Scholarlink

With the rapid expansion of e-education, the demand for reliable solutions to uphold academic integrity has grown significantly. Online proctoring systems equipped with deep learning technologies have become crucial. These systems can automatically detect and prevent cheating using the advanced features of deep learning, ensuring credibility and fairness in online exams. The objective of this paper is to evaluate deep learning methodologies for online examination investigation. Experimental findings indicate that both CNN and ViT models effectively detect irregular activities in online examinations. ViT slightly outperforms CNN, achieving an accuracy of 94.04%. This study provides valuable insights for the development of AI-based proctoring systems for online examinations.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

CNN and Vision Transformer Models for Detecting Cheating in Online Examinations: A Comparative Evaluation

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Siham Essahraui,
  • Khalid El Makkaoui,
  • Mouncef Filali Bouami,
  • Ibrahim Ouahbi

摘要

With the rapid expansion of e-education, the demand for reliable solutions to uphold academic integrity has grown significantly. Online proctoring systems equipped with deep learning technologies have become crucial. These systems can automatically detect and prevent cheating using the advanced features of deep learning, ensuring credibility and fairness in online exams. The objective of this paper is to evaluate deep learning methodologies for online examination investigation. Experimental findings indicate that both CNN and ViT models effectively detect irregular activities in online examinations. ViT slightly outperforms CNN, achieving an accuracy of 94.04%. This study provides valuable insights for the development of AI-based proctoring systems for online examinations.