Research map / Adversarial robustness and security / Black-box and transfer attacks
A Geometry-Inspired Decision-Based Attack
Yujia Liu, Seyed-Mohsen Moosavi-Dezfooli, Pascal Frossard
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 changed | Where | Measured | Effect |
|---|---|---|---|
| restrict gradient estimation to a 75x75 low-frequency DCT subspace (ablation) | qFool · ImageNet non-targeted, VGG-19 | MSE: 1.16e-5 -> 6.15e-6 | Improved |
| restrict gradient estimation to a DCT subspace (ablation) | qFool · ImageNet non-targeted, ResNet-50 | MSE: 1.03e-4 -> 4.14e-5 | Improved |
| restrict gradient estimation to a DCT subspace (ablation) | qFool · ImageNet non-targeted, Inception-v3 | MSE: 2.48e-4 -> 1.80e-4 | Improved |
| use qFool-fullspace instead of Boundary attack | decision-based attack · ImageNet non-targeted, VGG-19 | MSE: 4.61e-4 -> 1.16e-5 | Improved |
| use qFool-fullspace instead of Boundary attack | decision-based attack · ImageNet non-targeted, Inception-v3 | MSE: 1.79e-3 -> 2.48e-4 | Improved |
| use qFool instead of Boundary attack for iterations | decision-based attack · ImageNet non-targeted | number of iterations: ~500 -> ~3 | Improved |
| use white-box C&W instead of qFool-subspace | attack · ImageNet non-targeted, VGG-19 | MSE: 6.15e-6 -> 2.22e-7 | Hurt |
| use targeted qFool instead of Boundary attack | decision-based attack · ImageNet targeted | distance vs queries | Mixed |
| use subspace instead of full space in targeted attack (ablation) | qFool · ImageNet targeted, VGG-19, 50,000 queries | MSE: 4.40e-4 -> 1.12e-4 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- DeepFool — DeepFool: A Simple and Accurate Method to Fool Deep Neural Networks (CVPR 2016) · better on 0 of 4 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
- Triangle Attack: A Query-Efficient Decision-Based Adversarial Attack (ECCV 2022 · similarity 0.87)
- Efficient Decision-Based Black-Box Adversarial Attacks on Face Recognition (CVPR 2019 · similarity 0.87)
- QAIR: Practical Query-Efficient Black-Box Attacks for Image Retrieval (CVPR 2021 · similarity 0.86)
- GeoDA: A Geometric Framework for Black-Box Adversarial Attacks (CVPR 2020 · similarity 0.85)
- SparseFool: A Few Pixels Make a Big Difference (CVPR 2019 · similarity 0.85)
- Amnesia as a Catalyst for Enhancing Black Box Pixel Attacks in Image Classification and Object Detection (NeurIPS 2024 · similarity 0.83)
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.
