Research map / Bandits and online learning / Multi-armed bandits
Adversarial Attacks on Stochastic Bandits
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 |
|---|---|---|---|
| Manipulate rewards of non-target arms online, without knowing their means | ε-greedy · Stochastic bandit; ε_t=1/t in the simulation | target-arm pulls: 6 -> 9994 | Helped |
| Manipulate rewards of non-target arms online, without knowing their means | UCB · Two-arm stochastic bandit, T=10^7 | target-arm pulls: 156 -> 10^7-2 | Helped |
| Increase the UCB attack parameter Δ_0 (ablation) | UCB attack · Two-arm stochastic bandit | cumulative attack cost | Helped |
| Use a constant attack A greater than the maximum reward gap | Heuristic constant attack on ε-greedy and UCB · Two-arm bandit with maximum gap 1 | target-arm pulls | Helped |
| Use a constant attack A smaller than the maximum reward gap | Heuristic constant attack on ε-greedy and UCB · Two-arm bandit with maximum gap 1 | attack success | Hurt |
| Increase the non-target reward gap Δ_1 (ablation) | ε-greedy attack · Two-arm Gaussian bandit simulation | cumulative attack cost | Hurt |
Closest papers by meaning
- Stochastic Bandits Robust to Adversarial Attacks (ICLR 2025 · similarity 0.93)
- Adversarial Attacks on Adversarial Bandits (ICLR 2023 · similarity 0.93)
- Data Poisoning Attacks on Stochastic Bandits (ICML 2019 · similarity 0.93)
- Practical Adversarial Attacks on Stochastic Bandits via Fake Data Injection (NeurIPS 2026 · similarity 0.92)
- Stealthy Adversarial Attacks on Stochastic Multi-Armed Bandits (AAAI 2024 · similarity 0.92)
- Sparsity-Agnostic Linear Bandits with Adaptive Adversaries (NeurIPS 2024 · 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.