Szczegóły publikacji
Opis bibliograficzny
Comparative analysis of non-commercial plagiarism detectors for computer science education / Paulina GACEK, Bartosz Gdowski, Konrad Szymański, Wojciech Żmuda // W: Proceedings of the 18th International Conference on Computer Supported Education [Dokument elektroniczny] : May 18–20, 2026, Benidorm, Spain , Vol. 3 / eds. Edmundo Tovar, Tania Di Mascio, Christoph Meinel. — Wersja do Windows. — Dane tekstowe. — [Spain] : ScitePress, [2026]. — ( CSEDU ; ISSN 2184-5026 ). — e-ISBN: 978-989-758-833-4. — S. 1972–1982. — Wymagania systemowe: Adobe Reader. — Tryb dostępu: https://www.scitepress.org/Link.aspx?doi=10.5220/001483650000... [2026-09-01]. — Bibliogr. s. 1981–1982, Abstr. — Dostęp po zalogowaniu
Autorzy (4)
Słowa kluczowe
Dane bibliometryczne
| ID BaDAP | 169727 |
|---|---|
| Data dodania do BaDAP | 2026-09-02 |
| DOI | 10.5220/0014836500004021 |
| Rok publikacji | 2026 |
| Typ publikacji | materiały konferencyjne (aut.) |
| Otwarty dostęp |  |
| Creative Commons | |
| Konferencja | International Conference on Computer Supported Education 2026 |
Abstract
As Computer Science education shifts toward automated assessment to manage growing class sizes, maintaining academic integrity has become an increasingly complex challenge. Source code plagiarism is uniquely difficult to detect due to the limited syntactic entropy of programming languages and the natural logic convergence inherent in introductory assignments. This paper presents a comprehensive qualitative and quantitative evaluation of four prominent non-commercial detection systems: MOSS, JPlag, DOLOS, and copydetect. Our analysis reveals critical trade-offs between detection sensitivity and specificity. While JPlag and MOSS demonstrate high resilience against false positives, MOSS exhibits significant vulnerability to obfuscation attacks. Conversely, copydetect offers high robustness to such attacks but suffers from elevated false-positive rates. Furthermore, we evaluate the reporting capabilities of these tools, highlighting DOLOS’s superior cluster-based visualizations for identifying complex collusion groups. By synthesizing these empirical findings, we provide a practical guideline for educators to select tools that balance operational ease with the necessary resilience to safeguard academic integrity in modern programming courses.
