microsoft / CyberBattleSim

An experimentation and research platform to investigate the interaction of automated agents in an abstract simulated network environments.
MIT License
1.63k stars 257 forks source link

Consider designing defender as RL agent #96

Open Screamer-Y opened 1 year ago

Screamer-Y commented 1 year ago

I'm curious if this project will follow up by designing defender as an agent that can be trained with reinforcement learning algorithms and how we will achieve this based on the existing environment? Thanks in advance.

forrestmckee commented 1 year ago

@Screamer-Y I'm hopeful for the same. this paper introduces an RL based defender agent that leverages a knowledge graph. However, I don't think the code is released.

2twentytwo2 commented 1 year ago

您好,您的邮件我已收到, 看到后,我会立即回复您。 This is an automatic reply, confirming that your e-mail was received.Thank you

Screamer-Y commented 1 year ago

@forrestmckee Thank you so much for sharing this informative paper! I will check it out later.