Skip to main content

Increasing Reliability and Security in Handwritten Signature Biometrics

Islam, Tasmina (2017) Increasing Reliability and Security in Handwritten Signature Biometrics. Doctor of Philosophy (PhD) thesis, University of Kent,. (Access to this publication is currently restricted. You may be able to access a copy if URLs are provided)

PDF
Restricted to Repository staff only until June 2022.
Contact us about this Publication Download (6MB)
[img]

Abstract

Biometric systems are increasingly being used as very convenient and efficient person identification systems, even within large populations. To be able to provide a successful, secure and reliable authentication process, different mechanisms (e.g. developing improved features or classification algorithms or template protection schemes - revocable biometrics, etc.) have been developed to protect systems from attacks and are deployed across many different modalities. The idea of revocability, where a fixed, unchanging biometric template is transformed into a revocable template, has been studied for both physiological and behavioural biometrics. But the concept of "natural revocability" introduced in this study, which most behavioural biometric modalities (and the handwritten signature in particular) offer, provides the possibility of adopting an extremely simple and intuitive strategy for the revocation process without the need for complex mathematical processing, because this is entirely under the user's control. This approach, however, has not been studied in relation to biometrics and data revocability hitherto and the lack of databases to support this type of investigation therefore imposes the need for generation of new data. The study reported in this thesis investigates the handwritten signature as the target biometric modality of interest in relation to natural revocability by means of an experimental study starting with the collection of "live" signature samples, and wide-ranging subsequent analysis. The suitability and effectiveness of natural revocability in handwritten signature biometrics as a practical option is investigated by observing how "stability" of the form of the signature changes over a period of time. The characteristics of potential revocability are also investigated by analysing performance and invoking the "biometric menagerie" notation for individual behaviour, while a more practically-oriented test of the viability of the natural revocability concept is performed by evaluating recognition performance. A feature-based analysis of the concept is presented by investigating some features commonly used in signature processing in both longstanding original and naturally revoked new signatures of a group of writers, and exploring the relationship between features, signature style and their effect in relation to original signatures and new signatures. This study also explores the development of a type of feature relating specifically to the concept of hesitancy (or its converse, fluency) for signature processing, which appears to be of particular relevance to the study reported here and investigates its impact on signature development in the context of natural revocability and signature verification more generally, using an objective measure of the power of the feature. The results from the experiments and the analysis provided suggest that if a sufficient time period is allowed then there is a high likelihood of convergence in terms of stability between a highly practised and long-standing signature and an alternative new representation, which also can be reliably recognised without degrading recognition accuracy, supporting the suitability and viability of the natural revocability concept. Exploring the influences of objectively defined hesitation features in creating the new signature also reveals that signers are more hesitant initially in signing the new signatures than the original - as might be expected - but gradually the hesitancies reduce with time, showing signers' increasing confidence in signing new signatures as time progresses; similarly, investigating recognising genuine and forged signatures shows that the hesitation is higher in forgery signatures than in genuine signatures, supporting the qualitative definition of hesitancy applied in a typical forensic scenario, an encouraging and effective performance improvement in discriminating genuine and forgery signatures. The study reported brings together two related ideas: the possibilities of adopting a natural revocability strategy in relation to security and reliability in handwritten signature analysis, and the development of a feature which may be particularly effective in the area of handwriting analysis, together with the aim to throw new light on issues relating to security in practical biometric systems.

Item Type: Thesis (Doctor of Philosophy (PhD))
Thesis advisor: Fairhurst, Michael
Uncontrolled keywords: Electronic Engineering Biometrics
Divisions: Faculties > Sciences > School of Engineering and Digital Arts
SWORD Depositor: System Moodle
Depositing User: System Moodle
Date Deposited: 02 Mar 2018 10:10 UTC
Last Modified: 29 May 2019 20:19 UTC
Resource URI: https://kar.kent.ac.uk/id/eprint/66265 (The current URI for this page, for reference purposes)
  • Depositors only (login required):

Downloads

Downloads per month over past year