DEVELOPMENT OF AN AUTOMATED EYE GAZE TRACKING SYSTEM FOR DETECTING MALPRACTICE IN ONLINE ASSESSMENT
Abstract
This study describes the development an automated eye gaze monitoring system that can identify
possible malpractice in online tests. The initiative is driven by the growing reliance on digital
learning and the need to ensure academic integrity in remote environments, where traditional
proctoring methods are impractical. The system leverages computer vision techniques, utilizing
Python and libraries like OpenCV and MediaPipe to monitor eye gaze direction in real-time. A
rule-based approach identifies suspicious behaviors, such as prolonged off-screen stares or gaze,
or rapid eye movements, which may indicate malpractice. A user-friendly web interface, built with
HTML, CSS, and JavaScript, is supported by a FastAPI backend for efficient processing. The
system’s effectiveness was tested with 10 students from various academic levels (100 to 400).
Performance metrics, including gaze direction accuracy, malpractice detection accuracy. The
system demonstrated an average accuracy of over 85% during testing, as observed in the
experiment conducted in Chapter 4, with low computational demand. By focusing exclusively on
eye movement, the system offers a privacy-respecting approach to remote proctoring, avoiding
invasive methods like facial recognition or audio monitoring. It lays a robust foundation for future
improvements, such as integrating machine learning or compatibility with learning management
systems, providing a scalable solution for educational institutions such as the University of Agbor,
to secure online assessments and CBT based examinations
Full-Text Access Notice
In accordance with the NERD Policy on promoting peer-reviewed publication, public access to the full text of a project, thesis or dissertation is restricted for three years, allowing the author and supervisors sufficient time to pursue peer-reviewed publication.
During this period, researchers with legitimate academic or research purposes may request authorisation directly from the author to enable NERD to release the indexed full texts of the work using the form below.