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

Подождите немного. Документ загружается.

2D ﬂat plane introduce distortions, noise and incon-

sistencies on the captured ﬁngerprint image. To over-

come these problems, a new approach to capture

ﬁngerprints has been proposed [6, 7], called touchless

▶ contactless ﬁngerprinting . Because of a lack of

contact between the ﬁnger and any rigid surface, the

skin does not deform during the capture and the re-

peatability of the measure is improved.

The approaches used to capture a ﬁngerprint

based on touchless technology can be grouped in two

main families:

▶ Reﬂection-based Touchless Finger

Imaging (RTFI) and

▶ Transmission-based Touchless

Finger Imaging (TTFI). Figure 3 highlights the two

approaches. In the RTFI approach, the light generated

by monochromatic light sources and reﬂected on the

ﬁnger skin is collected by the detector. In the TTFI

approach, the light penetrating the ﬁnger is collected

by the detector positioned in front of the ridge–valley

pattern.

Since both the light reﬂecting on or penetrating the

valleys and the light reﬂecting on or penetrating

the ridges are collected by the detector, the ﬁnal

image has a contrast lower than that in the traditio nal

FTIR technology. This has a huge impact on the min-

utiae extraction algorithm and thus, the advantage

of a lack of skin deformation is negatively compen-

sated by thi s low contrast. Moreover, the illumina-

tion not being perfectly perpendicular to the skin

surface, shadowing effects of the ridges on the valley

provide a wrong representation of small details (min-

utiae, pores, island, branches, etc.). Sophisticated

illumination techniques are required to avoid this rep-

resentation problem and increase the ﬁnal image con-

trast. The consequence is an increase of the size and

ﬁnal costs of these devices.

Another disadvanta ge of this technology is repre-

sented by the easy methods that can be used to attack

these devices, which cann ot be deﬁnitively used for

high-securi ty applications. In contrst to the FTIR

case, where the ridge–valley 3D structure is important

to generate an image, the touchless approach cannot

discriminate between a 2D and a 3D pattern. Hence,

presenting a photograph or a simple drawing of a

ﬁngerprint to the sensor, a new ﬁngerprint image sim-

ilar to the synthetic one is generated and the access is

granted. Finger positioning, sensor usability, and user

convenience must be still addressed.

Solid-State Sensors

The ﬁrst solid- state ﬁngerprint capture device appeared

on the market only in the middle of 1990s. It was a

CMOS sensor capable of measuring the electrical ca-

pacity between the ﬁnger skin and the sensing surface

(Fig. 4), which is composed by many squared pixels.

Each pixel and the corresponding skin portion can

be considered as an electrical capacitor with capacity

C ¼ eA ∕d, where A represents the pixel area and d the

distance between the skin and the pixel and e is the

permittivity (a constant depending on the material) of

the dielectric contained between the two capacitor

Fingerprint, Palmprint, Handprint and Soleprint Sensor. Figure 3 ▶ Touchless or contactless capture approach:

(a) reﬂection-based touchless ﬁnger imaging; (b) transmission-based touchless ﬁnger imaging.

540

Fingerprint, Palmprint, Handprint and Soleprint Sensor

plates. Each pixel produces a graylevel value propor-

tional to its distance from the skin.

Another approach used to capture the ridge–valley

pattern is based on the Radio Frequency (RF) electrical

ﬁeld (Fig. 5). A signal generator produces a low-level

RF ﬁeld traveling through the ﬁnger. The signal is then

collected by AC sensors after being attenuated by the

ﬁnger skin. The attenuation level of the signal is a

function of the ridges and the valleys; the sensor array

calculates the attenuation to synthesize the ﬁngerprint

structure. RF signal can be dynamically optimized in

frequency and level to obtain the best possible image.

Using pyroelectric materials, it is possible to mea-

sure the difference of temperature between ridges and

valleys. This approach is used while the ﬁnger is swiped

on the small sensor surface (Fig. 6). This type of

devices are called

▶ sweep sensors [8, 9]. The thermal

sensing elements detect temperature difference be-

tween valleys and ridges during the ﬁnger movement.

