Authentication in Liveness Detection Utilizing CNN and MobileViT Algorithm

Liveness detection is a critical component in biometric security systems, aiming to distinguish between live and spoofed biometric samples to ensure system authentication.Image technology advancements are one of the factors that lead to attacks on liveness detection.The use of camera and mask, have made it less difficult to generate attacks that target the liveness detection system, including deep fake, replay, and print attacks.A reliable approach is required to more accurately identify these attacks.Recent advances in deep learning have shown significant promise in addressing these challenges by learning robust and adaptive features directly from raw biometric data.This paper provides an experimental research of deep learning approaches for liveness detection, focusing on Convolutional Neural Networks (CNNs) including EfficientNetV2S, EfficientNetV2M, EfficientNetV2L, and comparing with MobileViT for facial recognition in liveness detection.The datasets employed are NUAA, Synthetic, and iBeta 1.This paper examines the strengths and limitations of each method, and evaluation metrics used in the field, and highlight the latest breakthroughs in improving detection accuracy and robustness against diverse replay and print attacks.Experimental results show that EfficientNetV2S outperforms other algorithms, both in terms of accuracy and false detection rate.