Advanced CNN-Based Hybrid Biometric Verification System for Enhanced Identity Authentication

Nicholas, Natalia Ellen (2025) Advanced CNN-Based Hybrid Biometric Verification System for Enhanced Identity Authentication. Masters thesis, Dublin, National College of Ireland.

Preview	PDF (Master of Science) Download (1MB) | Preview
Preview	PDF (Configuration Manual) Download (825kB) | Preview

Abstract

Existing biometric authentication systems often face challenges in trait diversity, adaptability to missing modalities, and energy efficiency. This study proposes a novel hybrid multimodal biometric authentication framework that addresses these challenges by using Siamese Convolutional Neural Networks (CNNs) based on deep learning in nine biometric traits (ear, face, ecg, opisthenar, periocular, palm touch, palm touchless, voice, and finger), introducing one of the broadest trait scopes yet addressed in biometric research. The system incorporates the fusion of feature and score levels to improve the accuracy and robustness of the verification system. Each trait-specific model was trained and evaluated using the LUTBIO dataset, with high performance traits such as face, finger, palm touchless, periocular (VGG16) and ecg (MobileNetV3Small) that achieve AUC values greater than 0.98. The fusion methods demonstrated further improvement, with the feature level generating an AUC of 0.9547 and the score-level through XGBoost with an AUC of 0.9895. Furthermore, testing was performed for generalization, and some trait models (ear, finger, palm touchless) were tested on public Kaggle datasets mean AUCs were 0.6367, 0.6934, and 0.6311, respectively, revealing how performance drops as data changes. A key contribution of this work is the inclusion of energy and carbon footprint profiling using the CodeCarbon library. All models exhibited low emissions (≤ 0.024 kg CO2),supporting feasibility on modest hardware. The proposed system demonstrates potential for identity verification in healthcare, education, and border control, but implementation and field validation are kept for future work. This work advances the field by promoting an inclusive, sustainable, and scalable approach to multimodal biometric authentication.

Item Type:	Thesis (Masters)
Supervisors:	Name Email Fajemisin, Ade UNSPECIFIED
Subjects:	Q Science > QH Natural history > QH301 Biology > Methods of research. Technique. Experimental biology > Data processing. Bioinformatics > Artificial intelligence Q Science > Q Science (General) > Self-organizing systems. Conscious automata > Artificial intelligence T Technology > TK Electrical engineering. Electronics. Nuclear engineering > Biometric Identification
Divisions:	School of Computing > Master of Science in Artificial Intelligence
Depositing User:	Ciara O'Brien
Date Deposited:	02 Jun 2026 11:37
Last Modified:	02 Jun 2026 11:37
URI:	https://norma.ncirl.ie/id/eprint/9335

Actions (login required)

View Item