This technology is claimed to overcome the skin con-

dition issues of optical sensors. However, the resulting

images are not rich in gray level values, i.e., dynamic

range. Sweep sensors are ver y attractive because of

their small size and low cost. This makes easier their

integration in handheld and mobile devices.

The big advantage of the solid-state technology is

represented by their smaller dimensions and lower

costs with respect to the optical technology. Since

they can be manufactured very thin and their power

consumption needs are very low, solid-state sensors

can be mounted on cards, handheld devices or laptops

and used as logon means. This has an implication on

the range of applications in which solid-state ﬁnger-

print sensors can be involved with respect to the

Fingerprint, Palmprint, Handprint and Soleprint Sensor. Figure 4 Capacitive principle for the capture of the

ridge–valley structure.

Fingerprint, Palmprint, Handprint and Soleprint Sensor. Figure 5 Radio Frequency Field principle used to capture the

ridge–valley structure.

Fingerprint, Palmprint, Handprint and Soleprint Sensor

541

optical devices. However, external environmental fac-

tors (temperature, humidity, dust, etc.) are the major

drawbacks of this technology. The sensing area is a chip

completely open to the external world. Thus, special

electrostatic protection method s must be used to avoid

that external electrostatic charges destroy the chip sur-

face. The same human skin can be the cause of the

surface destruction, since the human body is usually

electrically charged. Dust is another common vehicle

of electrostatic charges that can quickly and easily

degrade the sensin g surface characteristics.

The use of solid-state sensor is mainly limited to

ﬁngerprint for their small sensing area. Palmprint and

soleprint would require very large silicon areas that

would make these sensors completely unaffordable in

term of costs.

Even if their introduction on the market has been

revolutionary and the expert envision new kinds of

applications for ﬁngerprint recognition (domotic,

health-care, id-card and credit-card protection, etc.)

their reduced lifetime and their high sensitivity to the

external environmental factors limit the wide-spreading

of these devices.

Ultrasonic Sensors

The ability to obtain images using ultrasound is based

upon the reﬂection and transmission coefﬁcients of

ultrasound as it propagates through media of varying

acoustic impedance. What makes sound waves valuable

for the imaging of the ridge–valley pattern is that they

can both reﬂect and pass through objects. The charac-

teristics of sound waves make it possible for high-fre-

quencies to pass through substances and accurately

measure the ridges and valleys of a ﬁngerprint even if

in presence of dirt, grease, ink, moisture, dye, or other

substances routinely found on ﬁngers.

The capture principle of a ultrasonic device is high-

lighted in Fig. 7. An ultrasonic wave generator pro-

duces high-frequency sound impulses. These imp ulses

reﬂect on each material found on their path producing

echos. The strength of each echo depends on the mate-

rial and the shape of the object on which they were

generated. Special receptors are used to translate the

echos in an electrical signal.

Livescan imaging the ﬁngerprints of children 5

years and younger is a technically challenging task,

since the ridge structure is usually very ﬁne and con-

tains high ‘‘spatial frequencies,’’ meaning that the

ridges very close together. The spatial frequency

of the ﬁngerprint directly determines the resolution

that the imaging device needs to accurately image the

ﬁnger. Most live-scan ﬁngerprint scanners have been

designed to image adult ﬁngers where a high-resolu-

tion scan is unnecessary. High-resolution ultrasonics is

the only technology that can reliably and repeatedly

capture clear and useful images of a young child’s

ﬁngerprint.

Next Generation

Although most of the technologies mentioned earlier

are quite new (some of them are still in the prototyping

phase), the research and the development continues to

Fingerprint, Palmprint, Handprint and Soleprint

Sensor. Figure 7 Capture principle of an ultrasonic

capture device.

Fingerprint, Palmprint, Handprint and Soleprint

Sensor. Figure 6 An example of sweep sensor.

