Enhancing Biometric Authentication Efficiency: A Hybrid Approach Exploiting Iris Modality and Leveraging One-Class SVM

Biometric characteristics play a vital role in the authentication and identification process, particularly in devel-oping resilient and secure systems with the main objective of safeguarding private data and ensuring secure access. Typically, two different modalities are commonly utilized: behavioural and physiological. Within the realm of physiological modalities, the iris stands out prominently due to its exceptional uniqueness and stability, rendering it exceedingly valuable in the context of security. This paper introduces a novel biometric-based authen-tication system that utilizes the iris as a modality. The proposed system is hybrid as it combines a pre-trained Convolutional Neural Network for feature extraction with the efficiency of a One-Class Support Vector classifier. To optimize the classification task and generate relevant features, Linear Discriminant Analysis is also employed. We evaluated the performance of the proposed system using two publicly available benchmark datasets. The system yield promising performance, achieving an accuracy of 99.07% and 99.68% with the CASIA-Iris-V1 and CASIA-Iris-Interval datasets, respectively.