WordBlitz: An Efficient Hard-Label Textual Adversarial Attack Method Jointly Leveraging Adversarial Transferability and Word Importance

WordBlitz: An Efficient Hard-Label Textual Adversarial Attack Method Jointly Leveraging Adversarial Transferability and Word Importance

Existing textual attacks mostly perturb keywords in sentences to generate adversarial examples by relying on the prediction confidence of victim models. In practice, attackers can only access the prediction label, meaning that the victim model can easily defend against such hard-label attacks by den...

Full description

Bibliographic Details
Main Authors: Xiangge Li, Hong Luo, Yan Sun
Format: Article
Language:English
Published: MDPI AG 2024-04-01
Series:Applied Sciences
Subjects:
natural language processing
textual attack
hard label
adversarial samples
model robustness
Online Access:https://www.mdpi.com/2076-3417/14/9/3831

Internet

https://www.mdpi.com/2076-3417/14/9/3831

Similar Items