542

Fingerprint, Palmprint, Handprint and Soleprint Sensor

bring new ideas to this ﬁeld. The study of the physiol-

ogy and the formation of the furrow pattern allowed

to propose new ﬁngerprint and palmprint capture

approaches. It is important to mention here the Opti-

cal Coherence Tomography (OCT) which is an inter-

ferometric, noninvasive, optical tomographic imaging

technique offering millimeter penetration (approxi-

mately 2–3 mm in tissue) with micrometer-scale axial

and lateral resolution. OCT is like an optical version of

ultrasound imaging. The technique is already routinely

used in medicine, but has not had a forensic applica-

tion until now. The technique provides a transparent

3D structural picture by sending light though the pat-

tern of natural secretions left on a surface by a ﬁnger

and combining the reﬂected beam w ith a ‘‘reference

beam’’ produced by bouncing light from a laser off a

mirror. This produces an interference pattern at

a photodetector the same as those found in a digital

camera which can then be used to reconstruct an image

of the original ﬁngerprint.

This tech nology together with multispectral and

touchless imag ing must be still further developed to

demonstrate their superiority with respect to the FTIR

approach that still remains the most used method to

capture ﬁngerprints and palmprints.

Summary

Livescan furrow sensors represent a family of devices

used to capture ﬁngerprints, palmprints, handprints,

and soleprints. The same anatomical characteristics of

the skin present on ﬁnger tips, palms, and soles allow

the use of the same tech nology for the capture of these

biometric treats.

The ink-on-paper method is ﬁrst method used to

capture ﬁngerprints and palmprints. Optical devices

try to overcome the inconvenience of the ink on the

skin and provide a good alternative method to the

legacy ink-on-paper. Solid-state sensors are very at-

tractive for their very small size and reduced costs,

but they can only be used to capture ﬁngerprints.

Moreover, environmental factors limit the life time of

these devices.

Multispectral and touchless imaging technologies try

to overcome the limitation of the optical devices, but

their relatively higher costs and very low interoperability

with legacy technology limit their wide-spreading.

Related Entries

▶ Biometric Recognition

▶ Biometric Sensor and Device, Overview

▶ Fingerprint Recognition, Overv iew

▶ Fingerprint Veriﬁcation

References

1. Ashbaugh, D.R.: Quantitative-Qualitative Friction Ridge Analy-

sis: An Introduction to Basic and Advanced Ridgeology. CRC,

Boca Raton (1999)

2. Criminal Justice Information Services Division: Electronic Finger-

print Transmission Speciﬁcation. Department of Justice (1999)

3. Rowe, R.K., Nixon, K.A.: Fingerprint enhancement using a mul-

tispectral sensor. In: Proceedings of SPIE Conference on Biomet-

ric Technology for Human Identiﬁcation, vol. 5779. pp. 81–93.

Orlando, USA (2005)

4. Rowe, R.K., Corcoran, S.P., Nixon, K.A., Ostrom, R.E.: Multi-

spectral Imaging for Biometrics. In: Proceedings of SPIE Con-

ference on Spectral Imaging: Instrumentation, Applications, and

Analysis, vol. 5694, pp. 90–99. Orlando, USA (2005)

5. Rowe R.K., Nixon, K.A., Butler, P.W.: Advances in Biometrics.

Sensors, Algorithms and Systems, chap. Multispectral Finger-

print Image Acquisition. Springer, Berlin (2008)

6. Parziale, G.: Advances in Biometrics. Sensors, Algorithms and

Systems, chap. Touchless Fingerprint Technology. Springer,

Berlin, New York, USA (2008)

7. Parziale, G., Diaz-Santana, E.: The surround imager: a multi-

camera touchless devide to acquire 3D rolled-equivalent ﬁnger-

prints. In: Proceedings of IAPR International Conference on

Biometrics (ICB), pp. 244–250. Hong Kong, China (2006)

8. Parziale, G., Bishof, H.: Image reconstruction and on-the-ﬂy

minutiae extraction of ﬁngerprints acquired with sweep sensors.

In: Proc. of 28th Workshop of the Austrian Association of

Pattern Recognition, pp. 241–248. Austria (2004)

9. Clausen, S.: Advances in Biometrics Sensors, Algorithms and

Systems, chap. A single-line AC capacitive ﬁngerprint swipe

sensor. Springer, New York, USA (2008)

Fingerprints Hashing

JEAN-FRANC¸OIS MAINGUET

Grenoble, France

Synonyms

Biometric encryption; Cancelable biometrics; Finger-

print encryption; Fuzzy extractor; Fuzzy vault; Intri-

cated biometrics

Fingerprints Hashing

543

Definition

Fingerprint hashing is merging ﬁngerprint recognition

and cryptographic methods. The aim is to perform a

recognition using ﬁngerprint while, at the same time,

hiding the private information related to the ﬁnger-

print, thus enabling public ﬁngerprint templates.

