Towards Efficient Source Code Plagiarism Detection: An N-gram-based Approach

摘要

In this day and age, plagiarism has become a serious problem requiring the attention of the academic community at large. The problem, or plague as described in the literature, is very common in written works especially among university students due to various reasons such as time pressure, lack of understanding of what constitutes plagiarism, and the wealth of digital resources available on the Internet which make "copy/paste" activities almost natural! In order to deter students from submitting plagiarized work, educators must have a practical way to detect plagiarism. Currently there are many tools to help educators detect plagiarism within free-text essays but only a few that focus specifically on source code plagiarism. This paper proposes a new technique based on N-grams for this purpose. Our work improves on the state-of-the-art in this area by going beyond simple pair-wise submission comparisons as is the case with almost all techniques in the literature. Furthermore, our approach has empirically demonstrated improved efficiency compared to other approaches in the literature without noticeable sacrifices in accuracy.