一种基于网络防御智能体的决策方法
By constructing a network system topology graph and generating agents using sigma rules, deploying attacking and defending agents, and utilizing reinforcement learning training and state encoders for action decisions, the computational complexity and instability issues in multi-agent reinforcement learning are resolved, thereby improving the adaptability and learning stability of the agents.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- JIANGSU RUINING XINCHUANG TECH CO LTD
- Filing Date
- 2025-07-18
- Publication Date
- 2026-07-17
AI Technical Summary
In complex network environments, the computational complexity of single-agent reinforcement learning increases with the environmental state and action space size, leading to impractical learning. In multi-agent reinforcement learning, agents cannot utilize information from other agents, resulting in an unstable learning process with poor convergence.
Construct a network system topology map, generate agents based on sigma rules, match network threats based on threat intelligence, deploy attack and defense agents, train agents through reinforcement learning, make action decisions using agent state encoders and evaluation networks, optimize agent strategies, and conduct joint evaluation using the local environmental states of other agents.
This technology enables agents to adapt to changes in the environment and other agents' policies during multi-agent reinforcement learning, improving learning stability, reducing computational costs, and effectively responding to dynamic network attacks.
Smart Images
Figure CN120639464B_ABST