Athens University Economics and Business (AUEB)
Statistical methods for the forensic examination of handwriting and dynamic signature data
Abstract
dc:descriptionNowadays, the evaluation of forensic evidence increasingly relies on statistical methodologies, which provide a coherent framework for quantifying evidential strength and addressing uncertainty. In the domains of handwriting and dynamic signature examination, the adoption of probabilistic models is enabling more objective, transparent, and scientifically grounded evaluations. This dissertation contributes to that ongoing development through two complementary studies.The first study introduces a novel statistical approach that aims at the identification of valid and useful patterns in handwriting examination via Bayesian modeling. Beginning with handwritten manuscript images, loop-character contours are characterized through Fourier-based features, including the first four pairs of coefficients and the surface size. Six Bayesian models are examined for such handwritten features. These models arise from two likelihood structures: (a) a Bayesian Normal model and (b) a Bayesian MANOVA model. For each likelihood, three different prior specifications are considered: a conjugate approach, a hierarchical Normal-Inverse-Wishart, and a Normal-LogNormal-LKJ prior. The Bayesian MANOVA model with hierarchical prior formulations is of primary interest because it can incorporate the within- and between-writer variability, as well as the between-character variability, which are particular distinguishing elements. Bayes factors are employed to assess the support for the competing propositions of interest (the person of interest (PoI) is the writer of the questioned document or not), aligning with the forensic international guidelines ENFSI. The study explores model comparison, the discriminative power, and a sensitivity analysis with respect to prior elicitation.The second study addresses the evaluation of the dynamic signature. Dynamic signatures are a behavioural biometric modality that captures the temporal characteristics of the signature. Hidden Markov Models (HMMs), including both Bakis and Ergodic structures, are used to model the sequential multivariate nature of signature data and to assess the likelihood ratios associated with the competing propositions of interest (questioned signature is genuine and was made by PoI or simulated). The study explores feature selection, automated determination of the hidden states, and the use of combined discriminative dynamic features to reduce the misleading support between competing propositions.Overall, these studies demonstrate how statistical methodologies can significantly advance forensic handwriting and dynamic signature examination. The integration of rigorous Bayesian and HMM-based techniques contributes to the development of more robust, interpretable, and scientifically grounded tools for the evaluation of evidence in both traditional and digital forensic contexts.
Degree
thesis:*- Grantor dc:publisher
- Athens University Economics and Business (AUEB)
- Year dc:date
- 2026
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Tzai, Lampis
- Τζάι, Λάμπης
Subjects
dc:subject × 24- Δικανική στατιστική
- Αξιολόγηση αποδεικτικών στοιχείων
- Μπεϋζιανή μοντελοποίηση
- Κρυφά μαρκοβιανά μοντέλα
- Χειρόγραφα δεδομένα
- Ψηφιακές δυναμικές υπογραφς
- Forensic statistics
- Evaluation of evidence
- Bayesian modeling
- Hidden Markov models
- Handwriting data
- Dynamic signatures
- Μαθηματικά
- Φυσικές Επιστήμες
- Στατιστική και Πιθανότητες
- Νομική Επιστήμη
- Κοινωνικές Επιστήμες
- Εγκληματολογία και Ποινικό δίκαιο - Ποινολογία
- Mathematics
- Natural Sciences
- Statistics and Probability
- Law
- Social Sciences
- Criminology and Penology
Rights
- Language dc:language
- eng
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/61934