BIOPLAG: abordagem de detecção de plágio em código-fonte utilizando bioinformática

Abstract

The unauthorized reuse of source code characterizes plagiarism in programming, which can affect everything from the performance of students in programming courses to the quality of software development in companies. Through the realization of a systematic mapping study, the approaches of automatic detection of plagiarism in programming were analyzed to identify the used techniques, the evaluation test procedures, and the supported programming languages. It was found that the solutions do not include the different techniques used by plagiarists to change the source codes. This work created an approach, named BIOPLAG, capable of improving the automatic detection of plagiarism levels in source code. The functioning of the created approach is based on Bioinformatics and Computer Science techniques: tokens of programming language elements, mapping of source codes in synthetic biological sequences, and alignment of biological sequences. The implementation of BIOPLAG was evaluated through seven test scenarios containing 336 source codes implemented in C language used in 168 different tests, considering in each scenario the evaluative performance parameters: precision, recall and measure F. All examples of Plagiarized source codes were produced from three real experiments developed with the participation of students from undergraduate, graduate, and programmers from a software development company in the region. The results obtained were compared with two tools considered to be the reference in state of the art: MOSS and JPLAG. As a result, BIOPLAG performed better in four and equal in three test scenarios considering the indicators of precision, recall, and measure F.