AN EXPERIMENTAL INVESTIGATION ON DIFFERENT EPOCHS AND SPLITTING DATA RATIOS FOR STUDENT AUTHENTICATION SYSTEM USING CONVOLUTIONAL NEURAL NETWORK (CNN) BASED FACE RECOGNITION

Abstract

In today's interconnected world, traditional username and password-based authentication methods are insufficient to safeguard sensitive data. This challenge is noticeable in Malaysian academic institutions, where these methods face vulnerabilities such as security breaches, forgotten passwords, and low user satisfaction. Weak passwords, reuse, and fake credentials further expose users to cyberattacks, highlighting the need for improved security and user experience. Face recognition using Convolutional Neural Networks (CNN) offers a promising solution, combining enhanced security with user-friendly identity verification. This study evaluates the performance of CNN based face recognition for improving authentication systems in Malaysian educational institutions. Experiments demonstrated the effectiveness of Student Authentication System Using CNN Based Face Recognition, achieving a maximum Average Recognition Accuracy of 100% and a minimum of 83% using varying epochs and data-splitting ratios. In conclusion, this approach has the potential to enhance security, usability, student experience, staff productivity, and institutional reputation.