An Optimised Artificial Neural Network Model for a Three-Level Authentication Security Scheme Utilising Fingerprint, Facial Recognition, and Optical Character Recognition

Abstract

The rapid increase in the use of digital technologies in daily activities has created both opportunities and threats. The paper reports an optimized Artificial Neural Network (ANN) model for implementing a three-tier authentication system using fingerprint biometrics (Level 1), facial recognition (Level 2) and Optical Character Recognition (OCR) (Level 3). The model is created using a multi-layer perceptron optimized using Adam and L2 regularization in order to have better accuracy and stability under environmental changes. On NIST SD4, LFW, and IAM datasets, an overall accuracy of 97.8% was reached with a false acceptance rate (FAR) of less than 1.0% was attained through experimental evaluation. The results show that the suggested model is better than unimodal techniques by about 16%, which proves its possible ability to protect e-learning and administrative systems at Nigerian universities.