AI-Powered Online Exam Proctoring System for Secure and Scalable Assessment

The rise of remote education has introduced new challenges in maintaining academic integrity during online examinations. Traditional methods of in-person or manual remote proctoring are either impractical or resource-intensive, leading to increased incidents of academic dishonesty. This project proposes an AI-powered online proctoring system capable of monitoring student behaviour in real time to ensure fair and secure assessment environments. The system integrates facial recognition, eye gaze tracking, voice activity detection, and browser tab monitoring to detect behaviours such as looking away, multiple face presence, or window switching. Built using MediaPipe FaceMesh, OpenCV, and TensorFlow, it provides real-time anomaly detection and alerting. A Python-based backend coordinates proctoring logic, while Streamlit and React.js power the examiner and student interfaces. Detected events are logged to Cloud Firestore and accompanied by automated screenshots, enabling scalable, post-exam review. This framework offers academic institutions a robust, scalable, and automated approach to uphold integrity in digital assessments. Key Words: Academic Integrity, AI-based Monitoring, Eye Gaze Tracking Online Proctoring.