计算机与数字工程2025，Vol.53Issue(2)：505-509,5.DOI:10.3969/j.issn.1672-9722.2025.02.036
基于多粒度分层建模的恶意URL检测模型
Malicious URL Detection Model Based on Multi-granularity Hierarchical Modeling
摘要
Abstract
Malicious URL detection is very important for network security protection.To solve the problem of loss of feature in-formation in traditional machine learning and insufficient context modeling of existing deep learning methods,this paper presents a malicious URL detection model based on multi-granularity hierarchical modeling.This method models two feature granularities,character and vocabulary.For each feature granularity,firstly convolution neural network is used to model local context information,attention mechanism is introduced to further model the context information to get the enhanced feature representation,and the com-bination of feature multi-granularity modeling and context hierarchical modeling sufficiently extracts the feature representation of URL for malicious detection.The experimental results show that the accuracy of the model is 98%,which improves the performance of the existing methods.关键词
恶意URL/分层建模/卷积神经网络/注意力机制Key words
malicious URL/hierarchical modeling/convolutional neural network/attention mechanism分类
信息技术与安全科学引用本文复制引用
肖军弼,牟丹..基于多粒度分层建模的恶意URL检测模型[J].计算机与数字工程,2025,53(2):505-509,5.基金项目
中国高校产学研创新基金项目"SD-WAN网络中关键业务通信保障技术研究"(编号:2021FNA02007)资助. （编号:2021FNA02007）
