A Novel Approach for Detecting Logic Similarity in Plagiarised Source Code

Source code plagiarism is a persistent problem in computer science education. Many tools have been developed to identify indications of source code plagiarism in large data sets. These tools are good at identifying simple cases of plagiarism (e.g. renaming identifiers or shuffling declarations), but they are vulnerable to semantics-preserving obfuscations. This is due to a reliance on analyzing the structure of source code, and not the implemented program logic. In this paper, a novel approach to source code plagiarism detection is proposed that compares two programs for logic similarity. This approach analyses the logic embedded in a program as a means of identifying similarity. The approach is evaluated on a data set of simulated plagiarism. The evaluation demonstrates that the approach is resilient to semantics-preserving transformations.