Detection of Iris Template Attacks using Machine Learning and Deep Learning Methods

Iris recognition is a biometric identification method that uses patterns in the iris, the coloured ring around the eye's pupil, to identify individuals. One of the most vulnerable attacks in iris detection is the template attack. An iris template attack is a type of biometric spoofing where an attacker attempts to gain access to a secure system by presenting a fabricated iris image or video as the authentic iris of the legitimate user. It is essential to detect these attacks to ensure the security and reliability of iris recognition systems and to prevent potential misuse and abuse. Machine learning and deep learning methods can potentially be used to detect template attacks on iris recognition systems, such as attempts to deceive the system using forged or altered iris templates. These methods involve training a model on a large dataset of iris images with both genuine and altered and using the trained model to classify new iris images as genuine or attacked. Hence, we propose a methodology that uses Logistic Regression (LR) and Convolutional Neural Network (CNN) to detect the Iris template attack. Through extensive experiments with the CASIA-IrisV1 dataset, CNN achieves an average accuracy of 98.75% and surpasses LR in recognizing iris template attack. Meanwhile, the performance of CNN is enhanced by applying the max pooling property to achieve 100% accuracy.