ProctorSense: An AI-Enhanced Online Proctoring System with Proactive and Analytical Monitoring Techniques

This paper presents ProctorSense, an AI-enhanced online proctoring system designed to maintain academic integrity in remote examination environments. The system integrates both proactive and analytical monitoring techniques to detect and prevent cheating behaviors. The proactive component focuses on preventing window switching and screenshot attempts, while the analytical component employs head tracking and eye movement detection to identify suspicious behaviors. Based on a comprehensive survey of educators and students, the system was developed to address the most common cheating methods in online examinations. Experimental results demonstrate that the head-turning pattern classifier achieved an 84.3% overall accuracy, with particularly strong performance in detecting left (95.1% F1-score) and right (91.6% F1-score) head movements. The system proved effective in monitoring reading comprehension questions, establishing clear threshold intervals for automated detection: suspected cheating rates above 50% strongly indicate abnormal behavior, while rates below 25% suggest normal testing behavior. This research contributes to the growing need for reliable online proctoring solutions in educational institutions, offering a comprehensive approach to maintaining academic integrity in remote learning environments.