dsv — questions remain…

1
Dynamic signature verification (DSV) is the capture and verification of a signature to confirm the identity of a user. The technology offers a number of advantages over other biometrics. People are familiar with writing their signatures to authorise transactions, and do not see it as a threatening technology. It can also be very inexpensive to implement. The primary reason that interest in DSV is growing is not because of algorithm improvements, but because it is becoming increasingly easy and possible to capture signatures, not only from special tablets or pens, but also from PDAs, computer displays with touch screens and the latest laptops. This is significant as it can make it possible for DSV to be implemented as a software only solution. Use of such devices for electronic capture of signatures is now becoming widespread in applications such as signing to acknowledge receipt of a parcel delivery. (Note, however, that simple signature capture systems do not record or compare dynamic features of signatures.) The significant difficulties and frustra- tion which users have with passwords and PINs today, combined with the growth of e- and m-commerce, remote working, remote banking, together with the growth of digital connectivity through broadband and wireless networks is creating an ever-greater need for innovative solutions for authentication of an individual. To date there have not been major trials of DSV systems that test performance amongst different sections of population, age groups, languages, etc. With many other biometric methods such as fingerprint, facial or iris recognition, large databases of sample images have been constructed enabling comparison algorithms to be tested thoroughly offline. Therefore, an independent dynamic signature technology benchmark similar to FVC 1 for fingerprint is still awaited. 2 Some questions remain open. How unique are people’s signatures? What is the chance that people with the same name might have similar signatures? Does this put people with common names at a disadvantage compared to those with unusual names? For example, in one major UK enterprise alone, there are currently 285 people called J. Smith. Current DSV technologies can have problems with people who have either very long or very short signatures. With short signatures, there is insufficient information to be confident in the uniqueness of the signature. However, long signatures can be too large for some capture devices, and algorithms can find it difficult to locate consistently replicable features in such signatures. What is the distribution of signatures over the population in terms of length and complexity? Although DSV algorithms offer many advantages over a visual comparison of signatures, they may be less sensitive to the shape of a signature. A commercial DSV algorithm interpreted the two signatures shown in Figure 1 as coming from the same person, where a human being would classify them as being different. In some cases it may be desirable to be able to extract the original signatures from the user template. This can be useful for legal reasons in situations where the ordinary static image of a person’s signature is needed, or to resolve disputes about verification. The above example shows that DSV might not always be able to resolve such disputes in a court of law. How do DSV algorithms compare with handwriting experts in analysing signatures? DSV can be used alongside digital signatures, for which there are already clear legal regulations in many countries and in the European Union. Some of the dynamic components of a signature which are commonly taken into consideration are the total time taken to sign and the speed of strokes. These tend to differ accordingly to what the user is signing for. If it is a job application, an individual may sign more slowly than if it is a phone bill. Furthermore, a person’s signature naturally varies according to mood, posture and the physical characteristics of the pen. Thus, ideally, enrolment should involve capturing signatures over a long time frame in order to allow for these variations. However, this is difficult and costly in practice. Also, the signature template can be updated every time a matching signature is recorded to accom- modate long-term changes to signatures. How does this affect the strength of DSV as a security mechanism? How much do signa- tures change over a lifetime? This may not have such an effect in the context of the same application, but it might have consequences if the same enrolment data is being used across multiple applications. Can algorithms successfully distinguish between someone signing carefully, and a forgery? In some applications, a user could sign a phrase which is not their usual signature, thus enabling users to be authenticated without revealing their signature. How consistently and distinctively can users write text which is not their usual signature? How many different signatures can users remember? In conclusion, there are clear technology and application drivers for the use of DSV. Over the next few years, we expect to see further advances in signature capture devices, especially the move towards paper- like flexible displays that will improve the potential for signing electronically. These advances, together with improvements in analysis algorithms, and the falling cost of computer processing will strengthen the case for the use of DSV. Whilst there are encouraging signs about the adoption of the technology, further research is still needed. References 1. http://bias.csr.unibo.it/fvc2002. 2. Marek Rejman-Greene et al, Roadmap for Biometrics in Europe to 2010, http://www.eubiometricforum.com. This article was contributed by Luann Rragami, Maurice Gifford and Nicholas Edwards at BT Exact. For further information email: [email protected] FEATURE Biometric Technology Today • November/December 2003 7 DSV – questions remain... In an age of e- and m-commerce and digital connectivity, the need to capture signatures securely and to verify the identity of people is becoming increasingly important. However questions surrounding dynamic signature verification technology still remain. Figure 1. These two signatures were identified as coming from the same person by a DSV algorithm.

Post on 05-Jul-2016

217 views

Category:

Documents


3 download

TRANSCRIPT

Page 1: DSV — questions remain…

Dynamic signature verification (DSV) isthe capture and verification of a signature toconfirm the identity of a user. Thetechnology offers a number of advantagesover other biometrics. People are familiarwith writing their signatures to authorisetransactions, and do not see it as athreatening technology. It can also be veryinexpensive to implement.

