Irregularity detection with face recognition and eye movements tracking in learning management system online exams

In recent years, distance education has become rapidly widespread in the globalizing world due to the rise of technology and pandemic outbreaks. With the transfer of education and training activities to the digital environment, various requirements related to distance education have emerged. The most important of these requirements are authentication, security and fair assessment and evaluation in distance education. This study aims to meet these requirements with the Learning Management System (LMS), which is called OKİNAR and developed within the scope of this thesis. In this system, face images are added to the LMS system along with other information when students are registered to the educational institution. In the LMS system, there are video conferencing rooms where students can participate in exams defined by the educational institution by turning on their cameras before starting their exams. In this system, the images from the camera are also saved as video files during the exam process to ensure exam security. After the participants are authenticated in the recorded videos, a list of participants who are likely to commit irregularities in the exam is compiled by analyzing the directions in which they look at places other than the screen. In other words, in this study, a solution has been developed to facilitate the detection of exam irregularities with face recognition technology. It is thought that this list will contribute to the re-evaluation of situations that are overlooked while making irregularities from the camera in online exams, and will facilitate follow-up in exams with many participants. With this system, as a result of the trials, it was determined that 234 out of 250 students did not commit irregularities, 16 students were interested in areas other than the screen, 9 out of 16 students committed irregularities, and 7 students did not take a position suitable for evaluation.