Plagiarism Detection in Code-Based Assignments

Abstract

Many engineering and computer science programs include computer based courses in which students are evaluated on code that they write individually or in groups. Plagiarism of code is typically difficult to detect , because courses have several markers, who cannot readily compare current submissions to those of students in other classes or in previous years. Our team is currently developing a plagiarism detection Web service that allows instructors to submit code and detect similarities between assignments and between projects. The interface is Web-based and is designed to smoothly integrate into instructors' marking activities. The presented Web service compares all up-loaded assignments, and identifies structural similarities. Since our service is based on structural analysis, superficial changes to variable names or control statements are not enough to conceal plagiarism. The results of similarity analysis are summarized in a simple format that indicates which assignments are the most similar and reports differences between assignments at the fine grain level of lexical tokens. The strength of this automated approach is that it removes the burden of plagiarism detection from instructors. In addition, it is fully repeatable and also provides a high level of accuracy with no false negatives.