Title: Detecting AI-Assisted Cheating in Online Exams through Behavior Analytics Show Chinese title

Title: 通过行为分析检测在线考试中的AI辅助作弊

Abstract: AI-assisted cheating has emerged as a significant threat in the context of online exams. Advanced browser extensions now enable large language models (LLMs) to answer questions presented in online exams within seconds, thereby compromising the security of these assessments. In this study, the behaviors of students (N = 52) on an online exam platform during a proctored, face-to-face exam were analyzed using clustering methods, with the aim of identifying groups of students exhibiting suspicious behavior potentially associated with cheating. Additionally, students in different clusters were compared in terms of their exam scores. Suspicious exam behaviors in this study were defined as selecting text within the question area, right-clicking, and losing focus on the exam page. The total frequency of these behaviors performed by each student during the exam was extracted, and k-Means clustering was employed for the analysis. The findings revealed that students were classified into six clusters based on their suspicious behaviors. It was found that students in four of the six clusters, representing approximately 33% of the total sample, exhibited suspicious behaviors at varying levels. When the exam scores of these students were compared, it was observed that those who engaged in suspicious behaviors scored, on average, 30-40 points higher than those who did not. Although further research is necessary to validate these findings, this preliminary study provides significant insights into the detection of AI-assisted cheating in online exams using behavior analytics. Show Chinese abstract

Abstract: 人工智能辅助作弊在在线考试的背景下已成为一个重大威胁。先进的浏览器扩展现在可以使大型语言模型（LLMs）在几秒钟内回答在线考试中提出的问题，从而危及这些评估的安全性。在本研究中，使用聚类方法分析了在监督的面对面考试中，52名学生在在线考试平台上的行为，目的是识别可能与作弊相关的可疑行为的学生群体。此外，还比较了不同聚类中的学生的考试成绩。本研究中定义的可疑考试行为包括在问题区域选择文本、右键点击和失去对考试页面的焦点。提取了每个学生在考试期间执行这些行为的总频率，并采用了k均值聚类进行分析。研究结果表明，学生根据其可疑行为被分为六个集群。发现六个集群中的四个集群的学生，约占总样本的33%，以不同水平表现出可疑行为。当比较这些学生的考试成绩时，发现参与可疑行为的学生平均得分比没有参与的学生高出30-40分。尽管需要进一步的研究来验证这些发现，但这项初步研究为使用行为分析检测在线考试中的人工智能辅助作弊提供了重要的见解。