ETRI-Knowledge Sharing Plaform

KOREAN
논문 검색
Type SCI
Year ~ Keyword

Detail

Journal Article 강화학습 기반 사이버 공격 시뮬레이션 및 에뮬레이션 환경 기술 동향
Cited - time in scopus Download 25 time Share share facebook twitter linkedin kakaostory
Authors
김범석, 문대성, 구기종, 최양서, 유재학
Issue Date
2026-02
Citation
전자통신동향분석, v.41, no.1, pp.82-95
ISSN
1225-6455
Publisher
한국전자통신연구원
Language
Korean
Type
Journal Article
DOI
https://dx.doi.org/10.22648/ETRI.2026.J.410108
Abstract
With the increasing sophistication and automation of cyberattacks, simulation and emulation environments that replicate attack processes in virtual settings have become essential for evaluating defensive strategies. Moreover, research has been conducted on the application of (RL) in cyber ranges to enable autonomous penetration testing. This study provides a comparative analysis of the characteristics and limitations of RL-based cyber-attack simulation environments, including Network Attack Simulation, Cyber Battle Simulation,and the Autonomous Pentesting framework based on Reinforcement Learning, as well as emulation environments such as Cyber Game for Intelligent Learning and Pentesting Gym. In addition, this paper summarizes the characteristics of hybrid environments, including Network Attack Simulation and Emulation and the Generalizable Autonomous Pentesting Framework. In particular, the analysis includes structural differences, Markov decision process design, scalability to large-scale networks, and generalization capability across diverse scenarios. In the future, providing high-fidelity learning environments that closely resemble real networks will require integrated training architectures that combine simulation and emulation, along with expansion toward multi-agent environments.
Keyword
Cyber Attack, Emulation, Generalization, Penetration Testing, Reinforcement Learning, Scalability, Simulation
KSP Keywords
Analysis of the characteristics, Attack Simulation, Battle simulation, Combine simulation, Comparative analysis, Cyber attacks, Generalization capability, High-fidelity, Hybrid environments, Integrated training, Intelligent Learning(iLearning)
This work is distributed under the term of Korea Open Government License (KOGL)
(Type 4: : Type 1 + Commercial Use Prohibition+Change Prohibition)
Type 4: