Research map / Adversarial robustness and security / Certified and empirical robustness
Dual Manifold Adversarial Robustness: Defense against Lp and non-Lp Adversarial Attacks
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 on-manifold adversarial training term to standard AT (ablation) | AT [PGD-5] · OM-ImageNet test OM-PGD-50 | accuracy: 7.23 -> 20.53 | Helped |
| add on-manifold adversarial training term to standard AT (ablation) | AT [PGD-5] · OM-ImageNet test clean | accuracy: 73.31 -> 77.96 | Helped |
| add on-manifold term to standard AT (ablation) | AT [PGD-5] · OM-ImageNet test PGD-50 | accuracy: 38.88 -> 37.86 | No clear effect |
| add standard (off-manifold) AT term to on-manifold AT (ablation) | OM-AT [OM-PGD-5] · OM-ImageNet test PGD-50 | accuracy: 0.25 -> 37.86 | Helped |
| add standard AT term to on-manifold AT (ablation) | OM-AT [OM-PGD-5] · OM-ImageNet test OM-PGD-50 | accuracy: 27.53 -> 20.53 | Hurt |
| add on-manifold term to standard AT (ablation) | AT [PGD-5] · OM-ImageNet test Fog attack | accuracy: 19.76 -> 31.78 | Helped |
| add on-manifold term to standard AT (ablation) | AT [PGD-5] · OM-ImageNet test L1 attack | accuracy: 21.21 -> 29.68 | Helped |
| add on-manifold term to standard AT (ablation) | AT [PGD-5] · OM-ImageNet test Gabor attack | accuracy: 50.43 -> 51.61 | No clear effect |
| add on-manifold term to TRADES (ablation) | TRADES · OM-ImageNet test OM-PGD-50 | accuracy: 8.92 -> 26.82 | Helped |
| add on-manifold term to TRADES (ablation) | TRADES · OM-ImageNet test PGD-50 | accuracy: 46.06 -> 42.57 | Hurt |
| add on-manifold term to TRADES (ablation) | TRADES · OM-ImageNet test clean | accuracy: 69.88 -> 73.17 | Helped |
| add on-manifold term to standard AT (ablation) | AT [PGD-5] · natural out-of-manifold images, L1 attack | accuracy: 21.21 -> 29.68 | Helped |
Closest papers by meaning
- Enhancing the Adversarial Robustness via Manifold Projection (AAAI 2025 · similarity 0.92)
- On the Need for Topology-Aware Generative Models for Manifold-Based Defenses (ICLR 2020 · similarity 0.91)
- Adversarial Purification with the Manifold Hypothesis (AAAI 2024 · similarity 0.91)
- Explaining the role of Intrinsic Dimensionality in Adversarial Training (ICML 2025 · similarity 0.88)
- Manifold Constraints for Imperceptible Adversarial Attacks on Point Clouds (AAAI 2024 · similarity 0.87)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.87)
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.