Software for Detecting Suspected Plagiarism: Comparing Structure and Attribute-Counting Systems

The early automated systems for detecting plagiarism in student programs employed attribute-counting techniques in their comparisons of programs. It has been claimed that the more recent, automated systems, which encode structural information, are more effective in their detection of plagiarisms than systems based on attribute-counting. To explore the validity of these claims, a comparison is presented of two systems based on attribute counting and a structure-metric system. The major result of this study is that the systems based on structural information consistently equal or better the performance of systems based on attribute-counting metrics. A distinction is also made between plagiarisms due to novice programmers and those of more experienced programmers. 1. INTRODUCTION The aim of this study is to compare a number of different systems for detecting potential plagiarisms in program texts. Plagiarism is a persistent problem in many university courses, particularly those with pro..