Research map / Adversarial robustness and security / Certified and empirical robustness
On the Robustness of Randomized Ensembles to Adversarial Perturbations
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 independently trained or MRBoost members with BARRE-trained REC (boosting with OSP) | randomized ensembles, ResNet-20 M=5 · CIFAR-10 l_inf eps=8/255 ARC | robust accuracy: 47.35 vs 46.66 (MRBoost-R) vs 45.77 (IAT) | Helped |
| Use BARRE-trained REC instead of single AT model | randomized ensembles · CIFAR-10 ResNet-18 l_inf eps=8/255 ARC | robust accuracy: 52.13 vs 48.72 | Helped |
| Randomized ensembling of MRBoost members (MRBoost-R) vs deterministic MRBoost ensemble | MRBoost ensembles · CIFAR-10 MobileNetV1 l_inf eps=8/255 | robust accuracy: 48.91 (REC, M=5) vs 49.36 (deterministic, M=4) | Hurt |
| Replace OSP with equiprobable sampling for MRBoost-R on MobileNetV1 (ablation) | MRBoost-R M=5 CIFAR-10 · CIFAR-10 MobileNetV1 | robust accuracy: 48.05 (equiprobable) vs 48.91 (OSP) | Hurt |
| Adding ensemble members via BARRE at fixed compute (81 MFLOPs) increases robustness (ablation) | BARRE REC ResNet-20 · CIFAR-10 l_inf eps=8/255 ARC | robust accuracy: 44.38 (M=2) -> 45.41 (M=3) -> 46.44 (M=4) | Helped |
| Using the first classifier as adversarially trained AT ensemble member (sharing f1 across RECs) | randomized ensembles · CIFAR-10 l_inf eps=8/255 ARC | robust accuracy | No clear effect |
Closest papers by meaning
- Adversarial Vulnerability of Randomized Ensembles (ICML 2022 · similarity 0.90)
- On the Role of Randomization in Adversarially Robust Classification (NeurIPS 2023 · similarity 0.90)
- Theoretical evidence for adversarial robustness through randomization (NeurIPS 2019 · similarity 0.89)
- Treatment of Statistical Estimation Problems in Randomized Smoothing for Adversarial Robustness (NeurIPS 2024 · similarity 0.87)
- Adversarial Robustness via Deformable Convolution with Stochasticity (ICML 2025 · similarity 0.87)
- AUTE: Peer-Alignment and Self-Unlearning Boost Adversarial Robustness for Training Ensemble Models (AAAI 2025 · similarity 0.86)
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.