Enhanced multi-modal biometric recognition: A deep learning approach with optimized feature extraction

This increased need to have a dependable biometric authentication has inspired literature that oriented towards designing a multi-modal biometric recognition scheme based on iris, face, and fingerprint modalities. Biometric authentication requires powerful and safe systems. In this paper, we present a streamlined multi-modal biometric system that combines iris, face, and fingerprint modalities. A Deep Learning Model (EDLM) with Improved Secretary Bird Optimization (ISBO) is applied to tuning the hyperparameters and improve the features extraction. The system is highly accurate and efficient as it uses U-Net to perform iris segmentation, Enhanced ResNet to perform facial recognition, and a CNN hierarchical model to process fingerprints. The high-level performance has been proved by experiment, with high accuracy of 97.8 percent, false acceptance and false rejection rates low. The framework is also a scaled and reliable solution to real-world biometric authentication through data security using cancellable biometrics and through differential privacy. The system must provide the protection of user data through cancellable biometrics and differential privacy. The given framework is more efficient than the traditional methods because it is written in Python and tested on benchmark datasets. The results show that this system has proven to be a scalable and secure biometric authentication system and can be considered a major breakthrough in the field of multi-modal biometric.