The primary reason that interest in DSV isgrowing is not because of algorithmimprovements, but because it is becomingincreasingly easy and possible to capturesignatures, not only from special tablets orpens, but also from PDAs, computer displayswith touch screens and the latest laptops. Thisis significant as it can make it possible forDSV to be implemented as a software onlysolution.

Use of such devices for electronic captureof signatures is now becoming widespreadin applications such as signing toacknowledge receipt of a parcel delivery.(Note, however, that simple signaturecapture systems do not record or comparedynamic features of signatures.)

The significant difficulties and frustra-tion which users have with passwords andPINs today, combined with the growth of e-and m-commerce, remote working, remotebanking, together with the growth of digitalconnectivity through broadband andwireless networks is creating an ever-greaterneed for innovative solutions forauthentication of an individual.

To date there have not been major trials ofDSV systems that test performance amongstdifferent sections of population, age groups,languages, etc. With many other biometricmethods such as fingerprint, facial or irisrecognition, large databases of sample imageshave been constructed enabling comparison

algorithms to be tested thoroughly offline.Therefore, an independent dynamic signaturetechnology benchmark similar to FVC1 forfingerprint is still awaited.2

Some questions remain open. Howunique are people’s signatures? What is thechance that people with the same namemight have similar signatures? Does this putpeople with common names at adisadvantage compared to those withunusual names? For example, in one majorUK enterprise alone, there are currently 285people called J. Smith.

Current DSV technologies can haveproblems with people who have either verylong or very short signatures. With shortsignatures, there is insufficient information tobe confident in the uniqueness of thesignature. However, long signatures can be toolarge for some capture devices, and algorithmscan find it difficult to locate consistentlyreplicable features in such signatures. What isthe distribution of signatures over thepopulation in terms of length and complexity?

Although DSV algorithms offer manyadvantages over a visual comparison ofsignatures, they may be less sensitive to theshape of a signature. A commercial DSValgorithm interpreted the two signaturesshown in Figure 1 as coming from the sameperson, where a human being would classifythem as being different.

In some cases it may be desirable to be ableto extract the original signatures from theuser template. This can be useful for legalreasons in situations where the ordinarystatic image of a person’s signature is needed,or to resolve disputes about verification. Theabove example shows that DSV might notalways be able to resolve such disputes in acourt of law. How do DSV algorithmscompare with handwriting experts inanalysing signatures? DSV can be usedalongside digital signatures, for which thereare already clear legal regulations in manycountries and in the European Union.

Some of the dynamic components of asignature which are commonly taken intoconsideration are the total time taken to signand the speed of strokes. These tend to differ

accordingly to what the user is signing for. Ifit is a job application, an individual may signmore slowly than if it is a phone bill.Furthermore, a person’s signature naturallyvaries according to mood, posture and thephysical characteristics of the pen. Thus,ideally, enrolment should involve capturingsignatures over a long time frame in order toallow for these variations. However, this isdifficult and costly in practice. Also, thesignature template can be updated every timea matching signature is recorded to accom-modate long-term changes to signatures.How does this affect the strength of DSV as asecurity mechanism? How much do signa-tures change over a lifetime?

This may not have such an effect in thecontext of the same application, but itmight have consequences if the sameenrolment data is being used acrossmultiple applications. Can algorithmssuccessfully distinguish between someonesigning carefully, and a forgery?

In some applications, a user could sign aphrase which is not their usual signature, thusenabling users to be authenticated withoutrevealing their signature. How consistentlyand distinctively can users write text which isnot their usual signature? How many differentsignatures can users remember?

In conclusion, there are clear technologyand application drivers for the use of DSV.Over the next few years, we expect to seefurther advances in signature capturedevices, especially the move towards paper-like flexible displays that will improve thepotential for signing electronically. Theseadvances, together with improvements inanalysis algorithms, and the falling cost ofcomputer processing will strengthen thecase for the use of DSV. Whilst there areencouraging signs about the adoption ofthe technology, further research is stillneeded.

References1. http://bias.csr.unibo.it/fvc2002.2. Marek Rejman-Greene et al, Roadmapfor Biometrics in Europe to 2010,http://www.eubiometricforum.com.

This article was contributed by LuannRragami, Maurice Gifford and NicholasEdwards at BT Exact. For furtherinformation email: [email protected]

FEATURE

Biometric Technology Today • November/December 20037

DSV – questions remain...In an age of e- and m-commerce and digital connectivity, the need to capturesignatures securely and to verify the identity of people is becoming increasinglyimportant. However questions surrounding dynamic signature verificationtechnology still remain.

Figure 1. These two signatures wereidentified as coming from the same personby a DSV algorithm.

btt novdecmasthead.qxd 22/10/2003 15:12 Page 7