DESIGN AND IMPLEMENTATION OF A HYBRID EXAMINATION MALPRACTICE PREVENTION SYSTEM

This study presents the design and implementation of a hybrid examination malpractice prevention system created to address the issues of academic integrity in exams and manual inefficient cheating detection process in schools, universities and colleges across the globe. The system makes use of a hybrid architecture consisting of a genetic algorithm for seating arrangement module and a YOLO model alongside facial movement tracking for the cheating detection with a FastAPI backend, SQL for storage and a HTML and CSS based front end. The system takes in a csv file containing consistent student data to generate a seating arrangement that makes sures students taking the same courses are sat as far apart from each other as possible and keeping students with a high risk of cheating towards the front to be invigilated better, students that have cheated before and have lower grades have a higher risk of cheating. It then goes to the camera based cheating detection to start monitoring students, it maps key points to each main upper body feature (nose, eyes, shoulders etc.) and flags it when a student turns their head to look at another students answers, when my custom YOLO model detects notes and phones. The system generates a seating arrangement of 100-500 students within 3 minutes and 501-2000 within 5 minutes and the camera based cheating detection is sufficiently efficient and detects frames at 1280 by 720 pixels in under 35ms per frame on average about 30 – 45 frames per second.