ANTI-EXAM CHEATING DETECTION SYSTEM

Abstract

Science and technology are increasingly developing, the applications of Automation, Electronics, IoT, and Artificial
Intelligence technologies into practical applications have been applied in many different areas of our modern life.
Including management of activities in education. In particular, managing cheating in exams is always considered an
urgent problem, consuming a lot of human resources as well as inevitably missing when the number of people
performing supervision in each room is limited. This research proposes a solution that utilizes these interdisciplinary
knowledges to build a cheating behavior detection system in computer-based examination rooms. The proposed
system is implemented with two monitoring layers including: (1) Cheating detection system at the examination room
door using two cascade layers of checking by RFID card and a online facial recognition; (2) Cheating detection system
in the examination room. In order to detect cheating in the exam room, we first define, build, collect and anotate a
cheating image database. Then, we use YOLO V5 model with a new objective function to increase the efficiency of the
cheating action detection. Our proposed method gives a higher results than the original YOLO V5 model on our
EPUExCheating from 7.2% to 9.88%.