Research map / Adversarial robustness and security / Black-box and transfer attacks
Boosting Decision-based Black-box Adversarial Attacks with Random Sign Flip
Weilun Chen, Zhaoxiang Zhang, Xiaolin Hu, Baoyuan Wu
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 |
|---|---|---|---|
| add random sign flip step (ablation) | SFA · CIFAR-10 untargeted | attack success rate: 9.1% -> 95.4% | Improved |
| add random sign flip step (ablation) | SFA · CIFAR-10 targeted | attack success rate: 0.2% -> 99.4% | Improved |
| add dimensionality reduction (ablation) | SFA · ImageNet untargeted, Inception-v3 | attack success rate: 91.1% -> 92.1% | Improved |
| add dimensionality reduction (ablation) | SFA · ImageNet untargeted, ResNet-50 | attack success rate: 96.9% -> 98.2% | Improved |
| add dimensionality reduction (ablation) | SFA · ImageNet untargeted, VGG-16 | attack success rate: 99.0% -> 98.7% | Improved |
| add dimensionality reduction (ablation) | SFA · ImageNet targeted, ResNet-50 | attack success rate: 98.6% -> 99.3% | Improved |
| add dimensionality reduction (ablation) | SFA · ImageNet targeted, Inception-v3 | attack success rate: 97.0% -> 95.8% | Hurt |
| add dimensionality reduction (ablation) | SFA · ImageNet targeted, ResNet-50 | average queries: 29,440 -> 22,538 | Improved |
| add dimensionality reduction (ablation) | SFA · ImageNet untargeted, VGG-16 | average queries: 2,112 -> 1,754 | Improved |
| use SFA instead of HSJA | SFA · ImageNet untargeted, Inception-v3 | attack success rate: 19.4% -> 92.1% | Improved |
| use SFA instead of HSJA | SFA · ImageNet targeted, Inception-v3 | attack success rate: 0.0% -> 95.8% | Improved |
| use SFA instead of HSJA | SFA · CIFAR-10 untargeted | attack success rate: 87.1% -> 95.4% | Improved |
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.
- Sign-OPT — Sign-OPT: A Query-Efficient Hard-label Adversarial Attack (ICLR 2020) · better on 2 of 2 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
- Understanding the Robustness of Randomized Feature Defense Against Query-Based Adversarial Attacks (ICLR 2024 · similarity 0.88)
- Sign Bits Are All You Need for Black-Box Attacks (ICLR 2020 · similarity 0.88)
- Meta-Learning the Search Distribution of Black-Box Random Search Based Adversarial Attacks (NeurIPS 2021 · similarity 0.87)
- Decision-Based Adversarial Attacks: Reliable Attacks Against Black-Box Machine Learning Models (ICLR 2018 · similarity 0.87)
- Sign-OPT: A Query-Efficient Hard-label Adversarial Attack (ICLR 2020 · similarity 0.87)
- Efficient Black-box Adversarial Attacks via Bayesian Optimization Guided by a Function Prior (ICML 2024 · similarity 0.86)
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.
