It has been claimed that Jankowitz's algorithm avoids the limitations of previous approaches to the automated detection of plagiarism. In order to establish the veracity of such claims, a number of tests were conducted to assess the and of Jankowitz's approach. In the context of this paper, robustness refers to a tool's ability to detect similarity when a perpetrator has disguised their plagiarism. Efficiency is viewed as the ratio of matches which a detection system reports to the number of cases of plagiarism which are confirmed by manual inspection. The results indicate that this Jankowitz's algorithm cannot distinguish between cheating and laudable instances of software re-use. They also indicate that this approach provides limited support during the assessment of simple programs because there is too little evidence upon which to base accusations of plagiarism. It is concluded that current educational practice in the teaching of software engineering limits the utility of attempts to automate the detection of plagiarism.
Keywords: plagiarism; Jankowitz's algorithm; programming style; computer aided instruction.
The Department of Computer Science,
The University of York, Heslington, The United Kingdom, YO1 5DD.
E-mail: [email protected], Telephone: (0904) 433376.
C.W. Johnson and I.C. Wand
A CRITICAL EVALUATION OF JANKOWITZ'S ALGORITHM FOR
THE DETECTION OF PLAGIARISM IN STUDENT PROGRAMS