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

Подождите немного. Документ загружается.

Zero Effort Forgery

STEPHEN J. ELLIOTT

Purdue University, West Lafayette, IN, USA

Synonym

Zero effort impostor attempt

Definition

An impostor attempt is classed as ‘‘zero-effort’’ if the

individual submits his/her own biometric feature as if

he/she were attempting successful veriﬁcation against

his/her own template, but the comparison is made

against the template of another user. In the case of

dynamic signature veriﬁcation, an impostor would

therefore sign his/her own signature in a zero-effort

attempt. In such cases where impostors may easily

imitate aspects of the required biometric, a second

impostor measure based on ‘‘active impostor attempts’’

may be required [1]. The active impostor attempt

is the one in which an individual tries to match the

stored template of a different individual by present-

ing a simulated or reproduced biometric sample or

by intentionally modifying his/her own biometric

characteristics [1].

Introduction

The concept of zero-effort forgery centers around

an individual’s capability to submit his/her own bio-

metric feature as if he/she were attempting a successful

veriﬁcation against his/her own template. Zero-effort

is a typical methodology of extracting perfor-

mance rates (false accept, false reject) of a biometric

system. However, in the realm of dynamic signature

veriﬁcation, the concept of the impostor distri-

bution, hence zero effort, is often overlooked. This

deﬁnition is broken into two parts:an overview of

dynamic signature veriﬁcation (to explain the con-

cept of the modality that has the exception for zero

effort) and a discus sion of the importance of the

impostor dataset (as it relates to dynamic signature

veriﬁcation and the concept of zero-effort as explained

earlier).

Overview of Dynamic Signature

Verification

Biometrics can be divided into two main categories:

behavioral andphysiological. Fingerprint and iris are

examples of physiological biometrics, whereas voice

and signature are examples of behavioral biometrics.

Signature veriﬁcation aims at authenticating an indi-

vidual, bas ed on their signing characteristics. Appli-

cations such as document authenticity, ﬁnancial

transactions, and paper-based transactions have all,

at one time, used the signature to convey the intent

to complete a transaction [ 2,3]. Signature veriﬁcation

can be divided into two classes, depending on how the

signature data are collected and analyzed. The ﬁrst is

the digitized signature. Here, the input is a static

image, extracted from a document, check, or receipt

that is veriﬁed after being scanned. The second type is a

dynamic signature, where the user signs on a digitizer,

and the signature, along with traits or characteristics, is

collected in real time.

Dynamic signature veriﬁcation has a number of

statistical features that can be derived from the basic

set of data that a digitizer provides, and these vary

signiﬁcantly across algorithms. The typical consensus

is that these input variables are gathered from a dig-

itizer and include x, y (Cartesian co-ordinates), p (pres-

sure or force), and t (time) [4]. These input variables

are then used to create the global and local features

2009 Springer Science+Business Media, LLC

described in various accounts in the literature [4–6].

Fairhurst and Ng [7] outlined 61 features [8], Nalwa [9],

observed that the temporal ‘‘characteristics of the pro-

duction of an online signature are the key to the

signature’s veriﬁcation’’ (p. 5). Typical signature func-

tions include pressure vs. time, horizontal and vertical

components of position, velocity, acceleration, and

force, all against time. Another way of characterizing

a signature is through the analysis of the ‘‘stroke’’ that

is, the pen down–pen up movement of the pen on the

digitizer. All these various dynamic traits that are col-

lected during the act of signing are said to make an

impostor signature easier to detect.

Forgery Levels – The Impostor

Data Set

One of the interesting characteristics of dynamic sig-

nature veriﬁcation is the concept of an impostor

attempt. An impostor attempt is deﬁned as a ‘‘zero-

effort’’ if the individual submits his/her own biomet-

ric feature as if he/she were attempting successful

veriﬁcation against his/her own templa te, but the

comparison is made against the template of another

user [1 ]. This makes sense for an opportunistic forger

for almost all biometrics, barring dynamic signature

veriﬁcation. In the case of dynamic signature veriﬁca-

tion, ‘‘B’’ would sign ‘‘B’’’s name trying to get into

‘‘A’’’s account. Thus, zero-effort forgery, on the part of

the imposter, makes no sense. ‘‘B’’ would not have

any information about the signature that ‘‘A’’ pro-

vides. Therefore, in order to examine the FAR and

FRR of a dynamic signature veriﬁcation system, an

examination of the impostor dataset construction is

of great importance. In the literature, there are several

different examples of the impostor dataset. Biometric

testing and evaluation typically create a probability of

an impostor signature getting accepted as a genuine

signature. These outcomes are generally denoted by a

False Accept, or a False Reject under conditions of

zero effort. Such an attempt is deﬁned as ‘‘where an

impostor uses his or her own biometric sample and

claims the identity of a different enrollee’’ [10].