Introduction

Keeping a database in a safe place is not easy. Even with

good encryption methods and special care, databases

containing sensitive information, such as bank account

numbers, are vulnerable to being compromised.

Nobody wants something like that to happen when

dealing with ﬁngerprint identiﬁcation.

Security of a ﬁngerprint-based system can be

divided into two main areas:

1. The electronic security, which poses the question:

‘‘Is the electronic system, at the other end of the

wires, a real trustful authorized ﬁngerprint system ?’’

2. The liveness securit y, which asks a different ques-

tion: ‘‘Is the object touching the sensor a real ﬁnger,

alive and connected to a living person?’’

Liveness security is not addressed in this essay.

Fingerprint hashing is part of the electronic security

solution and deals with encryption.

Any bio metric system requires the storage of

a template (or reference). For ﬁngerprint systems, the

most common method consists in storing the minu-

tiae. This information is considered as private infor-

mation and should be protected and ciph ered, not only

for privacy reasons, but also against template replace-

ment. This prevents a hacker from replacing the

owner’s minutiae, or using reverse engineering to get

the minutiae locations and create a fake ﬁngerprint.

Even if biometric data cannot be considered as secret

(they are public information in its cryptographic sense,

always hiding ones face or voice is impossible), it is

important to protect biometric data and the additional

data that goes with them (name, bank account or

whatever).

The template storage problem is generally solved

using encryption. A template database can be created

and protected using a single key pair (Fig . 1). Everything

is ﬁne up to the date when the key is compromised or

badly protected, allowing hackers to access the data.

However, in the case of ﬁngerprints, this is ones very

private information that is stored, and so it seems

desirable to have an even better security scheme.

Fingerprints Hashing. Figure 1 Protecting fingerprint templates: Key is not protected.

544

Fingerprints Hashing

Another problem occurs when one tryies to create a

system enabling encryption/decryption features (like

▶ Pretty Good Privacy (PGP)). One needs to protect

the access to the encryption/decryption module, which is

done using a passwor d . If one tries to replace the pass-

word with a ﬁngerprint (Fig. 2), then one faces the

problem of protecting the template, and cannot use

the encryption/decryption scheme, because it is not yet

enabled! It is the same problem as ‘‘you cannot put the key

of the safe inside the safe itself.’ ’ One needs another safe.

There is also an additional problem from a secur ity

point of view. The result of matching is only one bit of

information that is easy to ﬁnd and hack (too low

entropy). It would be better to eliminate this weakness.

Desirable Features, Definitions

A better ﬁnge rprint system includes:

1. The storage of the template (minutiae) in a nonre-

versible way. It is still possible to perform a match,

but it is impossible to recover the original minutiae

and impossible to derive the secret key.

2. It is possible to revoke (cancel) a template. If a

template is not to be used anymore, it is possible

to forbid its use and create a new one.

3. There is no step with a single yes/no bit

corresponding to the match/no match result.

Properties #1 and #2 are generally linked, because it

would be very impractical and dangerous to use a

transform that is unique. Ea ch individual would have

a unique number ID for his or her whole life, impossi-

ble to change.

Fingerprint hashing is the use of a non-reversible

transform (similar to a hash function) over a ﬁnger-

print. It is also called ‘‘cancellable biometrics,’’ because

it is possible to cancel or to revoke the template.

Fingerprint hashing involves using some kind of cryp-

tographic scheme, similar to a hash function, but it is

not a hash function.

Property #3 requires a stronger merge between

biometrics and cry ptography. Having all the properties

at the same time is pretty hard to achieve and to prove,

but has been originally proposed under the name of

‘‘Biometric Encryption’’ [1, 2 ]. Unfortunately, ‘‘bio-

metric encryption’’ can be a simple combination of a

biometric template and a simple encryption scheme.

But it is much more; it is a real merge. In quantum

cryptography, the word ‘‘

▶ intricated’’ is used to des-

ignate the non-separable nature of some properties in

quantum mechanics, and so ‘‘Intricated Biometrics’’

seems a better designation.

