华侨大学学报（自然科学版）2025，Vol.46Issue(5)：528-538,11.DOI:10.11830/ISSN.1000-5013.202507014
基于图结构表示增强与特征融合的网络威胁检测方法
Network Threat Detection Method Based on Graph Structure Representation Enhancement and Feature Fusion
摘要
Abstract
To address the security threats from malicious network activities in financial projects involving real-world assets(RWA)within blockchain and internet of things applications,a network threat detection method based on graph structure representation enhancement and feature fusion is proposed.First,traffic records are modeled as event-driven heterogeneous graphs.Second,a post-event-level contrastive enhancement mechanism is designed to decouple the representation learning and enhancement processes,improving the discriminative power of node representations.Finally,a multi-source feature fusion scheme is introduced to optimize input representation through dimensionality reduction and feature selection.The experimental results show that the proposed method achieves an area under the receiver operating characteristic curve of 0.975,an area under the precision-recall curve of 0.966,and an F1-score of 0.928 on the CIC-Darknet dataset,thereby effectively en-hancing the detection capability of obnormal network activities.关键词
网络检测/图结构/表示学习/机器学习/对比增强/特征融合Key words
network detection/graph structure/representation learning/machine learning/contrast enhance-ment/feature fusion分类
计算机与自动化引用本文复制引用
游欣源,郭荣新,施一帆..基于图结构表示增强与特征融合的网络威胁检测方法[J].华侨大学学报（自然科学版）,2025,46(5):528-538,11.基金项目
国家自然科学青年基金资助项目(62306122) （62306122）
福建省科技项目引导性项目(2023H0012) （2023H0012）
