Adversarial attacks are label-preserving modifications to inputs of machine learning classifiers designed to fool machines but not humans. Natural Language Processing (NLP) has mostly focused on high-level attack scenarios such as paraphrasing input texts. We argue that these are less realistic in typical application scenarios such as in social media, and instead focus on low-level attacks on the character-level. Guided by human cognitive abilities and human robustness, we propose the first large-scale catalogue and benchmark of low-level adversarial attacks, which we dub Z\'eroe, encompassing nine different attack modes including visual and phonetic adversaries. We show that RoBERTa, NLP's current workhorse, fails on our attacks. Our dataset provides a benchmark for testing robustness of future more human-like NLP models.
@article{arxiv.2010.05648,
title = {From Hero to Z\'eroe: A Benchmark of Low-Level Adversarial Attacks},
author = {Steffen Eger and Yannik Benz},
journal= {arXiv preprint arXiv:2010.05648},
year = {2020}
}
Comments
Authors accidentally in wrong order; cannot be undone due to conference constraints. Accepted for publication at AACL 2020