Tracking Input Devices to Detect Cheating Using Machine Learning Techniques

Abstract

The holding of examinations online has risen in recent years, especially with the Covid-19 pandemic. Cheating in online examinations has also become rampant, causing concerns among examiners and institutions. This study explored the usage of machine learning to detect cheating in online exams. The research used input device data collected from participants which were fed into 3 machine learning models: Logistic Regression, SVM, and Random Forest. These were then combined to form an Ensemble model. Two feature sets were also used for the models: the first set consisted of four (4) features used in a previous study, while the second set consisted of seven (7) features. The study found that incorporating more features into the models produced better performance than when the models had only four (4) features. After evaluation, the Logistic Regression with Standardization had the best performance, having a sensitivity of 65% and specificity of 87%. This model was integrated into a quizzing system developed by the researchers. The system records the input device data of examinees while taking an exam, which is then fed into the machine learning model. This allows the system to detect which examinees cheated and the specific questions where they did.

Ähnliche Arbeiten