Research map / Adversarial robustness and security / Black-box and transfer attacks
Investigating Top-k White-Box and Transferable Black-Box Attack
Check 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 |
|---|---|---|---|
| replace CE with RCE | White-box attack on ResNet50 · ImageNet-compatible dataset | ICR: 752.90 -> 1000.00 | Helped |
| replace CW with RCE | White-box attack on ResNet50 · ImageNet-compatible dataset | ICR: 391.40 -> 1000.00 | Helped |
| replace CE with RCE | I-FGSM transfer from ResNet50 to DenseNet121 · ImageNet-compatible dataset, non-targeted | ICR: 34.16 -> 72.11 | Helped |
| replace CE with RCE | MI-DI-TI-FGSM transfer from ResNet50 to DenseNet121 · ImageNet-compatible dataset, non-targeted | ICR: 220.87 -> 482.58 | Helped |
| replace Po-Trip with RCE | MI-DI-TI-FGSM transfer from ResNet50 to VGG16bn · ImageNet-compatible dataset, targeted | targeted ASR@1: 37.20% -> 59.80% | Helped |
| replace CE with RCE | MI-FGSM transfer from ResNet50 to RN20 · CIFAR-10, non-targeted | ICR: 6.23 -> 8.23 | Helped |
| replace CE with RCE | MI-FGSM transfer from ResNet50 to RN20 · CIFAR-100, non-targeted | ICR: 24.31 -> 48.32 | Helped |
| target the least-likely class in RCE (ablation) | White-box attack on ResNet50 · ImageNet-compatible dataset | ICR: 1000.00 -> 687.36 | Hurt |
| set CE temperature from 1 to 1/100 (ablation) | White-box attack on ResNet50 · ImageNet-compatible dataset | ICR: 752.90 -> 76.89 | Hurt |
| increase CE temperature from 1 to 8 (ablation) | White-box attack on ResNet50 · ImageNet-compatible dataset | ICR: 752.90 -> 999.60 | Helped |
| increase iteration count while holding step size fixed (ablation) | I-FGSM transfer attack · ImageNet-compatible dataset | transfer rate | Helped |
| increase step size while holding iteration count fixed (ablation) | I-FGSM transfer attack · ImageNet-compatible dataset | transfer rate | Helped |
Closest papers by meaning
- GSBA$^K$: $top$-$K$ Geometric Score-based Black-box Attack (ICLR 2025 · similarity 0.84)
- PubDef: Defending Against Transfer Attacks From Public Models (ICLR 2024 · similarity 0.83)
- Stronger and Transferable Node Injection Attacks (AAAI 2024 · similarity 0.82)
- Enhancing Adversarial Transferability with Adversarial Weight Tuning (AAAI 2025 · similarity 0.81)
- Exploring the limits of strong membership inference attacks on large language models (NeurIPS 2025 · similarity 0.81)
- Adversarial Perturbations Are Formed by Iteratively Learning Linear Combinations of the Right Singular Vectors of the Adversarial Jacobian (ICML 2025 · similarity 0.81)
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.