Real-Time Face Authentication Using Denoised Autoencoder (DAE) for Mobile Devices

Abstract

FBA (facial-based authentication), a non-contact biometric technology, has been evolving since its inception.. FBA can be used to unlock devices by showing their faces in front of devices. DLTs (deep learning techniques) have been receiving increased interest in FBA applications. Many proposals have used DLTs in this area. This chapter proposes DAEs (denoise auto encoders) for real-time classification of human faces. The proposed scheme balances accuracy with constraints of resource and time. The proposed DAE technique uses MDCs (mobile device cameras) for FBAs as they can address spoof or Windows-based attacks. The proposed DAE technique eliminates possible attacks on windows by immediately recognizing impostors. Moreover, feature extraction in DAE is dynamic and thus authenticates humans based on their facial images. Facial videos collected from MDCs results in realistic assessments. Spoof attacks using MDCs for bypassing security mechanisms are identified by DLTs in authentication.