EPlag: A two layer source code plagiarism detection system

Abstract

In academic environments where students are partly evaluated on the assignments, it is necessary to discourage the practice of copying assignments of other students. The detection of plagiarism in code from large source code repositories, manual detection is fairly complex, if not impossible. Therefore, for fair evaluation there must be a fast, efficient and automatedlsemi-automated way to detect the assignments copied. Source Code metrics can be used to detect the source code plagiarism in programming assignments submitted by university students. In this paper we have developed a source code plagiarism detection system and tried to improve the existing techniques by separating the suspected files and the non-plagiarized files, thus reducing the dataset for further comparison. A number of source code metrics have been calculated, combined using similarity detection formula to give an aggregate view of the source code metrics. After that the suspected files are separated and then performed string-matching to detect the level of similarity.