Research map / Adversarial robustness and security / Certified and empirical robustness
Meta Invariance Defense Towards Generalizable Robustness to Unknown Adversarial Attacks
Lei Zhang, Yuhang Zhou, Yi Yang, Xinbo Gao
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 meta-learning over sampled attack combinations (ablation) | MID (LeNet5) · MNIST BIM_T | accuracy: 88.96 -> 90.64 | Improved |
| add distillation framework to meta learning (ablation) | MID (LeNet5) · MNIST BIM_T | accuracy: 90.64 -> 89.09 | Hurt |
| add adversarial consistency to meta+distillation (ablation) | MID (LeNet5) · MNIST BIM_T | accuracy: 89.09 -> 97.07 | Improved |
| add cyclic consistency to meta+distillation (ablation) | MID (LeNet5) · MNIST BIM_T | accuracy: 89.09 -> 97.99 | Improved |
| add both consistency constraints (ablation) | MID (LeNet5) · MNIST BIM_T | accuracy: 89.09 -> 98.78 | Improved |
| add meta-learning (ablation) | MID (ResNet18) · CIFAR10 FGSM_N | accuracy: 34.59 -> 37.17 | Improved |
| add meta-learning (ablation) | MID (ResNet18) · CIFAR10 CW_T | accuracy: 41.81 -> 64.57 | Improved |
| add adversarial and cyclic consistency to meta+distillation (ablation) | MID (ResNet18) · CIFAR10 FGSM_N | accuracy: 39.09 -> 41.63 | Improved |
| add both consistency constraints (ablation) | MID (ResNet18) · CIFAR10 benign | clean accuracy: 66.81 -> 64.48 | Hurt |
| use MID instead of adversarial training | ResNet18 · CIFAR10 black-box PGD_N | accuracy: 13.88 -> 64.84 | Improved |
| use MID instead of RSLAD | ResNet18 · CIFAR10 black-box CW_N | accuracy: 68.28 -> 64.34 | Hurt |
| use MID instead of AT | ConvNext-B · Tiny-ImageNet SPA | accuracy: 39.26 -> 33.89 | Hurt |
Closest papers by meaning
- On Fast Adversarial Robustness Adaptation in Model-Agnostic Meta-Learning (ICLR 2021 · similarity 0.88)
- Demystifying the Adversarial Robustness of Random Transformation Defenses (ICML 2022 · similarity 0.88)
- Practical Evaluation of Adversarial Robustness via Adaptive Auto Attack (CVPR 2022 · similarity 0.87)
- Rethinking Invariance Regularization in Adversarial Training to Improve Robustness-Accuracy Trade-off (ICLR 2025 · similarity 0.87)
- Enhancing the Adversarial Robustness via Manifold Projection (AAAI 2025 · similarity 0.86)
- Breaking Barriers in Physical-World Adversarial Examples: Improving Robustness and Transferability via Robust Feature (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.
