Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Vulnerability of Randomized Ensembles
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 |
|---|---|---|---|
| Replace APGD with ARC to evaluate a BAT randomized ensemble | ResNet-20 BAT REC · CIFAR-10 ℓ2 | robust accuracy (%): 69.21 -> 55.44 | Hurt |
| Replace APGD with ARC to evaluate a BAT randomized ensemble | ResNet-20 BAT REC · CIFAR-10 ℓ∞ | robust accuracy (%): 61.10 -> 40.71 | Hurt |
| Use a BAT randomized ensemble instead of a single AT model, evaluated with ARC | ResNet-20 · CIFAR-10 ℓ2 | robust accuracy (%): 62.43 -> 55.44 | Hurt |
| Use a BAT randomized ensemble instead of a single AT model, evaluated with ARC | ResNet-20 · CIFAR-10 ℓ∞ | robust accuracy (%): 45.66 -> 40.71 | Hurt |
| Replace APGD with ARC to evaluate a BAT randomized ensemble | BAT REC · SVHN ℓ2 | robust accuracy (%): 74.66 -> 60.15 | Hurt |
| Replace APGD with ARC to evaluate a BAT randomized ensemble | BAT REC · SVHN ℓ∞ | robust accuracy (%): 65.99 -> 52.01 | Hurt |
| Replace APGD with ARC to evaluate a BAT randomized ensemble | BAT REC · CIFAR-100 ℓ2 | robust accuracy (%): 41.91 -> 28.92 | Hurt |
| Replace APGD with ARC to evaluate a BAT randomized ensemble | BAT REC · CIFAR-100 ℓ∞ | robust accuracy (%): 33.37 -> 17.45 | Hurt |
| Replace APGD with ARC to evaluate a BAT randomized ensemble | BAT REC · ImageNet ℓ2 | robust accuracy (%): 49.62 -> 42.09 | Hurt |
| Replace APGD with ARC to evaluate a BAT randomized ensemble | BAT REC · ImageNet ℓ∞ | robust accuracy (%): 35.92 -> 19.54 | Hurt |
| Add random initialization to APGD | ResNet-20 BAT REC · CIFAR-10 ℓ2 | robust accuracy (%): 69.21 -> 69.28 | No clear effect |
| Add random initialization to APGD | ResNet-20 BAT REC · CIFAR-10 ℓ∞ | robust accuracy (%): 61.10 -> 62.80 | Hurt |
Closest papers by meaning
- Theoretical evidence for adversarial robustness through randomization (NeurIPS 2019 · similarity 0.92)
- On the Role of Randomization in Adversarially Robust Classification (NeurIPS 2023 · similarity 0.91)
- On the Robustness of Randomized Ensembles to Adversarial Perturbations (ICML 2023 · similarity 0.90)
- Adversarial Robustness via Deformable Convolution with Stochasticity (ICML 2025 · similarity 0.90)
- Treatment of Statistical Estimation Problems in Randomized Smoothing for Adversarial Robustness (NeurIPS 2024 · similarity 0.89)
- Towards Million-Scale Adversarial Robustness Evaluation With Stronger Individual Attacks (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.