Abstract
Security operations centers (SOCs) are increasingly turning to machine learning to help analysts manage the growing volume of alerts. Although many of these systems use static classifiers, they often lack the ability to adapt to changing threat contexts or analyst feedback. Reinforcement learning (RL) offers a way to train agents that can learn from experience and make context-aware decisions. In this work, we investigate how the reward function, rather than the algorithm itself, can influence agent behavior in simulated SOC environments.We introduce a risk-aware reward formulation that encodes operational factors such as threat criticality, confidence, and isolation cost. We then evaluate whether a standard RL agent (PPO) can learn more useful behavior purely through reward shaping. Our experiments show that reward design plays a critical role: agents trained with shaped rewards not only achieve better performance, but also demonstrate more adaptive and risk-sensitive behavior, especially under uncertainty. These findings highlight the importance of aligning reward signals with real-world decision trade-offs in practical security settings.
| Original language | English |
|---|---|
| Title of host publication | Computer Security. ESORICS 2025 International Workshops : AutonomousCyber 2025, CPS4CIP 2025, DisA 2025, HS3 2025, MIST 2025, Toulouse, France, September 25–26, 2025, Revised Selected Papers, Part III |
| Editors | Romain Laborde, Joaquin Garcia-Alfaro, Abbas Yazdinejad, Gregory Epiphaniou, Habtamu Abie, Silvio Ranise, Michal Choras, Michal Wozniak, Yuko Hara, Jan Tobias Mühlberg, Claudia Greco, Kim-Kwang Raymond Choo |
| Number of pages | 13 |
| Publisher | Springer |
| Publication date | Apr 2026 |
| Pages | 92-104 |
| ISBN (Print) | 978-3-032-16164-2 |
| ISBN (Electronic) | 978-3-032-16165-9 |
| DOIs | |
| Publication status | Published - Apr 2026 |
| Event | 30th European Symposium on Research in Computer Security, ESORICS 2025 - Toulouse, France Duration: 25 Sept 2025 → 26 Sept 2025 Conference number: 30 https://conf.laas.fr/esorics/ |
Conference
| Conference | 30th European Symposium on Research in Computer Security, ESORICS 2025 |
|---|---|
| Number | 30 |
| Country/Territory | France |
| City | Toulouse |
| Period | 25/09/2025 → 26/09/2025 |
| Internet address |
| Series | Lecture Notes in Computer Science (LNCS) |
|---|---|
| Volume | 16233 |
| ISSN | 0302-9743 |
Keywords
- Cyber-attack
- Cyber-physical system
- Key (lock)
- Reinforcement
- Reinforcement learning
- Risk-Aware Reinforcement Learning
- Operational Security
- SOC Alert Handling
- Autonomous Cyber Defense
Fingerprint
Dive into the research topics of 'Risk-Aware SOC Alert Handling in Adaptive Cyber Defense with Reinforcement Learning'. Together they form a unique fingerprint.Cite this
- APA
- Author
- BIBTEX
- Harvard
- Standard
- RIS
- Vancouver