Enhancing adversarial attack transferability with multi-scale feature attack

DOI10.1142/S0219691320500769zbMATH Open1493.68334OpenAlexW3095480789MaRDI QIDQ4990044FDOQ4990044

Authors: Caixia Sun, Lian Zou, Cien Fan, Yu Shi, Yifeng Liu 

Publication date: 28 May 2021

Published in: International Journal of Wavelets, Multiresolution and Information Processing (Search for Journal in Brave)

Full work available at URL: https://doi.org/10.1142/s0219691320500769

Recommendations

• A robust generative classifier against transfer attacks based on variational auto-encoders
• Black-box adversarial attacks by manipulating image attributes
• Spanning attack: reinforce black-box attacks with unlabeled data
• Robustifying models against adversarial attacks by Langevin dynamics
• Implicit adversarial data augmentation and robustness with noise-based learning

zbMATH Keywords

deep neural networksadversarial examplesblack-box attacksmulti-scale feature attack

Mathematics Subject Classification ID

Artificial neural networks and deep learning (68T07) Computer security (68M25)

Cited In (13)

• Vulnerability of classifiers to evolutionary generated adversarial examples
• Spanning attack: reinforce black-box attacks with unlabeled data
• A robust generative classifier against transfer attacks based on variational auto-encoders
• Black-box adversarial attacks by manipulating image attributes
• Towards improving fast adversarial training in multi-exit network
• Implicit adversarial data augmentation and robustness with noise-based learning
• Greedy attack and Gumbel attack: generating adversarial examples for discrete data
• An empirical study of derivative-free-optimization algorithms for targeted black-box attacks in deep neural networks
• Generating universal adversarial perturbation with ResNet
• Robustifying models against adversarial attacks by Langevin dynamics
• Generalizing universal adversarial perturbations for deep neural networks
• An adversarial attack based on multi-objective optimization in the black-box scenario: MOEA-APGA II
• Lagrangian objective function leads to improved unforeseen attack generalization

Uses Software

• ImageNet
• DeepSDF
• AlexNet
• Inception-v4

This page was built for publication: Enhancing adversarial attack transferability with multi-scale feature attack

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q4990044)