Authorship Obfuscation in Multilingual Machine-Generated Text Detection

Macko, D., Moro, R., Uchendu, A.², Srba, I., Lucas, J. S.³, Yamashita, M.³, Tripto, N. I.³, Lee, D.³, Simko, J., Bielikova, M.

² MIT Lincoln Laboratory, USA
³ The Pennsylvania State University, PA, USA

High-quality text generation capability of recent Large Language Models (LLMs) causes concerns about their misuse (e.g., in massive generation/spread of disinformation). Machinegenerated text (MGT) detection is important to cope with such threats. However, it is susceptible to authorship obfuscation (AO) methods, such as paraphrasing, which can cause MGTs to evade detection. So far, this was evaluated only in monolingual settings. Thus, the susceptibility of recently proposed multilingual detectors is still unknown. We fill this gap by comprehensively benchmarking the performance of 10 well-known AO methods, attacking 37 MGT detection methods against MGTs in 11 languages (i.e., 10 × 37 × 11 = 4,070 combinations). We also evaluate the effect of data augmentation on adversarial robustness using obfuscated texts. The results indicate that all tested AO methods can cause evasion of automated detection in all tested languages, where homoglyph attacks are especially successful. However, some of the AO methods severely damaged the text, making it no longer readable or easily recognizable by humans (e.g., changed language, weird characters).

Cite: Macko, D., Moro, R., Uchendu, A., Srba, I., Lucas, J. S., Yamashita, M., … & Bielikova, M. (2024). Authorship obfuscation in multilingual machine-generated text detection. In Findings of the Association for Computational Linguistics: EMNLP 2024 (pp. 6348-6368). https://doi.org/10.18653/v1/2024.findings-emnlp.369