The document proposes developing an automatic cheating detection system for exam halls using machine learning. It aims to identify various forms of cheating like copying or communicating without invading privacy. The system will use YOLOv3 object detection and ShuffleNets architecture for real-time detection from multiple camera feeds. When cheating is detected, the system will promptly alert administrators for intervention while minimizing false positives and negatives. The system is intended to enhance exam integrity and fairness while reducing human errors.