CLJan 15, 2024

Authorship Obfuscation in Multilingual Machine-Generated Text Detection

arXiv:2401.07867v338 citationsh-index: 32EMNLP
Originality Incremental advance
AI Analysis

This addresses the problem of detecting AI-generated disinformation in multilingual contexts, but it is incremental as it extends existing monolingual evaluations to a broader setting.

The authors investigated how authorship obfuscation methods affect multilingual machine-generated text detection, finding that all tested methods could evade detection across 11 languages, with homoglyph attacks being particularly effective, though some methods severely damaged text readability.

High-quality text generation capability of recent Large Language Models (LLMs) causes concerns about their misuse (e.g., in massive generation/spread of disinformation). Machine-generated 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 $\times$ 37 $\times$ 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).

Code Implementations2 repos
Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes