Unimodal biometric identification system on Resnet-50 residual block in deep learning environment fused with serial fusion

Abstract

Purpose: Multifactor authentication biometric identification systems are tending on high, advance across the globe, and are significant in every sphere of life, financial business transactions, access control systems, cross-country border gateways, etc as security measures can not be ignored. So, robust multimodal identification systems are globally accepted and continuously upgraded with technical innovations by researchers, and technologists to protect from spoof cyber-attack and security breaches. Multimodal systems are difficult to manipulate, and breach the security layer due to tightly coupled verification feature passcode used due to concatenation on fused feature vector build[1]up by extracting important features of multiple traits that’s the basic robustness of the system. In this proposed model we have experimented with fusing four biometric traits i.e. facial, fingerprint, and hand-written signature together to make robust system.

Design/Methodology/Approach: The proposed model developed on Resnet-50 residual block base architecture was developed, executing the serial fusion algorithm for concatenation of important features forming fused feature vector (Fm) and experimented with two hundred fifty samples of each subject in deep learning environment.

Finding: The finding of our research experiments is that handwritten signatures have the highest technical feasibility for the concatenation of biometric feature with facial, fingerprint & Iris traits.

Originality /Value: Multimodal identification system has great impact & value, in terms of social security at individuals as well as social, economic & multi fold angel. In this research paper, original scenario of real life have been incorporated like hand written signature concatenating with facial, fingerprint & Iris traits for verification of identification in off-line mode and country cross boarder gateways & others.

Paper Type: Empirical Research Paper.