Forensic Examination of Dynamic Signatures
Details
Download: thèse-LJ-OK.pdf (12302.19 [Ko])
State: Public
Version: After imprimatur
License: Not specified
State: Public
Version: After imprimatur
License: Not specified
Serval ID
serval:BIB_3596A4810DEE
Type
PhD thesis: a PhD thesis.
Collection
Publications
Institution
Title
Forensic Examination of Dynamic Signatures
Director(s)
Taroni Franco
Institution details
Université de Lausanne, Faculté de droit, des sciences criminelles et d'administration publique
Publication state
Accepted
Issued date
2022
Language
english
Abstract
Durant les dernières décennies, des changements profonds, tels l’internet et la numérisation ont impacté la vie des citoyens profondément. La digitalisation de la société a provoqué une évolu- tion rapide du concept de signature. La signature électronique, le fruit de ce développement, est un équivalent numérique de la signature manuscrite. La signature électronique manuscrite est un enregistrement de signatures, comprenant des données biométriques et kinématiques, telles la vitesse. Vu la simplicité de mise en place, le faible coût, la sécurité et le gain de temps occasionné, la signature dite dynamique connaît actuellement un grand essor. Les changements profonds de la signature manuscrite obligent les forensiciens à faire face à de nouveaux défis et problématiques. Ainsi, les litiges impliquant ces signatures sont inévitables et doivent être pris en charge. La nature logique et complexe des signatures électroniques manuscrites requiert des avancées méthodologiques pour garantir une examination scientifique et des résultats valides. Vu la nouveauté de ces signatures et la profondeur des changements, l’état des recherches est encore lacunaire. La complexité des données de signature impose l’intégration de nouvelles méthodes de traitement de données, ainsi que de banques de données de comparaison. Addi- tionellement, l’interprétation de ces données est fondamentale dans le processus forensique et requiert l’usage de probabilités et de statistiques pour exprimer les résultats de façon transpar- ente et scientifique. Cette thèse propose une approche théorique et statistique à la question des signatures imitées. Son but est de compléter la méthodologie d’examen de signatures dy- namiques. D’abord, l’état de l’art des signatures dynamiques est réalisé. Ensuite, des méthodes techniques et statististiques pour l’analyse de signatures dynamiques sont proposés, en liant les connaissances des domaines de biométrie, de neuroscience et de science forensique. Le mémoire approfondit les connaissances actuelles à travers d’analyses exploratoires et de statistiques réal- isés sur des données réelles, collectées dans le cadre du projet. Au coeur de la thèse est un modèle statistique Bayésien pour l’interprétation des données de signatures électroniques manuscrites. Ce dernier a été construit spécifiquement pour une application au problème de détection des signatures imitées. Ce modèle statistique et ses applications permettent d’étudier et de mieux comprendre le caractère dynamique des signatures. La validation du modèle avec des données de signatures réelles et d’imitations a permis de démontrer une performance et reproductibilité encourageante. Le modèle, dans une logique de transparence et de communication, permet à l’examinateur d’obtenir une évaluation quantifiée de la valeur de la preuve. Il permet ainsi la communication, en toute transparence, de résultats scientifiques justifiables appuyés sur des données empiriques.
--
In recent years, our society has rapidly evolved and been digitalized. These deep-rooted changes affect every citizen’s life, be it through government and commerce, communication or con- tracts. With the increasing number of important electronic transactions (e.g. e-government, e-commerce), the need for numerical authentication procedures has emerged. One of these so- lutions is electronic signatures, a numerical equivalent to its analogue homonym. Handwritten signatures have also been affected by the ongoing change. Dynamic signatures, complex record- ings of signature image and movements, as well as their biometric attributes have changed signa- ture examination. These electronic solutions have many advantages for businesses, such as their simple set-up, time- and cost-effectiveness, simple workflow, inherent security and ecological sus- tainability. As a result, dynamic signatures are currently enjoying an increase in popularity. Just as ’physical’ signatures are, dynamic signatures are important contractual elements. Conflict, and therefore, challenges to their authenticity, are inevitable. It falls to forensic experts to deal with the novel challenges related to dynamic signature authenticity. The complex handwritten electronic signature data requires a revision of existing methods in order to produce valid scien- tific evidence. Dynamic signatures are relatively novel and the state of their research is still in its beginning stages. The inherent data complexity calls for the integration of novel techniques, databases as well as adequate statistical models. A case assessment and interpretation method for signature forgery cases requires both statistics and probability to express results scientifi- cally. Such models are currently missing from the forensic and dynamic signature literature. This thesis proffers a logical and theoretical approach to the problem of forged signatures. It aims to do so by providing a deeper understanding of dynamic signatures, as well as proposing an extension to existing examination methodologies. To this effect, it features a state of the art on the forensically relevant knowledge, analysis methods and statistical tools, by uniting elements from biometrics, neuroscience and forensic science literature. The information relevant for an update of signature examination methodology is structured into four research propositions. At the heart of the thesis is a statistical Bayesian model constructed specifically for questioned and forged (i.e. simulated) signature scenarios. This model and its applications enabled the study and comprehension of the dynamic and kinematic features of signatures. Further, the validation of the model on known-source and large-scale data managed to produce encouraging results for evidence interpretation. The proposed model helps the forensic examiners quantify the value of disputed signature evidence. It may be used for the transparent communication of justifiable assessments of evidential value backed up by empirical data.
--
In recent years, our society has rapidly evolved and been digitalized. These deep-rooted changes affect every citizen’s life, be it through government and commerce, communication or con- tracts. With the increasing number of important electronic transactions (e.g. e-government, e-commerce), the need for numerical authentication procedures has emerged. One of these so- lutions is electronic signatures, a numerical equivalent to its analogue homonym. Handwritten signatures have also been affected by the ongoing change. Dynamic signatures, complex record- ings of signature image and movements, as well as their biometric attributes have changed signa- ture examination. These electronic solutions have many advantages for businesses, such as their simple set-up, time- and cost-effectiveness, simple workflow, inherent security and ecological sus- tainability. As a result, dynamic signatures are currently enjoying an increase in popularity. Just as ’physical’ signatures are, dynamic signatures are important contractual elements. Conflict, and therefore, challenges to their authenticity, are inevitable. It falls to forensic experts to deal with the novel challenges related to dynamic signature authenticity. The complex handwritten electronic signature data requires a revision of existing methods in order to produce valid scien- tific evidence. Dynamic signatures are relatively novel and the state of their research is still in its beginning stages. The inherent data complexity calls for the integration of novel techniques, databases as well as adequate statistical models. A case assessment and interpretation method for signature forgery cases requires both statistics and probability to express results scientifi- cally. Such models are currently missing from the forensic and dynamic signature literature. This thesis proffers a logical and theoretical approach to the problem of forged signatures. It aims to do so by providing a deeper understanding of dynamic signatures, as well as proposing an extension to existing examination methodologies. To this effect, it features a state of the art on the forensically relevant knowledge, analysis methods and statistical tools, by uniting elements from biometrics, neuroscience and forensic science literature. The information relevant for an update of signature examination methodology is structured into four research propositions. At the heart of the thesis is a statistical Bayesian model constructed specifically for questioned and forged (i.e. simulated) signature scenarios. This model and its applications enabled the study and comprehension of the dynamic and kinematic features of signatures. Further, the validation of the model on known-source and large-scale data managed to produce encouraging results for evidence interpretation. The proposed model helps the forensic examiners quantify the value of disputed signature evidence. It may be used for the transparent communication of justifiable assessments of evidential value backed up by empirical data.
Create date
18/07/2022 10:39
Last modification date
23/08/2022 9:50