The Application of Artificial Intelligence in Preventing Cheating in Online Exams: Utilizing YOLO Technique and Multi-Agent Systems (MAS)

With the expansion of online education, maintaining academic integrity in virtual exams has become a significant challenge for educational institutions. This research presents an innovative system that combines the YOLO object detection algorithm and Multi-Agent Systems (MAS) to combat cheating. The YOLO algorithm in this system detects unauthorized objects, such as mobile phones and tablets, by analyzing video streams. Meanwhile, MAS analyzes participants' abnormal behaviors, such as head movements, hand gestures, or gaze shifts, to identify signs of cheating. Experimental results showed that the system achieved an ۸۷.۹% accuracy rate in detecting cheating. The YOLO algorithm performed highly effectively in identifying unauthorized objects, while the MAS framework accurately detected unusual behaviors. With a processing speed of ۰.۱ to ۰.۱۵ seconds per frame, the system is well-suited for real-time applications. By integrating deep learning algorithms and agent-based systems, this study provides a scalable and efficient solution to enhance the security of online exams. Implementing this system can strengthen trust in online assessment processes and ensure educational fairness.

کلیدواژه ها:

Cheating ، Online Exam ، YOLO ، Multi-Agent System ، Deep Learning