融合LSTM、GNN和贝叶斯网络的网络安全态势评估与预测  

Network Security Situation Assessment and Prediction Based on LSTM,GNN and Bayesian Network

作　　者：魏巍 许丰宽[1] 毛思琪 WEI Wei;XU Fengkuan;MAO siqi(Hulunbuir University,Hailar,Inner Mongolia 021008;Dalian University of Technology,Dalian,Liaoning 116000)

机构地区：[1]呼伦贝尔学院,内蒙古呼伦贝尔021008 [2]大连理工大学,辽宁大连116000

出　　处：《呼伦贝尔学院学报》2025年第1期125-131,共7页Journal of Hulunbuir University

基　　金：2023年学院校级委托项目“基于高校校园网络安全设计探究——以呼伦贝尔学院科研系统网络安全防护为例”(2023XJWT08);2024年呼伦尔学院校级委托项目“呼伦贝尔市旅游安全与应急响应系统的AI驱动研究”(2024XJCG35)。

摘　　要：本文研究了基于多源数据分析的网络安全整体态势评估系统的结构组成与评估技术,并通过实验验证其应用效果。该系统通过多源数据获取日志信息、节点漏洞信息和节点服务信息,从获取的信息数据中提取脆弱性、威胁性和资产三种态势指标,使用长短期记忆网络(LSTM)、图神经网络(GNN)和贝叶斯网络对态势指标进行融合处理进行实验,评估网络安全整体态势。实验结果表明,未来6个月内网络整体安全态势“较为危险”(Quite Dangerous)的概率逐渐增加,从当前的0.25上升到0.40。这表明随着时间的推移,网络的安全态势可能恶化,风险增加。This paper introduces a holistic system for assessing network security posture by utilizing multi-source data analysis,and its effectiveness is validated through simulation experiments.The system gathers log information,node vulnerability details,and node service information through a multi-source data layer,from which three posture indicators—vulnerability,threat,and asset—are extracted.To integrate these indicators and evaluate the overall network security posture,Long Short-Term Memory(LSTM)networks,Graph Neural Networks(GNNs),and Bayesian networks are utilized.The results from the simulation indicate that the likelihood of the network security posture being categorized as“Quite Dangerous”will incrementally rise over the next six months,increasing the current probability from0.25 to 0.40.This implies that over time,the security posture of the network may worsen,resulting in an elevated risk.

关 键 词：多源数据分析 网络安全态势评估 LSTM GNN 贝叶斯网络 

分 类 号：TP393.08[自动化与计算机技术—计算机应用技术]