Research map / Adversarial robustness and security / Black-box and transfer attacks
SeRI: Gradient-Free Sensitive Region Identification in Decision-Based Black-Box Attacks
Feiyang Wang, Xingquan Zuo, Hai Huang, Gang Chen, Hangwei Qian
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 SeRI refinement | HSJA · ImageNet VGG19 non-targeted, 2000 queries | avg l2 norm: 8.18 -> 6.47 | Improved |
| add SeRI refinement | HSJA · ImageNet VGG19 targeted, 2000 queries | avg l2 norm: 72.3 -> 48.3 | Improved |
| add SeRI refinement | CGBA · ImageNet VGG19 targeted, 2000 queries | avg l2 norm: 77.4 -> 53.0 | Improved |
| add SeRI refinement | CGBA · ImageNet ViT targeted, 2000 queries | avg l2 norm: 36.6 -> 29.0 | Improved |
| add SeRI refinement | ADBA · ImageNet ViT non-targeted, 10000 queries | avg l2 norm: 3.65 -> 2.39 | Improved |
| add SeRI refinement | RayS · ImageNet VGG19 non-targeted, 10000 queries | avg l2 norm: 2.72 -> 1.56 | Improved |
| use SeRI instead of PAR region sensitivity (ablation) | RayS · CIFAR-100 non-targeted, 2000 queries | avg l2 norm: 2.54 -> 1.79 | Improved |
| use SeRI instead of PAR region sensitivity (ablation) | CGBA · CIFAR-100 non-targeted, 2000 queries | avg l2 norm: 1.71 -> 1.48 | Improved |
| use SeRI instead of PAR region sensitivity (ablation) | CGBA · ImageNet VGG19 non-targeted, 2000 queries | avg l2 norm: 2.81 -> 2.92 | Mixed |
| use SeRI instead of PAR region sensitivity (ablation) | CGBA · ImageNet VGG19 targeted, 10000 queries | avg l2 norm: 20.4 -> 21.1 | Hurt |
| use 4x4 split instead of 2x2 split (ablation) | CGBA+SeRI · ImageNet VGG19 non-targeted, 5000 queries | avg l2 norm: 1.589 -> 1.645 | No clear effect |
| use 1x2 split instead of 2x2 split (ablation) | CGBA+SeRI · ImageNet VGG19 non-targeted, 2000 queries | avg l2 norm: 3.234 -> 3.292 | No clear effect |
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.
- CGBA — CGBA: Curvature-aware Geometric Black-box Attack (ICCV 2023) · better on 7 of 7 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
- Consistency-Sensitivity Guided Ensemble Black-Box Adversarial Attacks in Low-Dimensional Spaces (ICCV 2021 · similarity 0.84)
- Trust Region Based Adversarial Attack on Neural Networks (CVPR 2019 · similarity 0.84)
- Bias in Zeroth-Order Normal Estimation for Decision-Based Attacks (ICML 2026 · similarity 0.84)
- Stage-wise Attention-Guided Region Sequencing for Adversarial Attacks on Large Vision-Language Models (NeurIPS 2026 · similarity 0.84)
- Robust NAS under adversarial training: benchmark, theory, and beyond (ICLR 2024 · similarity 0.84)
- Contrasting Adversarial Perturbations: The Space of Harmless Perturbations (AAAI 2025 · 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.
