Imitation Attacks and Defenses for Black-box Machine Translation Systems

TL;DR

This paper demonstrates that black-box machine translation systems can be stolen through query-based imitation, and proposes a defense that modifies outputs to reduce attack success, balancing security and translation quality.

Contribution

It introduces a method to steal black-box MT models via imitation and proposes a defense that disrupts adversarial training by altering translation outputs.

Findings

Imitation models can reach within 0.6 BLEU of target systems.

Adversarial examples can cause semantically-incorrect translations.

The proposed defense reduces attack success but impacts BLEU and speed.

Abstract

Adversaries may look to steal or attack black-box NLP systems, either for financial gain or to exploit model errors. One setting of particular interest is machine translation (MT), where models have high commercial value and errors can be costly. We investigate possible exploits of black-box MT systems and explore a preliminary defense against such threats. We first show that MT systems can be stolen by querying them with monolingual sentences and training models to imitate their outputs. Using simulated experiments, we demonstrate that MT model stealing is possible even when imitation models have different input data or architectures than their target models. Applying these ideas, we train imitation models that reach within 0.6 BLEU of three production MT systems on both high-resource and low-resource language pairs. We then leverage the similarity of our imitation models to transfer adversarial examples to the production systems. We use gradient-based attacks that expose inputs which lead to semantically-incorrect translations, dropped content, and vulgar model outputs. To mitigate these vulnerabilities, we propose a defense that modifies translation outputs in order to misdirect the optimization of imitation models. This defense degrades the adversary's BLEU score and attack success rate at some cost in the defender's BLEU and inference speed.