AI Paper Atlas

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

Decision-based Black-box Attack Against Vision Transformers via Patch-wise Adversarial Removal

Yucheng Shi, Yahong Han, Yu-an Tan, Xiaohui Kuang

NeurIPS 2022 · Adversarial robustness and security › Black-box and transfer attacks · approach: Robustness benchmarks and defenses · 2 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
initialize with PAR noiseHSJA · ILSVRC-2012 res-101median l2 distance: 35.13 -> 8.92Improved
initialize with PAR noiseSign-OPT · ILSVRC-2012 res-101median l2 distance: 50.18 -> 9.16Improved
initialize with PAR noiseBoundary · ILSVRC-2012 res-101median l2 distance: 16.25 -> 7.29Improved
initialize with PAR noiseCAB · ILSVRC-2012 res-101median l2 distance: 11.20 -> 4.73Improved
initialize with PAR noiseBBA · ILSVRC-2012 res-101median l2 distance: 9.00 -> 4.89Improved
initialize with PAR noiseCAB · ILSVRC-2012 vit_s16average l2 distance: 44.85 -> 24.96Improved
initialize with PAR noiseBoundary · ILSVRC-2012 vit_s16median l2 distance: 21.52 -> 11.09Improved
initialize with PAR noiseHSJA · ILSVRC-2012 vit_s16median l2 distance: 65.81 -> 14.55Improved
PAR alone versus initial Gaussian noisePAR · ILSVRC-2012 res-101median l2 distance: 58.60 -> 9.19Improved
lower minimum patch size (56 down to 7), initial patch 112 (ablation)PAR + Boundary · ILSVRC-2012 vgg-19median noise: 6.21 -> 4.31Improved
smaller minimum patch size 7 vs 56 (ablation)PAR + Boundary · ILSVRC-2012 vgg-19average query number: 16.80 -> 195.69Hurt
smaller initial patch size (112 to 14), min patch 7 (ablation)PAR + Boundary · ILSVRC-2012 vgg-19average query number: 195.69 -> 415.06Hurt

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