Characterizing students behavior in multi-user multi-computer testing environments

In recent years, multi-user, multi-computer testing environments have become widely utilized for examinations, quizzes, and programming assignments. This setup raises concerns about students accessing information from neighboring screens, facilitating potential solutions. In this paper, we propose a multi-user system capable of analyzing students’ behavior during testing sessions. To achieve this, we collected and labeled a dataset of students’ head postures, including instances of students looking left, looking right, and looking forward. The proposed system utilizes six facial landmarks extracted from detected faces to classify students’ behavior. We designed and trained a neural network to implement this system. Extensive testing demonstrates its effectiveness in identifying and characterizing the frequency and duration of gaze shifts. This research advances exam proctoring systems, bolstering their capability to detect and discourage unwanted behaviors in multi-user testing environments.