Development of AI-Powered Proctoring System for Computer Based Examination

Computer-based tests (CBT) are increasingly used in educational institutions, but ensuring exam integrity remains a challenge due to possibilities of cheating and unauthorized behavior. This final year project report presents the design and integration of the IntelliProctor Vision System, an AI-powered proctoring system tailored for computer-based examinations in resource-constrained environments. Developed as a lightweight web-based application, the system utilizes Flask for backend orchestration, SQLite for local data persistence, and computer vision / AI libraries including OpenCV, InsightFace (for facial recognition and verification), and YOLOv8n (for unauthorized object and person detection), alongside SoundDevice for real-time acoustic monitoring. System Capabilities & Architecture: Continuous Face & Identity Verification: Real-time embedding extraction and comparison to detect impersonation and face mismatches. Behavior & Object Monitoring: Real-time detection of multiple faces, gaze diversion (looking away), and unauthorized physical objects/devices via YOLOv8n. Audio & Screen Activity Tracking: Ambient noise/speech detection and browser-level tab-switching detection. Configurable Fairness Thresholds: Dynamic violation tolerances (e.g., face mismatch: 1, multiple faces: 2, looking away: 4, audio: 2, screen activity: 1) designed to reduce false positives. Audit Logging & Admin Dashboard: Persistent event logging paired with timestamped screenshots and integrity scoring for post-exam review. Key Performance Results: Detection Accuracy: Achieved an overall macro-average F1-score of 0.89 across all detection modules. Edge Performance: Maintained ~18 FPS with 78% CPU utilization on low-end hardware (Intel Celeron, 4GB RAM) with an overall storage footprint of under 1GB. Cloud Prototype Benchmark: Achieved ~32 FPS upon cloud container deployment on Hugging Face Spaces (IbnMuhd/IPS). Submitted to the Department of Computer Engineering, Ahmadu Bello University, Zaria, in partial fulfilment of the requirements for the award of Bachelor of Engineering (B.Eng) Degree in Computer Engineering.