Research map / Adversarial robustness and security / Certified and empirical robustness
Computational Asymmetries in Robust Classification
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 the strong heuristic attack pool to estimate decision-boundary distance | Heuristic attacks · MNIST | mean distance overestimate vs. MIPVerify: 4.09±2.02% | Mixed |
| Use the strong heuristic attack pool to estimate decision-boundary distance | Heuristic attacks · CIFAR10 | mean distance overestimate vs. MIPVerify: 2.21±1.16% | Mixed |
| Use heuristic attack-pool distances as predictors of the true decision-boundary distance | Strong attack pool · MNIST | R^2: 0.992±0.004 | Helped |
| Use heuristic attack-pool distances as predictors of the true decision-boundary distance | Strong attack pool · CIFAR10 | R^2: 0.997±0.003 | Helped |
| Switch from the strong to the balanced parameter set | Heuristic attack pool · MNIST | R^2: 0.992±0.004 -> 0.990±0.006 | No clear effect |
| Switch from the strong to the balanced parameter set | Heuristic attack pool · CIFAR10 | R^2: 0.997±0.003 -> 0.998±0.002 | No clear effect |
| Use only BIM, Brendel & Bethge, and Carlini & Wagner attacks (ablation) | Heuristic attack pool · MNIST+strong | R^2: 0.988±0.004 | Helped |
| Use only BIM, Brendel & Bethge, and Carlini & Wagner attacks (ablation) | Heuristic attack pool · MNIST+balanced | R^2: 0.986±0.005 | Helped |
| Use only BIM, Brendel & Bethge, and Carlini & Wagner attacks (ablation) | Heuristic attack pool · CIFAR10+strong | R^2: 0.935±0.048 | Mixed |
| Use only BIM, Brendel & Bethge, and Carlini & Wagner attacks (ablation) | Heuristic attack pool · CIFAR10+balanced | R^2: 0.993±0.003 | Helped |
| Drop Fast Gradient Sign Method and uniform noise from the pool (ablation) | Heuristic attack pool · All settings | mean R^2: ≪ 0.001 | No clear effect |
| Increase the number of attacks in the pool (ablation) | Heuristic attack pool · All settings | predictability | Helped |
Closest papers by meaning
- Asymmetric Certified Robustness via Feature-Convex Neural Networks (NeurIPS 2023 · similarity 0.90)
- Towards Deep Learning Models Resistant to Adversarial Attacks (ICLR 2018 · similarity 0.90)
- Theoretically Principled Trade-off between Robustness and Accuracy (ICML 2019 · similarity 0.89)
- Analyzing the Robustness of Nearest Neighbors to Adversarial Examples (ICML 2018 · similarity 0.89)
- On the Alignment between Fairness and Accuracy: from the Perspective of Adversarial Robustness (ICML 2025 · similarity 0.88)
- Provably Cost-Sensitive Adversarial Defense via Randomized Smoothing (ICML 2025 · similarity 0.88)
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.