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-...
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: | |
Online Access: | https://www.mdpi.com/2227-7390/10/8/1249 |
Similar Items
-
Towards Adversarial Robustness for Multi-Mode Data through Metric Learning
by: Sarwar Khan, et al.
Published: (2023-07-01) -
On the Effectiveness of Adversarial Training in Defending against Adversarial Example Attacks for Image Classification
by: Sanglee Park, et al.
Published: (2020-11-01) -
Maxwell’s Demon in MLP-Mixer: towards transferable adversarial attacks
by: Haoran Lyu, et al.
Published: (2024-03-01) -
Adversarial attacks on cooperative multi-agent deep reinforcement learning: a dynamic group-based adversarial example transferability method
by: Lixia Zan, et al.
Published: (2023-07-01) -
The Framework of Cross-Domain and Model Adversarial Attack against Deepfake
by: Haoxuan Qiu, et al.
Published: (2022-01-01)