A Simple Yet Efficient Method for Adversarial Word-Substitute Attack

Tianle Li, Yi Yang

NLP researchers propose different word-substitute black-box attacks that can fool text classification models. In such attack, an adversary keeps sending crafted adversarial queries…