PRACTICAL ASPECTS OF PLAGIARISM DETECTION IN COMPUTER SCIENCE E-LEARNING

We present some practical experiences in the application of e-learning methods and techniques in the computer science curriculum at Sofia University, where the e-learning system Moodle is being actively used to aid lectures and exercises in several programming courses, based on the languages C/C++ and Scheme. One of the grade components in those courses are homework assignments, which are electronically submitted by students using standard file upload functionality. The source code is evaluated by a teacher, rather than automatically, since in addition to program correctness other aspects of the submission are reviewed such as algorithmic efficiency, style, readability, modularity. Grades are individual and one of the conditions is that the submission needs to be original work. Unfortunately, very often collaboration between students ultimately leads to similar submissions due to plagiarism and subsequent skewing of grades. The present paper is focused on detection of plagiarism of electronically submitted homework assignments. We review several popular tools and techniques for source code analysis and evaluation of authorship and similarity. We compare their applicability in the light of several case studies based on actual student submissions and discuss their effectiveness in practical situations.