Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Robustness Guarantees for Random Deep Neural Networks
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 |
|---|---|---|---|
| Use randomly initialized deep networks | Deep neural networks · Random inputs; varying input dimension | relative adversarial distance: O(1/\sqrt{n}) | Mixed |
| Train with stochastic gradient descent on MNIST | Simplified residual network · MNIST random images | \ell^1 adversarial distance | Hurt |
| Train with stochastic gradient descent on MNIST | Simplified residual network · MNIST training and test images | \ell^1 adversarial distance | No clear effect |
| Train with stochastic gradient descent on CIFAR10 | Simplified residual network · CIFAR10 random, training, and test images | \ell^1 adversarial distance: roughly half order of magnitude | Hurt |
| Train on CIFAR10 rather than use random weights | Simplified residual network · CIFAR10 random, training, and test images | \ell^1 adversarial distance | Hurt |
Closest papers by meaning
- Theoretical evidence for adversarial robustness through randomization (NeurIPS 2019 · similarity 0.91)
- Certified Adversarial Robustness with Additive Noise (NeurIPS 2019 · similarity 0.90)
- Adversarial Examples in Multi-Layer Random ReLU Networks (NeurIPS 2021 · similarity 0.90)
- Robust Design of Deep Neural Networks Against Adversarial Attacks Based on Lyapunov Theory (CVPR 2020 · similarity 0.90)
- Adversarial Training Can Provably Improve Robustness: Theoretical Analysis of Feature Learning Process Under Structured Data (ICLR 2025 · similarity 0.89)
- Probably Approximately Global Robustness Certification (ICML 2025 · similarity 0.89)
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.