Cheating Detection in E-exams System Using EEG Signals

Abstract

Cheating in e-exams is a real problem that threatens academic integrity and undermines confidence in the feasibility of remote assessments. Many previous research papers and studies discussed the issue of cheating in e-exams to prevent or reduce it through the use of the available technologies such as the use of a web camera to monitor the examinee, some researchers proposed using specific software to restrict the examinee from accessing resources that are not permitted during the exam. This work aims to design a Semi-automatic, AI-based e-proctoring system that mitigates cheating in e-exams. This research proposed an innovative method to detect the possibility of cheating in the e-exams. This method relies on the use of IoT and the Muse2 devices to detect the examinee's physiological state and determine whether it is “Normal” or “Abnormal” through the examinee`s EEG signal, where the abnormal state indicates a possibility of cheating. Convolutional Neural Network (CNN) was used to distinguish the examinee's state. The collected data from 15 students at the fourth stage of the Computer Engineering Department/ University of Mosul ranging between 23 and 26 years old showed that there is an obvious difference between the “calm” or “Normal” state and “stress” or “Abnormal” state in the EEG signal of the volunteer. The accuracy of the system was obtained for many testing datasets. The dataset was divided into two main datasets; the 30 and 60 seconds duration time. The best accuracy obtained for the 30sec duration time was 97.37%, and 97.14% for the 60sec duration time. The researchers concluded that the EEG signal contains a lot of important information that can be utilized to detect the physiological state of the examinee and that the Muse2 device can be reliable to record the EEG signal.