Research map / Adversarial robustness and security / Certified and empirical robustness
Robust Learning for Data Poisoning Attacks
Yunjuan Wang, Poorya Mianjy, Raman Arora
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 |
|---|---|---|---|
| use negated loss instead of original cross-entropy when generating PGA poisoning attacks | poisoning attack generation (regime A) · MNIST linear model and CIFAR10 AlexNet | clean test accuracy under poisoning | Improved |
| train two-layer ReLU networks with SGD while widening network into a moderate range (not too wide) | two-layer ReLU networks under poisoning · MNIST and CIFAR10 | clean test accuracy vs width | Mixed |
| increase overall perturbation budget S beyond about sqrt(n) (corruption rate above ~250) | SGD on linear model under clean-label poisoning · MNIST linear model | excess error and excess loss vs sample size | Hurt |
| increase corruption budget with sample size (sublinear overall perturbation) | SGD on convex learning problems · theory (Theorem 2.2) | excess risk bound | No clear effect |
Closest papers by meaning
- Not All Poisons are Created Equal: Robust Training against Data Poisoning (ICML 2022 · similarity 0.89)
- Safety-Efficacy Trade Off: Robustness against Data-Poisoning (ICML 2026 · similarity 0.88)
- What Distributions are Robust to Indiscriminate Poisoning Attacks for Linear Learners? (NeurIPS 2023 · similarity 0.88)
- How Robust Are Randomized Smoothing Based Defenses to Data Poisoning? (CVPR 2021 · similarity 0.88)
- Sharpness-Aware Data Poisoning Attack (ICLR 2024 · similarity 0.87)
- Re-thinking Data Availability Attacks Against Deep Neural Networks (CVPR 2024 · 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.
