A Lightweight Browser-Based Facial Recognition Framework for Preventing Impersonation in Computer-Based Examinations

Abstract

The rise of Computer-Based Testing (CBT) in academic institutions has streamlined assessment processes but introduced challenges in preventing impersonation, threatening exam integrity. This study proposes a real-time facial recognition system to verify student identities during CBT, implemented at the University of Ilorin, Nigeria. Utilizing face-api.js for browser-based facial recognition, Express.js for backend processing, MongoDB for secure data storage, and WebSocket for real-time administrative monitoring, the system achieves a verification accuracy of 94.7% with a false positive rate of 2.1%. A similarity threshold of 50% ensures only verified students access the CBT platform, with failed attempts flagged for administrative review. The system employs an iterative design-based methodology, integrating liveness detection to counter spoofing and HTTPS encryption for data security. Testing with 1,200 student photographs under varied conditions confirmed robust performance, with response times averaging 2.3 seconds. This study contributes to educational technology by demonstrating a scalable, privacy-conscious biometric solution that enhances CBT security, reduces administrative overhead, and upholds academic credibility.