Research And Implementation Of Code Plagiarism Detection Technology In Online Measurement Systems

This paper proposed a more effective model for code plagiarism detection in program designing to make up for the lack of precision and the low speed in present plagiarism detection systems.Based on the new model,a more advanced detection system has been constructed for code plagiarism detection in the program designing course.The research contents are mainly as follows.First of all,to solve the problem of slow speed of detection,the thesis put forward an improved method of detecting similarity of source programs by combining attribute counting and structure calculation,which is proved effective.The result shows that it takes 15.1% less time to use the new method than to use the Longest Common Subsequence Algorithm.Then,in order to improve the precision of present detection systems,the code plagiarism detection in program designing is analyzed and the procedure of detection is optimized.By comparing the result of plagiarism detection by means of GST with that by combining attribute counting and structure calculation,it can be found that simple plagiarism in the model sample,the code sample from students' routine assignments,and that from computer-based exams can be detected easily using the new system with an accuracy of up to 95%,and this system is also 5.6% more precise than JPlag for sophisticated plagiarism,such as the increase of redundant codes and the replacement of equivalent structures.Lastly,the automation of code detection is achieved,including the pretreatment of codes,the effective choice of source programs,and the efficient abstraction of feature elements and the detection of similarity.To verify the effectiveness of the new detection system,three typical samples are selected to be detected respectively on the present system JPlag and the new one proposed in this paper.Through analyzing the detection of five common types of plagiarism in the three samples,it concludes that the precision in the new system has been improved by 7.3% compared with that in JPlag.Additionally,the new system can works smoothly and stably,which shows the experiment has achieved the expected goal.