计算机与现代化Issue(4)：88-94,7.DOI:10.3969/j.issn.1006-2475.2026.04.012
融合局部与全局特征的网络安全实体识别方法
A Network Security Entity Recognition Method Integrating Local and Global Features
摘要
Abstract
With the rapid development of information technology,network security entity recognition plays an important role in threat intelligence analysis and intrusion detection.Traditional entity recognition methods have significant limitations in the field of network security,especially when dealing with complex contextual dependencies and multi-level feature information.To solve these problems,this paper proposes a network security entity recognition method based on the fusion of local and global features.This method uses the BERT model to extract word-level features and captures global features through the BiLSTM/BiGRU layer to capture long-distance contextual dependencies.In order to further improve the model's ability to model local dependencies,a multi-head attention mechanism is introduced to extract local features and enhance the understanding of short-distance relation-ships between words.Finally,the local features and global features are fused by weighted summation,and sequence annotation is performed through the conditional random field layer to improve the accuracy of entity recognition.This paper conducts a weight distribution experiment on the DNRTI dataset to determine the optimal fusion ratio.The effectiveness of this method on the BERT-BiLSTM-MHAF-CRF and BERT-BiGRU-MHAF-CRF models is verified through comparative experiments.At the same time,the comparative experiment with the large model ChatGLM-6B further verifies the advantages of the proposed model in the task of network security entity recognition.关键词
实体识别/网络安全/注意力机制/特征融合/BERTKey words
entity recognition/network security/attention mechanism/feature fusion/BERT分类
信息技术与安全科学引用本文复制引用
付鲁森,年梅,张俊,吕美静..融合局部与全局特征的网络安全实体识别方法[J].计算机与现代化,2026,(4):88-94,7.基金项目
新疆维吾尔自治区自然科学基金资助项目(2023D01A46) （2023D01A46）
国家重点研发计划项目子课题(E1182101) （E1182101）
