On Adversarial Examples for Character-Level Neural Machine Translation
Ebrahimi, Javid, Lowd, Daniel, Dou, Dejing
–arXiv.org Artificial Intelligence
Evaluating on adversarial examples has become a standard procedure to measure robustness of deep learning models. Due to the difficulty of creating white-box adversarial examples for discrete text input, most analyses of the robustness of NLP models have been done through black-box adversarial examples. We investigate adversarial examples for character-level neural machine translation (NMT), and contrast black-box adversaries with a novel white-box adversary, which employs differentiable string-edit operations to rank adversarial changes. We propose two novel types of attacks which aim to remove or change a word in a translation, rather than simply break the NMT. We demonstrate that white-box adversarial examples are significantly stronger than their black-box counterparts in different attack scenarios, which show more serious vulnerabilities than previously known. In addition, after performing adversarial training, which takes only 3 times longer than regular training, we can improve the model's robustness significantly.
arXiv.org Artificial Intelligence
Jun-23-2018
- Country:
- Europe > France (0.04)
- North America > United States
- Oregon (0.04)
- Asia > Middle East
- Israel (0.04)
- Genre:
- Research Report (0.64)
- Industry:
- Information Technology > Security & Privacy (0.66)
- Government > Military (0.48)
- Technology: