ProctoAI: An AI-Powered Online Examination Proctoring System Using MERN Stack and Real-Time Computer Vision
Avinash Balasani, Pavan Bandaru, Sudhakar Banoth, Vamshi Banoth
Department of CSE (Artificial Intelligence & Machine Learning), ACE Engineering College
Affiliated to Jawaharlal Nehru Technological University, Hyderabad, Telangana, India
Under the Guidance of Mr Avinash , Associate Professor and Head, Department of CSE (AI & ML)
Abstract
The rise of online learning platforms has made it necessary to have strong ways to make sure that academic honesty is upheld during remote tests. This paper introduces ProctoAI, an AI-driven online exam proctoring system created with the MERN stack (MongoDB, Express.js, React.js, Node.js). The system uses TensorFlow.js and OpenCV-based Haar Cascade classifiers to give computers the ability to see in real time. This lets it keep an eye on students' behaviour through a webcam all the time. ProctoAI can find four types of suspicious activity: (1) when the student's face is missing, (2) when there are more than one face, (3) when a mobile phone is used with YOLO-based object detection, and (4) when a browser tab is switched. A rule-based weighted scoring system counts up violations and puts behaviour into one of three categories: Normal, Suspicious, or High Risk. Events that are detected are given a timestamp and saved in MongoDB. Instructors can access them through a special dashboard. Using JSON Web Tokens (JWT), the system lets students and teachers access it safely based on their roles. The tests show that violations are correctly detected, scores are correctly calculated, alerts are sent in real time, and the system works well for multiple users at the same time. ProctoAI is a scalable, automated alternative to human proctoring that makes online tests more fair and reliable while making them less dependent on manual supervision.
Keywords: AI Proctoring, Online Examination System, MERN Stack, TensorFlow.js, YOLO, Computer Vision, Face Detection, Tab Switching Detection, Secure Authentication, Real-Time Monitoring
