Research map / Adversarial robustness and security / Certified and empirical robustness
Online Adversarial Attacks
Andjela Mladenovic, Joey Bose, Hugo Berard, William L. Hamilton, Simon Lacoste-Julien, Pascal Vincent, Gauthier Gidel
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 an online selection algorithm instead of random selection (average aggregate improvement) | Online adversarial attacks · MNIST and CIFAR-10 | online fool rate: 7.5% and 34.1% | Improved |
| Use Virtual+ instead of Virtual or Optimistic | Online adversarial attacks · CIFAR-10, non-robust and robust classifiers | online fool rate | Improved |
| Use Virtual+ instead of Virtual or Optimistic | Online adversarial attacks · Robust classifiers, except MNIST at k=10 | online fool rate | Improved |
| Use Virtual+ instead of Single-Ref | Online adversarial attacks · MNIST, k=1000 | online fool rate: 15.3% | Improved |
| Use an online algorithm rather than Naive random selection | Online adversarial attacks · ImageNet non-robust classifiers | online fool rate | Improved |
| Use Virtual+ instead of the strongest single-threshold baselines | k-secretary algorithms · k=2, 3, and 4 | competitive ratio: C_2≥0.427, C_3≥ .457, C_4≥.4769 | Improved |
| Use Single-Ref instead of Virtual+ | k-secretary algorithms · 5≤k≤100 | competitive ratio | Hurt |
| Use FGSM rather than PGD with an online selection algorithm | Online adversarial attacks · Online blackbox transfer attacks | online fool rate | Mixed |
Closest papers by meaning
- Efficient Adversarial Attacks on High-dimensional Offline Bandits (ICLR 2026 · similarity 0.88)
- Prior Convictions: Black-box Adversarial Attacks with Bandits and Priors (ICLR 2019 · similarity 0.87)
- PubDef: Defending Against Transfer Attacks From Public Models (ICLR 2024 · similarity 0.87)
- Practical Adversarial Attacks on Stochastic Bandits via Fake Data Injection (NeurIPS 2026 · similarity 0.87)
- AttackBench: Evaluating Gradient-based Attacks for Adversarial Examples (AAAI 2025 · similarity 0.86)
- SuperDeepFool: a new fast and accurate minimal adversarial attack (NeurIPS 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.
