Multi-modal Biometrics’ Template Preservation and Individual Identification

We introduce a system that provides multi-modal template security in response to the rising vulnerability to biometric templates. The proposed work aims to provide a multi-modal biometric identification system with protected templates that do not degrade overall recognition performance. The presented shielded technique was compared against an unprotected multi-modal biometric recognition system to prove the above metric. Many criteria are used to determine the success of the recommended system, including training time, testing time, Equal Error Rate (EER), accuracy, and classifier performance. Unique characteristics are acquired using Speeded-up Robust Features (SURFs) and Histogram of Oriented Gradients (HoG) from three biometric modalities (fingerprint, face, and signature). With the aid of the bio-secure template security method, the extracted characteristics have been fused, and templates have been turned into new templates. The generated template can be altered by simply adjusting the seed's random matrix. A virtual database is developed to evaluate the recommended approach. The hybrid feature extraction method is also assessed in addition to the performance of the single feature extraction strategy. Finally, the classifier and deep neural network are trained to predict the provided individual. The findings reveal that the protected approach improves the system’s overall recognition performance, and the EER value remains lower at different feature counts. The acquired highest accuracy is 96%, and the lowest EER is 0.07% on the 20 vital hybrid feature points.