Beyond Weaponization: NLP Security for Medium and Lower-Resourced Languages in Their Own Right

TL;DR

This paper investigates the security of language models for lower- and medium-resourced languages, extending adversarial attacks to 70 languages, revealing limitations of monolingual and multilingual models in ensuring security.

Contribution

It extends adversarial attack methods to a wide range of languages and analyzes security issues specific to lower- and medium-resourced language models.

Findings

Monolingual models often lack sufficient parameters for security.

Multilingual models do not always improve security.

Lower-resourced languages require targeted security considerations.

Abstract

Despite mounting evidence that multilinguality can be easily weaponized against language models (LMs), works across NLP Security remain overwhelmingly English-centric. In terms of securing LMs, the NLP norm of "English first" collides with standard procedure in cybersecurity, whereby practitioners are expected to anticipate and prepare for worst-case outcomes. To mitigate worst-case outcomes in NLP Security, researchers must be willing to engage with the weakest links in LM security: lower-resourced languages. Accordingly, this work examines the security of LMs for lower- and medium-resourced languages. We extend existing adversarial attacks for up to 70 languages to evaluate the security of monolingual and multilingual LMs for these languages. Through our analysis, we find that monolingual models are often too small in total number of parameters to ensure sound security, and that while multilinguality is helpful, it does not always guarantee improved security either. Ultimately, these findings highlight important considerations for more secure deployment of LMs, for communities of lower-resourced languages.