AI Paper Atlas

Research map / Adversarial robustness and security / Certified and empirical robustness

Transferable Sparse Adversarial Attack

Ziwen He, Wei Wang, Jing Dong, Tieniu Tan

CVPR 2022 · Adversarial robustness and security › Certified and empirical robustness · approach: Adversarial training with feature consistency · 5 citations (OpenAlex)

Check your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
use the transferable sparse generator instead of GreedyFoolIncV3-source attack · ImageNet ℓ∞=255, IncV3→Res50fooling rate: 15.09 -> 63.76Improved
use the transferable sparse generator instead of GreedyFoolIncV3-source attack · ImageNet ℓ∞=10, IncV3→Res50fooling rate: 10.67 -> 45.32Improved
use the transferable sparse generator instead of GreedyFoolRes50-source attack · ImageNet ℓ∞=255, Res50→IncV3fooling rate: 29.12 -> 25.90Hurt
use the transferable sparse generator instead of GreedyFoolRes50-source attack · ImageNet ℓ∞=10, Res50→IncV3fooling rate: 12.35 -> 9.20Hurt
use the transferable sparse generator instead of GreedyFoolIncV3-source attack · ImageNet targeted bubble, IncV3→Res50fooling rate: 0.10 -> 10.38Improved
remove the decoupled location-mask branch and regularize perturbation directly (ablation)Transferable Sparse Adversarial Attack · ImageNet, IncV3 sourcesparsity: 0.46 -> 82.88Hurt
train without quantization (p=0) (ablation)Transferable Sparse Adversarial Attack · ImageNet ℓ∞=255, IncV3→Res50fooling rate: 63.76 -> 29.06Hurt
replace proposed quantization with STE (ablation)Transferable Sparse Adversarial Attack · ImageNet ℓ∞=255, IncV3→Res50fooling rate: 63.76 -> 50.06Hurt
remove sparse loss (ablation)Transferable Sparse Adversarial Attack · ImageNet, IncV3 sourcesparsity: 0.46 -> 100.00Hurt
remove quantization loss (ablation)Transferable Sparse Adversarial Attack · ImageNet ℓ∞=255, IncV3→Res50fooling rate: 63.76 -> 51.94Hurt
train the generator on one test image instead of many natural images (ablation)Transferable Sparse Adversarial Attack · ImageNet ℓ∞=255, IncV3→Res50fooling rate: 63.76 -> 43.97Hurt
increase sparse-loss weight from 5×10⁻⁶ to 2×10⁻⁴ (ablation)Transferable Sparse Adversarial Attack · ImageNet ℓ∞=10, IncV3→Res50fooling rate: 86.84 -> 32.82Hurt

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy