A New Approach to Cheating Detection in Online Exams

Due to the COVID-19 pandemic, there have been significant changes in different aspects of
education around the world, including the testing process. Online testing carried out without
any supervision replaced the traditional evaluation processes. However, their lack of validity
as a result of the increased possibility of cheating among the students concerns educators
about the results' authenticity. In such cases, providing additional processes to check the
results is a necessary step to administer safer online tests. In this research, a process with
high accuracy is proposed. This process is able to detect fraud based on data mining
techniques. It consists of several steps, including the use of statistical methods, similarity
criteria, student grades, K-nearest neighbor algorithms, artificial neural network, and
support vector machine. In this process, scoring has been done for each stage of the
operation, and the final statement to judge whether the cheating ever occurred or not will
be based on the sum score of each stage. The process uses the results of the students' exam
evaluation in order to identify abnormal scores in exams, and the recommender system is
able to detect students' cheating with 99.98% accuracy. The results show that the proposed
online test system has the ability to effectively reduce fraud and is able to help in providing a
valid online test.