Initial Suspicion on Detecting Code Plagiarism and Collusion in Academia: Case Study of Algorithm and Data Structure Courses

Abstract

In engineering education, some assessments require the students to submit program code, and since that code might be a result of plagiarism or collusion, a similarity detection tool is often used to filter excessively similar programs. To improve the scalability of such a tool, it is suggested to initially suspect some programs and only compare those programs to others (instead of exhaustively compare all programs one another). This paper compares the ef-fectiveness of two common techniques to raise such initial suspicion: focusing on the submissions of smart students (as they are likely to be copied), or the submissions of slow-paced students (since those students are likely to breach academic integrity to get higher assessment mark). Our study shows that the latter statistically outperforms the former by 13% in terms of precision; slow-paced students are likely to be the perpetrators, but they fail to get the submissions of smart students.