Knowledge Embedding Enabled Cyber Security Defense for Networked System: A Novel Risk Detection Method based on Knowledge Graph

Angxiao Zhao, Wenying Feng, Ning Hu, Dawei Zhao, Zhaoquan Gu

Abstract: Knowledge graph (KG) is universally recognized for Knowledge representation. However, valuable information is usually distributed across multi-source heterogeneous data, especially in cyberspace security, which increases the difficulty of mining latent risk in KG. In this paper, we define four kinds of knowledge patterns to form cyberspace security data, and use KG embedding methods to enable the computation and reasoning ability in low-dimensional vector space. Based on our method, we can enable more effective risk detection than common KG embedding methods. We conduct experiments on the Peng Cheng Cyber Range, the experimental results validate the higher prediction accuracy and detection performance. The method can greatly improve the cyberspace security defense capabilities for networked system.

External IDs:dblp:conf/cscwd/Zhao0HZG24