Enhance Domain-Invariant Transferability of Adversarial Examples via Distance Metric Attack

Enhance Domain-Invariant Transferability of Adversarial Examples via Distance Metric Attack

A general foundation of fooling a neural network without knowing the details (i.e., black-box attack) is the attack transferability of adversarial examples across different models. Many works have been devoted to enhancing the task-specific transferability of adversarial examples, whereas the cross-...

Full description

Bibliographic Details
Main Authors: Jin Zhang, Wenyu Peng, Ruxin Wang, Yu Lin, Wei Zhou, Ge Lan
Format: Article
Language:English
Published: MDPI AG 2022-04-01
Series:Mathematics
Subjects:
deep learning
distance metric
adversarial attack
cross-task
transferability
Online Access:https://www.mdpi.com/2227-7390/10/8/1249

Internet

https://www.mdpi.com/2227-7390/10/8/1249

Similar Items