Iris Liveness Detection for Biometric Access Control System in Smart Home Security using Deep Convolutional Neural Network.

Abstract

Biometric access control systems are essential for enhancing the security of smart homes. Among the various biometric modalities, iris recognition is a promising option due to its high accuracy and contactless nature. Nevertheless, presentation attacks, which try to trick the system using artificial or fake irises, may deceive iris recognition systems. To counter this threat, iris liveness detection (ILD) techniques are employed to distinguish between real and fake irises. In this paper, a novel and robust ILD method that combines handcrafted features and deep learning based features is proposed. The proposed method's performance is assessed across multiple machine learning classifiers and contrasted with existing ILD methods. The experimental results show that the proposed method achieves the lowest ACER values of 1.1% and 0.3% on the IIIT-D and Clarkson 2015 datasets, respectively, demonstrating its effectiveness and robustness against different types of presentation attacks.