Cancelable Template Generation Using Convolutional Autoencoder and RandNet

Abstract

The security of biometric systems has always been a challenging area of research to safeguard against the day-by-day introduction of new attacks with the advancement in technology. Cancelable biometric templates have proved to be an effective measure against these attacks while ensuring an individual’s privacy. The proposed scheme uses a convolutional autoencoder (CAE) for feature extraction, a rank-based partition network, and a random network to construct secured cancelable biometric templates. Evaluation of the proposed secured template generation scheme has been done on the face and palmprint modalities.