Research map / Adversarial robustness and security / Black-box and transfer attacks
Learning Black-Box Attackers with Transferable Priors and Query Feedback
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| alternate TIMI transferability steps with SimBA+ query steps | SimBA++ · ImageNet S1, Inception-V3 | AVG.Q: 725.2 -> 295.7 | Helped |
| learn the surrogate with HOGA using query feedback (ablation) | SimBA++ / LeBA · ImageNet S1, Inception-V3 | AVG.Q: 295.7 -> 243.8 | Helped |
| use a learned S1 surrogate at test time | SimBA++ · ImageNet S1, Inception-V3 | AVG.Q: 295.7 -> 230.6 | Helped |
| use a learned S1 surrogate at test time | SimBA++ · ImageNet S2, Inception-V3 | AVG.Q: 183.0 -> 151.3 | Helped |
| use surrogate saliency to guide coordinate selection and spatially smooth perturbations | SimBA · ImageNet S1, Inception-V3 | AVG.Q: 874.5 -> 725.2 | Helped |
| replace HOGA with a separately trained UNet gradient refiner (ablation) | LeBA · ImageNet S1, Inception-V3 | AVG.Q: 243.8 -> 286.5 | Hurt |
| use adaptive gradient compensation rather than fixed γ=1 (ablation) | LeBA · ImageNet S1, Inception-V3 | AVG.Q: 250.4 -> 243.8 | Helped |
| use adaptive gradient compensation rather than fixed γ=3 (ablation) | LeBA · ImageNet S1, Inception-V3 | AVG.Q: 248.3 -> 243.8 | Helped |
| add Forward Loss to Backward Loss (ablation) | LeBA · ImageNet S1, Inception-V3 | AVG.Q: 257.8 -> 243.8 | Helped |
| add Backward Loss to Forward Loss (ablation) | LeBA · ImageNet S1, Inception-V3 | AVG.Q: 276.8 -> 243.8 | Helped |
| use a ResNet-101 surrogate instead of ResNet-152 (ablation) | LeBA · ImageNet S1, Inception-V3 | AVG.Q: 243.8 -> 231.6 | Helped |
| use a ResNet-50 surrogate instead of ResNet-152 (ablation) | LeBA · ImageNet S1, Inception-V3 | AVG.Q: 243.8 -> 236.3 | Helped |
Closest papers by meaning
- Black-Box Adversarial Attack with Transferable Model-based Embedding (ICLR 2020 · similarity 0.92)
- Blackbox Attacks via Surrogate Ensemble Search (NeurIPS 2022 · similarity 0.91)
- Efficient Black-box Adversarial Attacks via Bayesian Optimization Guided by a Function Prior (ICML 2024 · similarity 0.91)
- Boosting Black-Box Attack With Partially Transferred Conditional Adversarial Distribution (CVPR 2022 · similarity 0.91)
- Understanding the Robustness of Randomized Feature Defense Against Query-Based Adversarial Attacks (ICLR 2024 · similarity 0.88)
- BRUSLEATTACK: A QUERY-EFFICIENT SCORE- BASED BLACK-BOX SPARSE ADVERSARIAL ATTACK (ICLR 2024 · similarity 0.87)
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.