Multimodal Biometric Authentication: Deep Learning Approach

Abstract

Authentication plays an important role in applications like data security. In this paper, we propose a Biometric authentication via multimodal biometrics (MBAS) where authentication is done with fingerprint and facial features. A multi-factor authentication system brings more security and reliability. The feature vectors are extracted using a deep learning model, and the feature vectors are merged. Both face and fingerprint features are extracted via transfer learning on different CNN models, and finally feature level fusion is used to authenticate the user. Various methods are compared in order to obtain the best level of accuracy. FaceNet, InceptionResNetV2, Xception, EfficientNetB3, and ResNet50 are the pre-trained models. Google Colab is used to train them. The performance of the models is analyzed using various datasets.