A Geometry-Inspired Attack for Generating Natural Language Adversarial Examples

Zhao Meng, Roger Wattenhofer
2020 Proceedings of the 28th International Conference on Computational Linguistics   unpublished
Generating adversarial examples for natural language is hard, as natural language consists of discrete symbols, and examples are often of variable lengths. In this paper, we propose a geometryinspired attack for generating natural language adversarial examples. Our attack generates adversarial examples by iteratively approximating the decision boundary of Deep Neural Networks (DNNs). Experiments on two datasets with two different models show that our attack fools natural language models with
more » ... h success rates, while only replacing a few words. Human evaluation
doi:10.18653/v1/2020.coling-main.585 fatcat:7wpnzeeywbbaldc5qq45ank4ym