Improved Facial Biometric Authentication Using MobileNetV2

One of the successful biometric technologies has been face recognition. Despite its enormous success, face recognition has drawbacks, much like any biometric technology. Face spoofing attacks, which are still possible with facial recognition technologies, have become a common threat to biometric applications. Many algorithms have recently been applied to publicly available datasets. However, efficiency has been a concern. One model cannot be a generalized solution for all attacks, so testing datasets over different models is needed to know all the possibilities. In this paper, a deep-learning method MobileNetV2 is proposed, which uses transfer learning and is a lighter convolutional network that is tested over the NUAA Dataset, which is publicly available. The proposed method shows better results than state-of-the-art methods over the same dataset.