Unauthorized Activity Detection During Online Exams

The rapid growth of online education has increased the demand for secure and reliable examination systems. However, traditional online exams are highly vulnerable to impersonation and cheating activities. This paper presents an AI-based smart exam monitoring system that integrates face recognition, object detection, and audio analysis to ensure secure and fair examination processes. The proposed system uses facial recognition techniques to authenticate users during registration and login, preventing unauthorized access and duplicate identities. A face encoding mechanism is employed to uniquely identify each user and eliminate duplicate registrations. During the examination, real-time monitoring is performed using a deep learning-based object detection model (YOLO) to identify suspicious objects such as mobile phones, laptops, and the presence of multiple individuals. Additionally, head and eye movement tracking is implemented using Haar Cascade classifiers to detect inattentive or suspicious behavior. The system also incorporates audio monitoring to detect abnormal sound levels, indicating potential malpractice. Whenever suspicious activity is detected, the system generates alerts, captures screenshots as evidence, and records the entire session for further review. An automated decision mechanism is implemented to terminate the exam if abnormal behavior persists beyond a defined threshold duration. The integration of multiple monitoring techniques enhances the overall accuracy and robustness of the system. This solution provides a scalable and efficient approach to maintaining academic integrity in online examinations.