Text Embedding Inversion Security for Multilingual Language Models
This addresses a security problem for users of multilingual NLP systems, highlighting a gap in existing defenses and offering a solution, though it is incremental as it builds on prior work on inversion attacks.
The paper tackles the security vulnerability of multilingual language models to text embedding inversion attacks, finding that multilingual models are more vulnerable and English-based defenses are ineffective, and proposes a simple masking defense that works for both monolingual and multilingual models.
Textual data is often represented as real-numbered embeddings in NLP, particularly with the popularity of large language models (LLMs) and Embeddings as a Service (EaaS). However, storing sensitive information as embeddings can be susceptible to security breaches, as research shows that text can be reconstructed from embeddings, even without knowledge of the underlying model. While defence mechanisms have been explored, these are exclusively focused on English, leaving other languages potentially exposed to attacks. This work explores LLM security through multilingual embedding inversion. We define the problem of black-box multilingual and cross-lingual inversion attacks, and explore their potential implications. Our findings suggest that multilingual LLMs may be more vulnerable to inversion attacks, in part because English-based defences may be ineffective. To alleviate this, we propose a simple masking defense effective for both monolingual and multilingual models. This study is the first to investigate multilingual inversion attacks, shedding light on the differences in attacks and defenses across monolingual and multilingual settings.