Research map / Adversarial robustness and security / Certified and empirical robustness
Finding Needles in a Moving Haystack: Prioritizing Alerts with Adversarial Reinforcement Learning
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 stochastic, state-dependent alert prioritization instead of Uniform allocation | Alert prioritization · CICIDS2017 intrusion detection | defender loss: 50% | Helped |
| Use ARL instead of Suricata's deterministic priorities | Alert prioritization · CICIDS2017 intrusion detection | defender loss | Helped |
| Use ARL when the defender's estimate of the attack budget is inaccurate | ARL · CICIDS2017 intrusion detection | defender loss: 5% | Hurt |
| Use ARL rather than Uniform, GAIN, or RIO | Alert prioritization · Fraud detection | defender loss: at least 25% | Helped |
| Use ARL when the defender's estimate of the attack budget is inaccurate | ARL · Fraud detection | defender loss: <10% | Hurt |
| Slightly underestimate rather than overestimate the attacker's budget | ARL · Fraud detection | defender loss: less than 3% | Helped |
| Use RIO rather than ARL against a Greedy attacker | Alert prioritization · Fraud detection, small defender budget | defender loss: 8% | Helped |
| Use RIO rather than ARL against a Greedy attacker | Alert prioritization · Fraud detection, large defender budget | defender loss: 13% | Helped |
| Account for adaptive attacks across time rather than optimizing each period independently | Alert prioritization · Fraud detection | defender loss | Helped |
Closest papers by meaning
- Optimal Attack and Defense for Reinforcement Learning (AAAI 2024 · similarity 0.87)
- Illusory Attacks: Information-theoretic detectability matters in adversarial attacks (ICLR 2024 · similarity 0.85)
- Adversaries With Incentives: A Strategic Alternative to Adversarial Robustness (ICLR 2025 · similarity 0.85)
- Rethinking Adversarial Policies: A Generalized Attack Formulation and Provable Defense in RL (ICLR 2024 · similarity 0.84)
- BadRL: Sparse Targeted Backdoor Attack against Reinforcement Learning (AAAI 2024 · similarity 0.84)
- Quickest Detection of Adversarial Attacks Against Correlated Equilibria (AAAI 2025 · similarity 0.84)
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.