English

Char-mander Use mBackdoor! A Study of Cross-lingual Backdoor Attacks in Multilingual LLMs

Computation and Language 2025-10-07 v3 Artificial Intelligence

Abstract

We explore \textbf{C}ross-lingual \textbf{B}ackdoor \textbf{AT}tacks (X-BAT) in multilingual Large Language Models (mLLMs), revealing how backdoors inserted in one language can automatically transfer to others through shared embedding spaces. Using toxicity classification as a case study, we demonstrate that attackers can compromise multilingual systems by poisoning data in a single language, with rare and high-occurring tokens serving as specific, effective triggers. Our findings expose a critical vulnerability that influences the model's architecture, resulting in a concealed backdoor effect during the information flow. Our code and data are publicly available https://github.com/himanshubeniwal/X-BAT.

Keywords

Cite

@article{arxiv.2502.16901,
  title  = {Char-mander Use mBackdoor! A Study of Cross-lingual Backdoor Attacks in Multilingual LLMs},
  author = {Himanshu Beniwal and Sailesh Panda and Birudugadda Srivibhav and Mayank Singh},
  journal= {arXiv preprint arXiv:2502.16901},
  year   = {2025}
}