The determination of the forger in dynamic signa-

ture veriﬁcation is somewhat a challenge, as a zero

effort forgery would require the impostor to sign his

or her own name while claiming the identity of a

different enrollee. Therefore, understanding how to

test and evaluate the signature in a forgery setting is

an interesting research question. In the literature,

there is no real consensus on how be st to challenge

a signature. Impostor datasets are created in numerous

ways and have the effect of changing the respective

performances of algorithms. This change can be done

through a different generation of the impostor signa-

tures. A review of the literature shows various perfor-

mance results from several studies, all of which have

different methodologies for collecting impostor signa-

ture datasets. The various studies and their respective

error rates (false accept, false reject, and the equal error

rate where appropriate) have been outlined earlier

[11]. The variances in error rates shown (0% to 50%

false accept rate and 0% to 20% false reject rate) can be

explained by a number of factors, one of which has to

do with how an imposter signature dataset is created.

Komiya and Matsumoto’s [12] study is particularly

interesting. This study had a database consisting of

293 genuine sign atures and 540 forgery signatures

from eight individuals [13] study dataset was com-

prised of 496 original signatures from 27 people.

Each person signed 11 to 20 times. The database

contained 48 forgeries that ‘‘fulﬁll the requirement on

the visual agreement and the dynamic similarity with

the original signature’’ (p. 5). [14] trained the algo-

rithm using 250 signatures per writer; of these 250

signatures, 100 were authentic signatures and 150

were random forgeries, classiﬁed as the genuine signa-

tures of other writers. [15] used 27 people in their

study, with the participants writing their own signa-

ture. The study also used 4 people who imitated the

signatures of these 27 people. Unfortunately, no fur-

ther information is provided on the selection of the

impostor or on what knowledge the forgers possessed

in order to forge the signatures. [2] used genuine

signatures from other individuals as forgeries. In addi-

tion, a group of synthesized signatures was created by

distorting real signatures through the addition of low

level noise and dilation or erosion of the various struc-

tures of the signature. [16] motivated the forgers by

offering a cash reward. [17] examined people’s signa-

tures over a four-month period to assess variability

over time. In the Signature Veriﬁcation Competition,

genuine signers created signatures other than their

own [18]. In [19], the authors deﬁned three different

levels of forgeries: the simple, statically skilled, and

1412

Zero Effort Forgery

timed (p. 643). [3] used signatures that ‘‘on casual

visual inspection would pass as authentic’’ (p. 201).

[20] provides three characteristics of forgery: the

random forgery deﬁned as one that belongs to a differ-

ent writer of the signature model; simple forgery

represented by a similar shape consistency with the

genuine signers shape; and the skilled forgery (p. 2).

[8] described two classes, the skilled forgery and

the random or zero effort forgery. For the skilled

forger, it is ‘‘signed by a person who has access to a

genuine signature for practice’’. In addition, it is

important to understand the handedness of the

genuine signer, and sometimes, the gender. Again,

the literature sometimes provides insights into the

impostor, and to some extent, the genuine data-

sets, all to varying degrees. For example, Lee, Berger,

Aviczer [19] collected 5,603 genuine signatures from

105 people, including 22 women and 5 individuals

who were left handed.

Therefore, when assessing the impostor datasets for

dynamic signature veriﬁcation, there are probably

many different levels from ‘‘zero-effort’’ at the lowest

level through to trained and skilled forger. This all

depends on the level of information given to the im-

postor. For example, level 0 would be the ‘‘zero-effort’’

where ‘‘B’’ has no relevant knowledge of ‘‘A’’. The

knowledge is zero; there is no knowledge of name:

verbal, printed, or signed. The next level is that ‘‘B’’

knows ‘‘A’’’s name – this information is revealed ver-

bally. Therefore, the forger might mistake ‘‘Stephen’’

for ‘‘Steven’’ or ‘‘Steve,’’ depending on the formality or

ceremony of the signature. Level 2 would be that ‘‘B’’

has seen a static image of ‘‘A’’’s signature prior to the

impostor attempt. Level 3 is that ‘‘B’’ can see a static

image of ‘‘A’’’s signature at the time of signing. Level 4

is that ‘‘B’’ is a ble to trace a sample of ‘‘A’’’s signature.

Level 5 is where ‘‘B’’ has recently witnessed a sample of

‘‘A’’’s signing, and level 6 is where ‘‘B’’ has repeatedly

witnessed ‘‘A’’ signing. Even this methodology takes

into consideration neither the Perceived Signature

Strength (PSS) of the original signature nor the moti-

vation of the impostor. The PSS is a concept that

indicates that an ‘‘opportunistic forger’’ will not forge

a signature that is ‘‘hard’’ to forge, as their success at

the point-of-sale may be not as high as an ‘‘easy’’

signature. This is more of an ‘‘opportunistic’’ forgery

as opposed to a more sophisticated attack on the

signature as outlined in previous research. For the

case of this study, an opportunistic forger is analogous

to an ‘‘opportunistic thief ’’ – working on their own

without any equipment [16].

