<![CDATA[ The increasing adoption of Computer-Based Testing (CBT) necessitates reliable security mechanisms to ensure examination integrity. This study aims to develop and evaluate a face recognition-based security system using Convolutional Neural Networks (CNN). The system is built using a dataset of 1,000 facial images from 100 individuals with variations in lighting, expression, and pose, and is designed for real-time identification and verification of examinees. The CNN model is employed to automatically extract facial features and perform identity classification. Experimental results show that the model achieves an accuracy of 95%, with false positive and false negative rates of 3% and 2%, respectively. The average recognition time of 0.75 seconds per individual indicates that the system is capable of real-time operation. Furthermore, the system demonstrates robustness against spoofing attacks, achieving detection rates of 98% for 2D photo attacks and 95% for video-based attacks. The system also integrates automated monitoring and anomaly detection mechanisms to enhance overall security. The results indicate that the CNN-based approach is effective in improving CBT security, reducing the risk of cheating, and maintaining the integrity of computer-based examinations.]]>
Copyrights © 2026
