A Secure Multi-Model Biometrics Using Deep Learning Model Based-Optimal Hybrid Pattern by the Heuristic Approach

Abstract

A new Deep Learning (DL)-based privacy preservation method using multimodal biometrics is implemented in this work. Here, the fingerprint, iris, and face are aggregated in the initial phase and fed to the Optimal Hybrid Pattern, where the Local Gradient Pattern and Local Weber Pattern are used. Thus, two sets of patterns from two diverse techniques for fingerprint, face, and iris are attained. Here, the Fitness-aided Random Number in Cheetah Optimizer (FRNCO) is used for optimization and also used for selecting the optimal Pixels to attain the optimal pattern. Further, these three pattern images are used to attain the histogram-based features, where the same FRNCO model is used for optimization. It is then forwarded to the final Deep Bayesian Network (DBN) with a Gated Recurrent Unit (GRU) termed the DB-GRU approach for acquiring the classified outcomes. The designed model is assimilated to recognize the efficacy of the developed model.