Research map / Adversarial robustness and security / Certified and empirical robustness
Stable Neural ODE with Lyapunov-Stable Equilibrium Points for Defending Against Adversarial Attacks
QIYU KANG, Yang Song, Qinxu Ding, Wee Peng Tay
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 SODEF layer on top of ODE (stability and diversity regularizers) | neural ODE · CIFAR100 AutoAttack L2 eps=0.5 | accuracy: 1.33 -> 79.10 | Improved |
| SODEF vs ODE under PGD | MNIST ResNet18 · MNIST PGD eps=0.3 | accuracy: 1.56 -> 45.25 | Improved |
| SODEF vs ODE under PGD | CIFAR10 · CIFAR10 PGD eps=0.1 | accuracy: 3.21 -> 55.59 | Improved |
| append SODEF to TRADES feature extractor | TRADES · CIFAR10 AutoAttack L2 | accuracy: 59.42 -> 67.75 | Improved |
| append SODEF to TRADES feature extractor | TRADES · CIFAR10 AutoAttack Linf | accuracy: 53.69 -> 57.76 | Improved |
| append SODEF to TRADES feature extractor | TRADES · CIFAR10 clean | accuracy: 85.48 -> 85.18 | No clear effect |
| increase integration time T from 1 to 10 (ablation) | SODEF · CIFAR100 AutoAttack L2 | accuracy: 2.00 -> 79.10 | Improved |
| fix last FC layer to diversity-promoting V instead of trainable (ablation) | SODEF · PGD eps=0.1 | accuracy: 34.9 -> 45.25 | Improved |
| fix last FC layer to V instead of trainable (ablation) | SODEF · FGSM eps=0.1 | accuracy: 51.6 -> 63.36 | Improved |
| transfer attack from ResNet18 instead of white-box attack | SODEF · CIFAR10 PGD eps=0.1 | accuracy: 55.59 -> 34 | Hurt |
| SODEF vs TisODE under FGSM | MNIST · MNIST FGSM eps=0.3 | accuracy: 36.70 -> 63.36 | Improved |
Closest papers by meaning
- Defending Against Adversarial Attacks via Neural Dynamic System (NeurIPS 2022 · similarity 0.90)
- Robust Design of Deep Neural Networks Against Adversarial Attacks Based on Lyapunov Theory (CVPR 2020 · similarity 0.90)
- Lyapunov-Stable Deep Equilibrium Models (AAAI 2024 · similarity 0.89)
- A Closer Look at the Adversarial Robustness of Deep Equilibrium Models (NeurIPS 2022 · similarity 0.87)
- Uniformly Stable Algorithms for Adversarial Training and Beyond (ICML 2024 · similarity 0.86)
- Stability and Generalization of Adversarial Training for Shallow Neural Networks with Smooth Activation (NeurIPS 2024 · similarity 0.85)
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.
