaker veriﬁcation system, the

UBM is a speaker-independent Gaussian Mixture

Model (GMM) trained with speech samples from a

large set of speakers to represent general speech char-

acteristics. Using a speaker-speciﬁc GMM trained with

speech samples from a particular enrolled speaker, a

2009 Springer Science+Business Media, LLC