Fingerprints Hashing. Figure 2 Protecting access to cryptographic secret keys using a fingerprint system: Fingerprint

template is not protected.

Fingerprints Hashing

545

Cancelable Biometrics

Fingerprint hashing seems pretty close to password

protection. A password is protected using a hash func-

tion, which is basically a method to transform some

data into a relatively small number, the hash value,

sometimes called ﬁngerprint (which causes confu-

sion), because of its uniqueness property (no collision

should occur). A hash function is not reversible, and in

most cases, some original data is lost as the result-

ing hash value is much shorter. This works well for

password storage. You just need to apply the same hash

function to the proposed password and perform a bit-

to-bit comparison for checking. It is not useful to

regain access to the original password.

Unfor tunately, this scheme cannot apply to a ﬁn-

gerprint, because you never enter exactly the same

ﬁngerprint image. Every acquisition is different, and

usual hash functions will return a different value,

forbidding a further comparison. The problem is

much more complex as it needs to be accepted that

there will be some variabili ty of the data. Fingerprint

hashing must use a non-reversible transform like a

hash function, but the comparison stops here. The

other properties of a hash function, such as ﬁxed

length and uniqueness, are not required, but there

needs to be a comparison, a match at the end, as

depicted in Fig. 3.

General concepts related to cancelable biometrics

have been discussed by Ratha et al. [3]. Davida et al. [ 4]

added data to create a non-reversible template. Lin-

nartz and Tuyls [5] proposed the use of speciﬁc shield

functions before a hash function.

Lumini and Nanni [6] proposed the BioHash, a

combination of a hash code, a Gram–Schmidt normal-

ization and using a Hamming distance for comparison.

This has been tested using the FVC-2002 ﬁngerprint

database.

Boult et al. [7] proposed another scheme called

BioToken, and also tested on the FVC-2002 and 2004

ﬁngerprint datab ases, which showed some enhance-

ments of the accuracy of the system.

As usual in cryptography, proving that the trans-

form is non-reversible or reversible with an extremely

long computation time is very hard to achieve.

Although there are some good reasons to believe that

some solutions exhibit the right properties, nothing is

mathematically proven yet. It took a long time for the

security of regular cryptographic schemes to be accepted,

and biometrics is in a similar situation, still in its infancy .

Intricated Biometrics

Cancelable biometrics shows interesting features,

but still shows the potential weakness of the

Fingerprints Hashing. Figure 3 Fingerprints hashing/cancellable biometrics: It is not possible to extract the original

biometric data from the template as the transform is non-reversible.

546

Fingerprints Hashing

match/no-match bit. This type of weakness may not be

as critical when speaking of a physical or logical access,

because of the need of a go/no go answe r. But in most

cases, the aim of the biometrically enabled system is to

provide a service, and secure systems always use some-

where a cryptographic key when a transmission is

involved in a non-secured environment.

A simple scheme use s the result of the match to

enable the decipher key, as shown in Fig. 4. The

template and the key are not protected. This requires

external means; another secret key and method. There

is still the one bit matc h/no match result. Intricated

biometrics proposes to merge the decipher key with

the template, so that neither the biometric template

(the minutiae) nor the decipher key can be obtained

from the stored template alone; they are intricated

(Fig. 5). The intricated biometric template can be

stored anywhere, even in a non-secured area.

Fingerprints Hashing. Figure 4 A simple use of biometrics to use a secret key: fingerprint and secret key are not yet

protected, and the 1-bit match/no match still exists.

Fingerprints Hashing. Figure 5 Intricated biometrics: It is not possible to get the secret key and the original

fingerprint data from the intricated fingerprint template, and the extracted key appears only for a short while for

deciphering. The 1-bit match/no match step is eliminated.

Fingerprints Hashing

547

When the decipher key is to be used, t he live

ﬁngerprint can be scanned. If the extracted minutiae

corresponds to the stored template, then the rig ht

decipher key will be regenerated, immediately used

to decipher the message (this is the service) and

destroyed (the key never leaves th e secure area). If

the extracted minutiae are not the genu ine min utiae,

a key is still generated but not the correct key. The

message is then incorrectly deciphered, giv ing a

meaningless result. At the end, there is no informa-

tion revealed, which is a very good p roperty o f

a secure system, and it is not p ossible to apply a

scheme such as hill-climbing, based on access of the

matching score.

