Research map / Adversarial robustness and security / Black-box and transfer attacks
Improving Black-box Adversarial Attacks with a Transfer-based Prior
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 |
|---|---|---|---|
| Add the transfer-gradient prior and optimize its sampling weight λ* rather than use uniform RGF sampling | RGF black-box attacks · ImageNet, Inception-v3 | AVG. Q: 1309 -> 745 | Helped |
| Add the transfer-gradient prior and optimize its sampling weight λ* rather than use uniform RGF sampling | RGF black-box attacks · ImageNet, VGG-16 | AVG. Q: 749 -> 331 | Helped |
| Add the transfer-gradient prior and optimize its sampling weight λ* rather than use uniform RGF sampling | RGF black-box attacks · ImageNet, ResNet-50 | AVG. Q: 673 -> 265 | Helped |
| Replace adaptive λ* with fixed λ=0.5 (ablation) | P-RGF · ImageNet, Inception-v3 | AVG. Q: 745 -> 1119 | Hurt |
| Replace adaptive λ* with fixed λ=0.05 (ablation) | P-RGF · ImageNet, Inception-v3 | AVG. Q: 745 -> 1021 | Hurt |
| Replace adaptive λ* with fixed λ=0.5 (ablation) | P-RGF · ImageNet, VGG-16 | AVG. Q: 331 -> 710 | Hurt |
| Replace adaptive λ* with fixed λ=0.05 (ablation) | P-RGF · ImageNet, VGG-16 | AVG. Q: 331 -> 624 | Hurt |
| Replace adaptive λ* with fixed λ=0.5 (ablation) | P-RGF · ImageNet, ResNet-50 | AVG. Q: 265 -> 635 | Hurt |
| Replace adaptive λ* with fixed λ=0.05 (ablation) | P-RGF · ImageNet, ResNet-50 | AVG. Q: 265 -> 511 | Hurt |
| Add the transfer-gradient prior to RGF | RGF · JPEG compression | ASR: 41.5% -> 61.4% | Helped |
| Add the transfer-gradient prior to RGF | RGF · Randomization | ASR: 19.5% -> 60.4% | Helped |
| Combine the transfer-gradient prior with a data-dependent prior | P-RGF · Guided denoiser | ASR: 51.4% -> 89.6% | Helped |
Closest papers by meaning
- Efficient Black-box Adversarial Attacks via Bayesian Optimization Guided by a Function Prior (ICML 2024 · similarity 0.93)
- Boosting Black-Box Attack With Partially Transferred Conditional Adversarial Distribution (CVPR 2022 · similarity 0.92)
- Prior Convictions: Black-box Adversarial Attacks with Bandits and Priors (ICLR 2019 · similarity 0.92)
- LRS: Enhancing Adversarial Transferability through Lipschitz Regularized Surrogate (AAAI 2024 · similarity 0.90)
- Understanding the Robustness of Randomized Feature Defense Against Query-Based Adversarial Attacks (ICLR 2024 · similarity 0.90)
- Enhancing Adversarial Transferability with Checkpoints of a Single Model's Training (CVPR 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.