Enhancing Online Exam Security: Deep Learning Algorithms for Cheating Detection

Cheating in online exams has become a significant concern in recent years. There are various forms of online exam cheating, including copying answers from external sources. Fear of failure can drive some students to cheat and obtain high grades effortlessly. Many reducing cheating detection methods have limitations in terms of accuracy and time complexity. Even while some clever cheating detection systems have a high detection rate for cheating in videos, they have issue of a significant runtime overhead. Simple procedures with low time complexity tend to have accuracy issues with high false alarm rate. To address these challenges, a unique hybrid deep learning technique based on CNN-BiGRU has been proposed. This approach utilizes recorded videos during an exam to detect cheating instances. The methodology incorporates facial recognition, video analysis, eye tracking and object detection to identify potential cheating behaviors. To evaluate the performance of the proposed approach, 80% of the dataset is used for training the model, while the remaining 20% is employed for testing purpose. Subsequently, accuracy and evaluation matrix are computed. The constructed model achieves an impressive accuracy rate of 92%.