Anti-Cheating Surveillance System

Maintaining discipline and fairness during examinations is an important responsibility for invigilators, especially in classrooms with many students. Although manual supervision remains essential, continuously observing every student at all times can be challenging. To support this process, this project presents an Anti-Cheating Surveillance System based on computer vision and deep learning for automated detection of suspicious student behaviour during examinations. The proposed system uses a dual-camera setup to monitor the examination hall from multiple angles, improving coverage and reducing missed activities. Video recordings are analysed to detect behaviours such as peeking, note passing, mobile phone usage, and unnecessary student interaction. The system is designed to assist invigilators in maintaining a fair examination environment. For dataset development, approximately 1 hour and 20 minutes of video data is recorded in a controlled classroom setting, from which around 1000 short video clips are extracted and labelled as normal or suspicious activities. Preprocessing steps such as resizing, grayscale conversion, cropping, and temporal segmentation are applied before training. A hybrid CNN-LSTM architecture is implemented, where MobileNetV2 is used for spatial feature extraction and the LSTM layer is used to capture temporal motion patterns across video frames. This combination enables the model to recognize actions over time more effectively than image-based approach

Poster