Scholarlink

<p>The rapid growth of online education has introduced new challenges in maintaining academic integrity, particularly in MOOCs where sophisticated cheating strategies, such as CAMEO (Copying Answers using Multiple Existences Online), have emerged. This study proposed a novel method to detect CAMEO-style cheating by identifying suspicious harvester and master accounts without relying on IP address tracking, which can be unreliable in shared or masked network environments. Using behavioral analytics and temporal patterns of task engagement, we analyzed data from 558 accounts in a Java programming course on the Priscilla MOOC platform. Our findings indicated that harvester accounts exhibited minimal engagement, low variability in task repetitions, answer purchases, and time spent on tasks, but high variability in relative scores. Immediate-mode CAMEO behavior was detected effectively, while batch-mode behavior remained more challenging due to delayed submissions. The method highlighted the importance of behavioral variability metrics in distinguishing suspicious accounts, providing an alternative to traditional IP-based detection methods. These results underscore the potential for behavioral analytics to strengthen academic integrity in online learning environments, while emphasizing the need for further research to validate and generalize detection methods across larger and more diverse datasets.</p>

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

Unmasking CAMEO cheating in MOOCs via behavioral and temporal analysis without IP tracking

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月22日

查看全文
  • Matúš Valko,
  • Michal Munk,
  • Dasa Munkova

摘要

The rapid growth of online education has introduced new challenges in maintaining academic integrity, particularly in MOOCs where sophisticated cheating strategies, such as CAMEO (Copying Answers using Multiple Existences Online), have emerged. This study proposed a novel method to detect CAMEO-style cheating by identifying suspicious harvester and master accounts without relying on IP address tracking, which can be unreliable in shared or masked network environments. Using behavioral analytics and temporal patterns of task engagement, we analyzed data from 558 accounts in a Java programming course on the Priscilla MOOC platform. Our findings indicated that harvester accounts exhibited minimal engagement, low variability in task repetitions, answer purchases, and time spent on tasks, but high variability in relative scores. Immediate-mode CAMEO behavior was detected effectively, while batch-mode behavior remained more challenging due to delayed submissions. The method highlighted the importance of behavioral variability metrics in distinguishing suspicious accounts, providing an alternative to traditional IP-based detection methods. These results underscore the potential for behavioral analytics to strengthen academic integrity in online learning environments, while emphasizing the need for further research to validate and generalize detection methods across larger and more diverse datasets.