Em Direção a uma Interface de Visualização e Aprendizado para o CybORG: um Ambiente para Agentes Cibernéticos Autônomos

• Gabriel Luiz de Souza Siqueira UFMG
• Bruno P. dos Santos UFBA
• Daniel F. Macedo UFMG
• José Marcos S. Nogueira UFMG
• Paulo H. L. Rettore UFLA

DOI: https://doi.org/10.5753/webmedia_estendido.2026.34214

Resumo

Cybersecurity training environments such as CybORG generate large amounts of simulation data, making experiment analysis difficult without dedicated visualization support. This work presents a lightweight visualization interface designed to support the inspection and interpretation of CybORG evaluation sessions, particularly in the context of the CAGE Challenge 4. Developed as part of the SiR.AI project, the proposed solution provides interactive visual components for analyzing rewards, network topology, and agent behavior without the complexity of a dedicated backend infrastructure. The tool facilitates the training and evaluation of Reinforcement Learning (RL), bridging the gap between theoretical concepts and practical application for beginner researchers and students.

Palavras-chave: CybORG, Reinforcement Learning, Cybersecurity, Visualization Interface, CAGE Challenge 4

Referências

Callum Baillie, Maxwell Standen, Jonathon Schwartz, Michael Docking, David Bowman, and Junae Kim. 2020. CybORG: An Autonomous Cyber Operations Research Gym. arXiv:2002.10667 [cs.CR] [link]

Harry Emerson, Liz Bates, Chris Hicks, and Vasilios Mavroudis. 2024. CybORG++: An Enhanced Gym for the Development of Autonomous Cyber Agents. arXiv:2410.16324 [cs.CR] [link]

Mitchell Kiely, David Bowman, Maxwell Standen, and Christopher Moir. 2023. On Autonomous Agents in a Cyber Defence Environment. arXiv:2309.07388 [cs.CR] [link]

Mitchell et al. Kiely. 2025. CAGE challenge 4: A scalable multi-agent reinforcement learning gym for autonomous cyber defence. AI Magazine 46, 3 (2025), e70021. DOI: 10.1002/aaai.70021 [link]

Mitchell et al. Kiely. 2025. Exploring the efficacy of multi-agent reinforcement learning for autonomous cyber defence: a CAGE challenge 4 perspective. In Proceedings of the Thirty-Ninth AAAI Conference on Artificial Intelligence and Thirty-Seventh Conference on Innovative Applications of Artificial Intelligence and Fifteenth Symposium on Educational Advances in Artificial Intelligence (AAAI’25/IAAI’25/EAAI’25). AAAI Press, Article 3260, 7 pages. DOI: 10.1609/aaai.v39i28.35158

Li Li, Raed Fayad, and Adrian Taylor. 2021. CyGIL: A Cyber Gym for Training Autonomous Agents over Emulated Network Systems. arXiv:2109.03331 [cs.CR] [link]

Abby Morris, Rachael Procter, and Caroline Wallbank. 2025. Evaluating Reinforcement Learning Agents for Autonomous Cyber Defence. Applied AI Letters 6, 3 (June 2025), 15 pages. DOI: 10.1002/ail2.125

Sean Oesch, Amul Chaulagain, Brian Weber, Matthew Dixson, Amir Sadovnik, Benjamin Roberson, Cory Watson, and Phillipe Austria. 2024. Towards a High Fidelity Training Environment for Autonomous Cyber Defense Agents. In Proceedings of the 17th Cyber Security Experimentation and Test Workshop (Philadelphia, PA, USA) (CSET ’24). Association for Computing Machinery, New York, NY, USA, 91–99. DOI: 10.1145/3675741.3675752

Maxwell Standen, Martin Lucas, David Bowman, Toby J. Richer, Junae Kim, and Damian Marriott. 2021. CybORG: A Gym for the Development of Autonomous Cyber Agents. arXiv:2108.09118 [cs.CR] [link]

Microsoft Defender Research Team. 2021. CyberBattleSim. [link]. Created by Christian Seifert, Michael Betser, William Blum, James Bono, Kate Farris, Emily Goren, Justin Grana, Kristian Holsheimer, Brandon Marken, Joshua Neil, Nicole Nichols, Jugal Parikh, Haoran Wei..

Melody Wolk, Andy Applebaum, Camron Dennler, Patrick Dwyer, Marina Moskowitz, Harold Nguyen, Nicole Nichols, Nicole Park, Paul Rachwalski, Frank Rau, and Adrian Webster. 2022. Beyond CAGE: Investigating Generalization of Learned Autonomous Network Defense Policies. arXiv:2211.15557 [cs.LG] [link]