信息工程大学学报2026，Vol.27Issue(1)：72-80,9.DOI:10.3969/j.issn.1671-0673.2026.01.010
基于多门共享—私有混合专家的漏洞CVSS度量预测系统
A CVSS Metrics Prediction Model of Vulnerability Based on Multi-Gate Shared and Private Mixture of Experts
摘要
Abstract
To address the subjectivity and inefficiency of manual common vulnerability scoring system(CVSS)assessment and the limitations of existing automated methods in capturing inter-metric relation-ships,a CVSS metrics prediction model of vulnerability based on a multi-gate shared and private mix-ture of experts(MSPMoE)is proposed.In the model,a shared-private expert mechanism is used to em-ployed enyto jointly leverage shared and unique features across different metrics.Firstly,the Distil-BERT model is fine-tuned on a collected vulnerability description corpus as the feature encoder.Then,a dynamic feature extraction module is designed to adaptively select optimal feature representa-tion strategies.Finally,the MSPMoE classifier is adopted,where shared experts are used to capture common patterns across metrics,while private experts are used to extract metric-specific features.A gating network is used to dynamically weight each expert's contribution,to balance shared and private knowledge for accurate CVSS metrics prediction.Experiments show that the proposed model outper-forms state-of-the-art methods on seven CVSS metrics,achieving an average accuracy of 83.32%across all eight metrics.关键词
CVSS度量预测/漏洞评估/多任务学习/自然语言处理/大模型Key words
CVSS metrics prediction/vulnerability assessment/multi-task learning/natural lan-guage processing/large language models分类
信息技术与安全科学引用本文复制引用
王晓龙,杜晔,郑天帅,陈奇芳,关昌昊..基于多门共享—私有混合专家的漏洞CVSS度量预测系统[J].信息工程大学学报,2026,27(1):72-80,9.基金项目
国家重点研发计划(2022YFB3105105) （2022YFB3105105）
北京市自然科学基金(L254063) （L254063）
