A Robust e-Invigilation System Employing Multimodal Biometric Authentication

Abstract

The significant growth in users of e-learning technologies and their use in courses has given rise to a major concern over protecting them from misuse; a significant concern is that of the potential for cheating or illicit assistance during online examinations. This paper presents the development of robust, flexible, transparent and continuous authentication mechanism for e-assessments. To monitor the exam taker and ensure that only the legitimate student is taking the exam, the system offers a continuous user identification employing multimodal biometrics; a security layer using an eye tracker to record the student’s eye movement; and, speech recognition to detect inappropriate communication. The focus of this paper in particular is the development and evaluation of 3D facial authentication. An experiment has been conducted to investigate the ability of the proposed platform to detect any cheating attempts. During the experiment, participants' biometric data, eye movement, and head movements have been collected using custom software. The 3D camera also captured the session using a built-in microphone and the system recognized speech (employing a speech recognition algorithm). 51 participants participated in this experiment. The FRR of all legitimate participants was 0 and 0.0063 in 2D and 3D facial recognition modes respectively. Furthermore, three participants were tasked with a series of eight scenarios that map to typical misuse. The results of the FAR and FRR of five of these threat scenarios in both 2D and 3D mode were 0 with two cases exhibiting an FAR of 0.11 and 0.076 in the 2D mode.