Research map / Adversarial robustness and security / Certified and empirical robustness
Defending Against Adversarial Attacks via Neural Dynamic System
Xiyuan Li, Xin Zou, Weiwei Liu
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 the asymptotically stable neural ODE denoiser in front of the classifier (ablation) | ASODE · MNIST PGD | accuracy (%): 0.36 -> 46.85 | Improved |
| add the asymptotically stable neural ODE denoiser in front of the classifier (ablation) | ASODE · MNIST FGSM | accuracy (%): 9.87 -> 65.13 | Improved |
| add the asymptotically stable neural ODE denoiser in front of the classifier (ablation) | ASODE · CIFAR-10 PGD | accuracy (%): 3.18 -> 57.35 | Improved |
| add the asymptotically stable neural ODE denoiser in front of the classifier (ablation) | ASODE · CIFAR-10 FGSM | accuracy (%): 47.54 -> 69.94 | Improved |
| use ASODE (nonautonomous, asymptotic stability on inputs) instead of SODEF | ASODE · MNIST PGD | accuracy (%): 45.25 -> 46.85 | Improved |
| use ASODE instead of SODEF | ASODE · CIFAR-10 PGD | accuracy (%): 55.59 -> 57.35 | Improved |
| use ASODE instead of SODEF | ASODE · MNIST FGSM | accuracy (%): 63.36 -> 65.13 | Improved |
| use ASODE instead of SODEF | ASODE · CIFAR-10 FGSM | accuracy (%): 68.05 -> 69.94 | Improved |
| use ASODE instead of TisODE-Net | ASODE · MNIST PGD | accuracy (%): 1.78 -> 46.85 | Improved |
| use ASODE instead of TisODE-Net | ASODE · CIFAR-10 PGD | accuracy (%): 3.82 -> 57.35 | Improved |
| use ASODE instead of SODEF on clean inputs | ASODE · MNIST clean | accuracy (%): 99.44 -> 99.44 | No clear effect |
| use ASODE instead of SODEF on clean inputs | ASODE · CIFAR-10 clean | accuracy (%): 95.00 -> 95.16 | No clear effect |
Closest papers by meaning
- Stable Neural ODE with Lyapunov-Stable Equilibrium Points for Defending Against Adversarial Attacks (NeurIPS 2021 · similarity 0.90)
- Robust Design of Deep Neural Networks Against Adversarial Attacks Based on Lyapunov Theory (CVPR 2020 · similarity 0.90)
- Towards Deep Learning Models Resistant to Adversarial Attacks (ICLR 2018 · similarity 0.88)
- Lyapunov-Stable Deep Equilibrium Models (AAAI 2024 · similarity 0.86)
- Stability and Generalization of Adversarial Training for Shallow Neural Networks with Smooth Activation (NeurIPS 2024 · similarity 0.86)
- Efficient Robustness Evaluation via Constraint Relaxation (AAAI 2025 · 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.
