AI Paper Atlas

Research map / Adversarial robustness and security / Black-box and transfer attacks

Transferable Adversarial Attack based on Integrated Gradients

Yi Huang, Adams Wai-Kin Kong

ICLR 2022 · Adversarial robustness and security › Black-box and transfer attacks · approach: Transferable adversarial examples · 18 citations (OpenAlex)

Read the paperCheck 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 random piecewise linear path integrated gradients instead of straight-line integrated gradients (ablation)TAIG · ImageNet val, untargeted, ε=0.03average attack success rate (%): 31.72 -> 52.82Improved
Use random piecewise linear path integrated gradients instead of straight-line integrated gradients (ablation)TAIG · ImageNet val, untargeted, ε=0.05average attack success rate (%): 52.13 -> 79.83Improved
Use random piecewise linear path integrated gradients instead of straight-line integrated gradients (ablation)TAIG · ImageNet val, untargeted, ε=0.1average attack success rate (%): 77.15 -> 96.90Improved
Use random piecewise linear path integrated gradients instead of straight-line integrated gradients (ablation)TAIG · Advanced defenses, ε=16/255average attack success rate (%): 34.84 -> 70.82Improved
Use random piecewise linear path integrated gradients instead of straight-line integrated gradients (ablation)TAIG · ImageNet val, targeted, ε=0.10average attack success rate (%): 5.15 -> 26.03Improved
Replace standard IFGSM gradients with integrated-gradient updatesILA · ImageNet val, untargeted, ε=0.1average attack success rate (%): 91.36 -> 96.57Improved
Use TAIG-R rather than IFGSM as the backendLinBP · ImageNet val, untargeted, ε=0.1average attack success rate (%): 94.50 -> 99.60Improved
Use TAIG-R rather than MIFGSM gradientsDI · ImageNet val, untargeted, ε=0.1average attack success rate (%): 70.16 -> 98.46Improved
Combine TAIG-R with other methods in targeted attacksTAIG-R · Targeted attackstransferabilityHurt
Combine TAIG-S with ILA in targeted attacksTAIG-S · Targeted attackstransferabilityHurt
Increase the number of AOA iterations from 100 to 300 (ablation)AOA · ImageNet val, ε=16/255average attack success rate (%): 32.07 to 32.19No clear effect
Increase the number of gradient calculations (ablation)SI and LinBP · ImageNet val, untargeted attacksaverage attack success rate (%)Improved

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