It is possible to reach these objectives, but it is hard

to achieve and to prove, especially for ﬁngerprints.

Cryptography is Accurate; Biometrics is

Fuzzy

In cancelable biometrics, a function similar to a

hash function had to be applied, but the data varia-

bility was a problem. Intricated biometric involves

re-generating a cryptographic key and the same prob-

lem. Every bit must be correct; no error is allowed. With

biometrics, there is always some uncertainty. Each time

a ﬁngerprint is scanned or applied to a sensor, it may

not be exactly the same area. The person may have a new

cut or scar; the ﬁnger could be dirty, wet, or dry.

A par tial solution would be to extract a stable

sequence from a ﬁngerprint image, always the same,

and then combine it with a cryptographic key. This is

like extracting a stable signal from a noisy, fuzzy envi-

ronment. Some research proposed the use of error-

correcting code, with the so-called ‘‘fuzzy extractor’’

[8, 9] that can be applied to different biometric mod-

alities, and then speciﬁcally over ﬁngerprint databases

[10, 11].

The ‘‘

▶ fuzzy vault’’ was proposed in 2002 by Juels

et al. [12]. The proposal involved secret being merged

with biometric data such as minutiae that does not

need to be in a speciﬁc order.

▶ Chaff points [ 13] are

added to hide the genuine minutiae. The experiment

was later enhanced using lattice [ 14], tested on the

FVC-2002 database and enhanced with helper data by

Uludag et al. [15, 16].

Soutar et al. [2] proposed in 1999 using ﬁlters to

extract stable characteristics of the ﬁngerprint and then

merged them with a secret.

One example scheme is:



Enroll



A set of M minutiae is extracted from a

ﬁngerprint.



A secret key is divided into M pieces of data;

each piece is linked to one minutiae.



Random chaff po ints are added, corresponding

to non-existing minutiae and wrong pieces of

secret key.



Recognition



A live set of N minutia e are extracted from a live

ﬁngerprint.



The matching minutiae enable extraction of the

correct piece of the secret key.

As the live minutiae may not be exactly the same, it

is important to introduce some kind of redundancy for

the secret key. A subset of the M enrolled minutiae is

needed to perform a match. Lagrange interpolation has

been proposed to recover the full secret key, with the

advantage of not depending on the order of the points

or minutiae.

But some problems arise:



Brute force attack: It is important to add enough

chaff points to hide the genuine points and to

create too many possible combinations for a brute

force attack to succeed.



Generating chaff points is not a simple operation,

becausecaremustbetakentoavoidﬂaws.The

chaff points must be indistinguishable from gen-

uine points. It is a similar problem to random

number generators, wher e it is difﬁcult to pro ve

that they are really random. Alwa ys using the same

chaff points would make it too easy to ﬁnd them.



Matching minutiae for key extractio n will likely

require more computation.



Chaff points may lead to wrong alignments,

especially with poor ﬁngerprints, making min-

utiae matching less robust.



Intricated ﬁngerprint template requires more

memory space than a simple template (but

another key and program would be needed for

protection).

Conclusion

Fingerprint hashing (intricated biometrics) seems to

be the ultimate protection scheme. This is not a proven

technolog y yet, but achieving the objectives would lead

548

Fingerprints Hashing

to a better protection of privacy without worrying

about databases.

Related Entries

▶ Encryption, Biometric

▶ Fake Finger Detection

▶ Fingerprint Features

▶ Fingerprint Matching, Automatic

▶ Fingerprint Templates

References

1. Cavoukian, A., Stoianov, A.: Biometric encryption: A positive-

sum technology that achieves strong authentication, security

and privacy. White paper, Information and privacy commiss-

ioner of Ontario, March (2007)

2. Soutar, C., Roberge, D., Stoianov, A., Gilroy, R., Vijaya Kumar,

B.V.K.: Biometric Encryption, chap. 22, McGraw-Hill (1999)

3. Ratha, N.K., Connell, J.H., Bolle, R.M.: Enhancing security and

privacy in biometrics-based authentication systems. IBM Syst. J.

