A Secure Online Proctoring System for Academic Integrity

Abstract

The fast development of online education has made the use of secure and reliable online examination systems more desired. Conventional internet-based exams in most cases encounter the problem of academic dishonesty and inefficiency in monitoring the exams. The current paper explains a proctoring system utilizing AI to examine learners online and guarantee academic integrity by observing the student behavior throughout the exams. The suggested system combines a web-based test platform and artificial intelligence and computer vision algorithms to identify suspicious activities on time. The webcam frames are analyzed by the system to detect behaviours that include, absence of faces, detection of multiple faces, and head pose deviation, movement of eyes and existence of illegitimate objects. Further, the activity of browsers like tab switching is tracked and cheating is discouraged. Rather than keeping recordings of large videos the system produces event logs with lists of timestamps, detected events, and risk scores which minimize storage space and facilitate viewing by instructors. In case the cumulative risk score goes beyond a specified limit, the system may automatically end the examination session. The architecture is made of a React based frontend, a Node.js based backend, and a Python based machine learning service based on computer vision libraries like OpenCV and YOLO. The offered system offers both efficient and scalable and privacy-aware to the system to undertake secure online tests and help educators to detect suspicious activity.