Smart Proctor: Lightweight, Privacy-Centric Remote Exam Monitoring

The rising popularity of online tests has sparked the fear of academic dishonesty, bandwidth overuse, and privacy of candidates. Traditional remote proctoring systems to a large degree rely on constant live video streaming and single front-facing camera surveillance, which leads to high network usage, scalability constraints, and high privacy risks because of transmitting and storing sensitive visual information about personal setting. To address these issues, this paper introduces Smart Proctor, a lightweight and privacy oriented AI based remote examination monitoring system that will guarantee secure and fair online examinations. The suggested architecture instead of streaming video, captures images in regular intervals and conducts intelligent analysis behaviors in real-time on the candidate native device through on-device AI. The use of a dual-camera system, consisting of front and side facing cameras is employed to reduce the areas of monitoring blindness as well as improve the alertness of suspicious activities like gazes, absence of candidates, the presence of more than one person, and unauthorized use of objects. Instead of sending uncoded images or video streams, the system constructs small non-sensitive metadata flags that are safely transmitted over to a backend server to aggregate, generate alerts and visualize using a proctor dashboard. Movement of computational processing to the client side and reducing data transmission to necessary metadata, Smart Proctor helps greatly reduce bandwidth needs, decrease server load, and enhances privacy preservation without undermining efficient monitoring.