Evolution Analysis of Homogenous Source Code and its Application to Plagiarism Detection

Due to intelligent softwares and worldwide internet environment, unauthorized source code theft/copying and partial plagiarism is widespread. So the detecting the plagiarized source codes and software is getting important, especially in academic field. Though there have been announced lots of studies for detecting plagiarized pair of codes, we did not find somewhat more fundamental work for understanding the mechanism of plagiarism. So in order to improve the plagiarism detecting algorithm, it is desirable to reveal the whole procedure of code plagiarism. The basic idea of this paper is that software plagiarism can be considered as an evolution process of a source code. The final goal of our paper is to reconstruct the phylogenetic tree with a set of plagiarized codes. The main contribution of this paper is to propose an asymmetric code similarity measure, which enables us to guess the direction of plagiarism. For this purpose, we applied local alignment approach to detect a region of similar codes with a novel adaptive matching score matrix. To show the effectiveness and efficiency of our procedure, we conducted experiments on 20 artificially generated phylogenetic trees whose roots are selected from four program groups of ICPC (International Collegiate Programming Contest). Experiment showed that the proposed algorithm can effectively estimate the evolution direction, which enables us to identify plagiarized codes more accurately and reliably.