DC-CNNPAD to Enhance the Detection Rate for Iris Presentation Attack

Biometrics involves the analysis and statistical assessment of unique physical and behavioural characteristics of an individual. It finds application in areas like identification, access control, and surveillance. In security systems, biometric-based recognition is replacing conventional methods. Iris recognition (IR) has gained prominence in contemporary biometric technology deployed across various devices for security purposes. Recent advancements in deep convolutional neural networks (CNNs), computer vision, and access to extensive training data have significantly enhanced the performance of IR systems over the last decade. A presentation attack refers to a scenario where an impostor generates fake biometric data to deceive the system. This study introduces an effective strategy to enhance the precision of detecting iris presentation attacks and reviews the evolution of CNN techniques from 2015 to 2022. The proposed solution is a Dual-Channel Convolutional Neural Network Presentation Attack Detector (DC-CNNPAD), designed to improve the accuracy of real iris detection. An experiment is conducted on the LivDet-2015 dataset to evaluate the model's effectiveness in identifying artefacts. The results obtained from the detection model on the sample dataset demonstrate highly favourable outcomes, and on LivDet-2015, the TDR is 98.70%.