Software System for Similarity Detection in the Picocomputer Assembly Programs

This paper tackles the problem of plagiarism in an academic environment with an emphasis on the detection of similarities between the source codes from student assignments in the programming courses. The detected similarity in these codes greatly helps a human expert to bring the final decision on which codes are plagiarised and to which extent. Since the manual comparison of the source codes is a tedious task, the system for automatic detection of similarities in the assembly programs written for the picoComputer architecture is envisioned and implemented. It relies on the application which first performs the scanning and tokenization of the source codes. The pair-wise similarity detection is carried out by the Greedy String Tiling algorithm upgraded with the hash-based Karp-Rabin modification. A convenient GUI is also provided for efficient communication for the users and the choice of necessary parameters. Two different approaches are pursued in the testing and evaluation of the system. The first test set consists of a starting program with several versions with intentional modifications to simulate plagiarism. The second test set represents a real workload which comprises 250 real source codes from the student assignments. In both cases, the system demonstrated good efficiency.