Backdoor Attacks on Multilingual Machine Translation

Anonymous

Abstract: While multilingual machine translation (MNMT) systems hold substantial promise, they also have security vulnerabilities. Our research highlights that MNMT systems can be susceptible to a particularly devious style of backdoor attack, whereby an attacker can inject poisoned data into a low-resource language pair in order to malicious translations in a high-resource language. Our experimental results reveal that injecting less than 0.01% poisoned data into a low-resource language pair can achieve an average 20% attack success rate in attacking high-resource language pairs. This type of attack is of particular concern, given the larger attack surface of languages inherent to low-resource settings. Our aim is to bring attention to these vulnerabilities within MNMT systems with the hope of encouraging the community to address the security concerns in machine translation, especially in the context of low-resource languages.

Paper Type: long

Research Area: Machine Translation

Contribution Types: Model analysis & interpretability, NLP engineering experiment, Approaches to low-resource settings

Languages Studied: English; Indonesian; Javanese; Malay; Tagalong; Tamil

Consent To Share Submission Details: On behalf of all authors, we agree to the terms above to share our submission details.