Protection of multi-biometric data based on deep learning and BioHashing
Fatima Bedad, Nassima Bousahba, Ikram Bennadji · 2023
Multi-biometrics is a nascent technological advancement that augments the process of individual authentication and identification via the use of numerous biometric modalities. Nevertheless, the susceptibility of multi-biometric systems lies in the difficulty of revoking biometric templates in the event of theft, hence giving rise to substantial apprehensions over privacy and security. In order to tackle this issue, a protection mechanism has been devised using a technique known as "Biohashing," which enables the reversible alteration of biometric characteristics. This particular method provides protection for the biometric template and enables the ability to revoke it if deemed appropriate. The proposed approach incorporates a biometric characteristic, namely fingerprints (referred to as instance 1 and instance 2), via the use of deep learning-based feature extraction, which is then followed by score fusion. The acquired findings illustrate the resilience of the "Biohashing" protection technique, successfully attaining a 0% error rate.