40(3), 614–634 (2001)

4. Davida, G.I., Frankel, Y., Matt, B.J., Peralta, R.: On the relation of

error correction and cryptography to an off-line biometric based

identiﬁcation scheme. In: Proceedings of the Workshop on Cod-

ing and Cryptography, Paris, France, pp. 129–138 (1999)

5. Linnartz, J.P., Tuyls, P.: New shielding functions to

enhance privacy and prevent misuse of biometric templates.

In: Proceedings of the Fourth International Conference on

Audio and Video based Biometric Person Authentication, Guild-

ford, UK, pp. 393–402 (2003)

6. Lumini, A., Nanni, L.: An improved biohashing for human

authentication. Pattern Recognit. 40, 1057–4065 (2007)

7. Boult, T.E., Scheirer, W.J., Woodworth, R.: Revocable Fingerprint

Biotokens: Accuracy and Security Analysis. In: Proceedings of

the IEEE Conference on Computer Vision and Pattern Recogni-

tion (CVPR’07), Minneapolis, USA, pp. 1–8, 17–22 June (2007)

8. Dodis, Y., Reyzin, L., Smith, A.: Fuzzy extractors: How to gener-

ate strong keys from biometrics and other noisy data. In:

Proceedings of the Eurocrypt 2004, pp. 523–540 (2004)

9. Burnett, A., Byrne, F., Dowling, T., Dury, A.: A biometric identity

based signature scheme. In: Proceedings of the Applied Cryptog-

raphy and Network Security Conference, New York, USA (2005)

10. Costanzo, C.R.: Biometric cr yptography: Key generation using

feature and parametric aggregation. Online techreport, School

of Engineering and Applied Sciences, Department of Computer

Science, The George Washington University, October (2004)

11. Al-Tarawneh, M.S., Khor, L.C., Woo, W.L., Dlay, S.S.: Crypto key

generation using contour graph algorithm. In: Proceedings of

the 24th IASTED International Multi-Conference Signal Proces-

sing, Pattern Recognition and Applications, Insbruck, Austria,

February (2005)

12. Juels, A., Sudan, M.: A fuzzy vault scheme. In: Lapidoth, A.,

Teletar, E. (eds.) Proceedings of the IEEE International Sympo-

sium on Information Theory, p. 408. IEEE Press (2002)

13. Chang, E.-C., Li, Q.: Hiding secret points amidst Chaff. In:

Proceedings of the Eurocrypt, Saint Petersburg, Russia (2006)

14. Zheng, G., Li, W., Zhan, C.: Cryptographic key generation from

biometric data using lattice mapping. In: Proceedings of the 18th

International Conference on Pattern Recognition (ICPR’06),

Washington, DC, USA, pp. 513–516. IEEE Computer Society

(2006)

15. Uludag, U., Jain, A.K.: Fuzzy ﬁngerprint vault. In: Proceedings

on Workshop: Biometrics: Challenges Arising from Theory to

Practice, August 2004, pp. 13–16 (2004)

16. Uludag, U., Jain, A.: Securing ﬁngerprint template: Fuzzy vault

with helper data. In: Proceedings of the 2006 Conference on

Computer Vision and Pattern Recognition Workshop, June

2006, pp. 163–170 (2006)

First Level Detail

This reﬂects the general ﬂow of the papillary ridges

which may form certain patterns such as arches, loops,

whorls, and deltas.

▶ Fingerprint Matching, Manual

Fisher Criterion

Fisher criterion is a discriminant criterion function

that was ﬁrst presented by Fisher in 1936. It is deﬁned

by the ratio of the between-class scatter to the within-

class scatter. By maximizing this criterion, one can

obtain an optimal discriminant projection axis. After

the sample being projected on to this projection axis,

the within-class scatter is minimized and the bet ween-

class scatter is maximized.

▶ Non-linear Techniques for Dimension Reduction

Fixed Pattern Noise

It is characterized by the same pattern of ‘‘hot’’ pixels

occurring with images taken under the same condi-

tions of temperature and exposure.

▶ Face Device

Fixed Pattern Noise

549