A Systematic Review of Deep Learning Based Online Exam Proctoring Systems for Abnormal Student Behaviour Detection

Abstract

In the last years, educational technology has advanced tremendously. Increasing numbers of schools and universities are embracing online learning to serve their students better. As a result of the COVID-19 epidemic, students now have more flexibility in their study schedules and may work at their speed to better themselves. AI-based proctoring solutions have also grabbed the industry by storm. Online proctoring systems (OPS) generally employ online technologies to ensure that the examination is conducted in a secure environment. A survey of current proctoring systems based on artificial intelligence, machine learning, and deep learning is presented in this work. There were 41 publications listed from 2016 to 2022 after a comprehensive search on Web of Science, Scopus, and IEEE archives. We focused on three key study questions: current approaches for AI-based proctoring systems, techniques/algorithms to be employed, datasets used, and cheating detection methods suggested in such systems. Analysis of AI-based proctoring systems demonstrates a lack of training in using technologies, methodologies, and more. To our knowledge, Machine Learning or Deep Learning-based proctoring systems have not been subjected to such a study. From a technology standpoint, our research focuses on detecting cheating in AI-based proctoring systems. New recently launched technologies are included in this review, where these technologies potentially substantially influence online education and the online proctoring system.