在线评测系统中代码剽窃检测技术的研究与实现

发布时间：2018-05-23 08:39

本文选题：属性计数法 + 结构度量技术　；参考：《西北农林科技大学》2017年硕士论文

【摘要】：为了高效检测程序设计课程中代码剽窃问题,减少或杜绝代码剽窃现象。针对目前已有的代码剽窃检测系统检测结果准确性较差、时间复杂度高等问题,本文提出一套完整的解决方案,并依据该方案构建出剽窃检测系统,用于程序设计课程的代码剽窃检测。主要研究内容如下:(1)针对已有检测系统运行时间复杂度高的问题,提出了一种基于属性计数法和结构度量技术相结合的程序相似度计算方法,通过提取代码属性特征和结构特征方法,降低了检测系统的时间复杂度。实验表明,检测系统的运行时间复杂度与最长公共子序列算法相比降低了15.1%。(2)针对已有检测系统准确性低的问题,通过分析代码剽窃的特点,优化了检测流程:优先使用GST(Greedy String Tilling)算法检测低级的代码剽窃问题,再依据决策函数判断是否采用属性计数法和结构度量技术相结合的方法进一步计算相似度,采用两次检测结果的综合评价得出最终检测结论,实验结果表明,对于构造样本、学生日常作业和上机考试提交的程序代码样本中的低级的剽窃手段,检测结果的准确率达95%;而对于较为高级的检测手段,比如增加冗余、等价结构替换等,本文设计的检测系统比JPlag系统的精度提高了5.6%。(3)实现了程序代码自动化检测系统一套,包括代码预处理、有效选择源程序的特征、高效提取特征元素及相似度检测算法的实现。为了验证本研究方法及构建出一套剽窃检测系统的有效性,选取三组典型样本在JPlag系统和本研究搭建的剽窃检测系统同时进行检测,最后比较检测结果。综合分析三组样本中常见的五类剽窃手段的检测结果比JPlag系统的精度提高了7.3%。软件测试表明,该系统能够稳定可靠地工作,较好的实现了设计的目标。
[Abstract]:To detect code plagiarism in programming courses efficiently, reduce or eliminate code plagiarism. Aiming at the problems of poor accuracy and high time complexity of the existing code plagiarism detection system, this paper proposes a complete solution, and constructs a plagiarism detection system based on this scheme. Code plagiarism detection for programming courses. The main research contents are as follows: (1) aiming at the problem of high time complexity of the existing detection system, a program similarity calculation method based on the combination of attribute counting and structure measurement is proposed. The time complexity of the detection system is reduced by extracting code attribute features and structural features. The experimental results show that the complexity of the detection system is 15.1than that of the longest common subsequence algorithm. (2) aiming at the problem of low accuracy of the existing detection system, the characteristics of code plagiarism are analyzed. This paper optimizes the detection flow: first using GST(Greedy String tracing algorithm to detect the low-level code plagiarism problem, and then judging whether to use attribute counting method and structure measurement technology to calculate the similarity according to the decision function. The final test results are obtained by comprehensive evaluation of the results of two tests. The experimental results show that, for the low-level plagiarism means of constructing samples, students' daily homework and program code samples submitted by computer examination, The accuracy of the detection results is 95%. For the more advanced detection methods, such as increasing redundancy and replacing equivalent structures, the precision of the detection system designed in this paper has been improved by 5.6% compared with that of the JPlag system.) the automatic detection system of program code has been realized. It includes code preprocessing, effective selection of source program features, efficient extraction of feature elements and implementation of similarity detection algorithm. In order to verify the effectiveness of this research method and construct a plagiarism detection system, three groups of typical samples were selected to simultaneously detect plagiarism in JPlag system and the plagiarism detection system built in this study. Finally, the detection results were compared. The accuracy of five kinds of plagiarism methods in three groups of samples is improved by 7.3% compared with that of JPlag system. The software test shows that the system can work stably and reliably, and achieve the design goal well.
【学位授予单位】：西北农林科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.1-4;G434

【参考文献】