References

1. Mansﬁeld, A and National Physical Laboratory (Great Britain):

Best practices in testing and reporting performance of biometric

devices (2002)

2. Hamilton, D., et al.: Low cost dynamic signature veriﬁcation

system. European Convention on Security and Detection,

pp. 202–206 (1995)

3. Nelson, W., Kishon, E., Use of dynamic features for signature

veriﬁcation. Decision Aiding for Complex Systems, Conference

Proceedings, 1991 IEEE International Conference on Systems,

Man and Cybernetics, vol.1, pp. 201–205 (1991)

4. Vielhauer, C., Steinmetz, R., Mayerhofer, A.: Biometric hash

based on statistical features of online signatures. Proceedings

of the 16th International Conference on Pattern Recognition,

Quebec City, Canda, vol. 1, pp. 123–126 (2002)

5. Leclerc, F., Automatic signature veriﬁcation: the state of the

Art–1989–1993. Int. J. Pattern Recog. Artif. Intell. 8, 643–660

6. Elliott S., Hunt, A.: Dynamic Signature Forgery Assessment and

Signature Strength Perception, Aeros. Electron. Syst. Mag. 23,

13–18 (2008)

7. Fairhust, M., Ng, S.: Management of access through biometric

control: A case study based on automatic signature veriﬁcation,

Universal Access in the Information Society, vol. 1, June 2001,

pp. 31–39

8. Kholmatov, A., Yanikoglu, B.: Biometric authentication using

online signatures. Computer and Information Sciences (ISCIS)

2004, pp. 373–380; http://www.springerlink.com/content/

a7q4keh1dv8d1u5f

9. Nalwa, V.: Automatic on-line signature veriﬁcation, Computer

Vision – ACCV’98, 1997, pp. 10–15; http://dx.doi.org/10.1007/

3–540–63930–6_98

10. Ohishi, T.: et al.: Pen-input on-line signature veriﬁcation

with position, pressure, inclination trajectories, Parallel and

Distributed Processing Symposium, Proceedings 15th Interna-

tional, 2001, pp. 1757–1763

11. Elliott, S., Hunt, A.: An assessement of dynamic signature forg-

ery and perception of signature strength. Procedding of the 40th

Annual IEEE International Carnahan Conference on Security

Technolgy, pp. 186–190 (2006)

12. Komiya Y., Matsumoto, T.: On-line pen input signature veriﬁca-

tion PPI (pen-Position/pen-Pressure/pen-Inclination), IEEE

SMC ‘99 Conference Proceedings. IEEE International Confer-

ence on Systems, Man, and Cyberbetics, vol. 4. pp. 41–46 (1999)

13. Schmidt, C., Kraiss, K.: Establishment of personalized templates

for automatic signature veriﬁcation. Proceedings of the Fourth

International Conference on Document Analysis and Recogni-

tion, vol. 1, pp. 263–267 (1997)

14. Lee, W.-S., Mohankrishnan, N., Paulik, M.: Improved Segmen-

tation through dyanmic time warping for signature veriﬁcation

Zero Effort Forgery

1413

using a neural network classiﬁer. ICIP 98. Proceeding of

the 1998 International Conference on Image Processing, vol. 2,

pp. 929–933 (1998)

15. Wu, Q,-Z., Jou, I-C Lee, S,-Y.: On-line signature veriﬁcation

using LPC cepstrum and neural networks. IEEE Trans. Syst.

Man Cybern., Part B, 27, pp. 148–153 (1997)

16. Mingming, M., Wijesona, W.: Automatic on-line signature

veriﬁcation based on multiple models., 2000. Proceedings

of the IEEE/IAFE/INFORMS 2000 Conference on Computation-

al Intelligence for Financial Engineering (CIFFr), pp. 30–33 (2000)

17. Han C.-C., e t al.: An on-line signature veriﬁcation sys-

tem using multi-template matching approaches Proceedings

of the IEEE 33rd Annual 1999 International Carnahan -

Conference on S ecurit y Technology, pp. 4 77–480 ( 1999)

18. Yeung, D., et al.: SVC2004: First International Signature Veriﬁ-

cation Competition In LNCS, Vol. 3072, pp. 16–22, Springer

(2004)

19. Lee, L., Berger, T., Aviczer, E.: Reliable online human signature

veriﬁcation systems. IEEE Trans. Syst. Man Cyber. 18, 643–647

(1996)

20. Justino, E., Bortolozzi, F., Sabourin, R.: The interpersonal and

intrapersonal variability inﬂuences on off-line signature veriﬁ-

cation using HMM. Proceedings of the XV Brazilian

Symposium on Computer Graphics and Image Processing,

pp. 197–202 (2002)

Zero Effort Impostor Attempt

▶ Zero Effort Forgery

Zone

The print Roman characters usually occupy three

zones: the upper, the middle and the bottom zones.

For instance, letters ‘‘a,’’ ‘‘c,’’ ‘‘e,’’ ‘‘m,’’ etc., appear in

the middle zone; ascendant letters ‘‘b,’’ ‘‘d,’’ ‘‘h,’’ etc.,

occupy both the upper zone and the middle zone;

descendent letters ‘‘g,’’ ‘‘p,’’ etc., occupy both the mid-

dle zone and the bottom zone. There are some letters

that may cover all the three zones such as ‘‘f’’ and ‘‘j.’’

▶ Signature Sample Synthesis

1414

Zero Effort Impostor Attempt