Research map / Adversarial robustness and security / Black-box and transfer attacks
Generalizable Black-Box Adversarial Attack With Meta Learning
Fei Yin, Yong Zhang, Baoyuan Wu, Yan Feng, Jingyi Zhang, Yanbo Fan, Yujiu Yang
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 MCG initialization | Square · CIFAR-10 untargeted | mean query number: 227.3 -> 39.7 | Improved |
| add MCG initialization | Square · CIFAR-10 targeted | mean query number: 504.7 -> 92.3 | Improved |
| add MCG initialization | CG-Attack · ImageNet untargeted WRN-50 | ASR: 89.6% -> 99.2% | Improved |
| add MCG initialization | SimBA-DCT · ImageNet targeted ResNet-18 | ASR: 56.0% -> 66.8% | Improved |
| add MCG initialization | Signhunter · ImageNet targeted Inception-V3 | ASR: 100.0% -> 99.7% | No clear effect |
| add MCG initialization | CG-Attack · JPEG-Compress-WRN-50 defended | ASR: 39.0% -> 91.7% | Improved |
| add MCG initialization | NES · JPEG-Compress-WRN-50 defended | ASR: 13.2% -> 98.9% | Improved |
| add MCG initialization | Signhunter · FastAdv-ResNet-50 adversarially trained | ASR: 89.4% -> 89.4% | No clear effect |
| add MCG initialization | MetaAttack · Imagga tagging API | ASR: 16.0% -> 61.0% | Improved |
| add MCG initialization (surrogate and target on different datasets) | NES · CIFAR-10 to CIFAR-100 open-set | mean query number: 2054.0 -> 773.1 | Improved |
| add meta-training in place of plain flow training (ablation) | MCG w/ Square · ImageNet ResNet-18 | FASR: 35.2% -> 57.9% | Improved |
| add surrogate fine-tuning with target query feedback (ablation) | MCG w/ Square · ImageNet ResNet-18 | FASR: 57.9% -> 60.1% | 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.
- AdvFlow — AdvFlow: Inconspicuous Black-box Adversarial Attacks using Normalizing Flows (NeurIPS 2020) · better on 1 of 1 shared result
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
- Training Meta-Surrogate Model for Transferable Adversarial Attack (AAAI 2023 · similarity 0.92)
- Query-Efficient Black-Box Adversarial Attacks Guided by a Transfer-Based Prior (TPAMI 2021 · similarity 0.91)
- Meta Gradient Adversarial Attack (ICCV 2021 · similarity 0.91)
- Improving Black-box Adversarial Attacks with a Transfer-based Prior (NeurIPS 2019 · similarity 0.91)
- Enhancing Adversarial Transferability with Checkpoints of a Single Model's Training (CVPR 2025 · similarity 0.88)
- Enhancing Adversarial Transferability with Adversarial Weight Tuning (AAAI 2025 · similarity 0.88)
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.
