Incorporating IR-based Code Model Transformation into Token-Based Source Code Plagiarism Detection

Detecting source code plagiarism automatically has become an important problem for ensuring academic integrity. Automated refactoring tools have made detecting plagiarism more difficult, resulting in a need for more advanced, and therefore time-consuming, approaches. In this thesis, we evaluate IR-based transformations as a means of decreasing syntactic diversity to improve the detection accuracy of existing token-based source code plagiarism detection systems. Our results show that such transformations can provide a measurable improvement to result quality at comparable runtime cost to existing approaches.