Yassen, Fahad (2026) A Hybrid Graph Neural Network Framework for Cyberattack Detection Based on Network Traffic Graphs. CENTRAL ASIAN JOURNAL OF MATHEMATICAL THEORY AND COMPUTER SCIENCES, 7 (3). ISSN 2660-5309

Text CAJMTCS_954_A+Hybrid+Graph+Neural.pdf Download (534kB)

Abstract

The increasing complexity of modern communication networks and the growing diversity of cyberattacks have exposed the limitations of traditional intrusion detection systems that treat network traffic records as independent instances. Although graph neural networks (GNNs) have shown promising performance in cybersecurity applications, most existing approaches rely on a single graph learning architecture and fixed graph construction strategies. To address these limitations, this paper proposes HGCF-Net, a Hybrid Graph Neural Network framework that integrates Graph Convolutional Networks (GCNs) and Graph Attention Networks (GATs) through an adaptive feature fusion mechanism for cyberattack detection. The proposed framework transforms network traffic flows into graph-structured representations to capture both global communication patterns and local neighborhood dependencies. Experiments conducted on the CSE-CIC-IDS2018 dataset demonstrate that the proposed model achieved an Accuracy of 99.28%, an F1-score of 99.09%, and an AUC of 99.63%, outperforming several recent graph-based intrusion detection methods. The results confirm the effectiveness and robustness of HGCF-Net for intelligent cyberattack detection in modern network environments.

Item Type:	Article
Uncontrolled Keywords:	Graph Neural Networks, Cyberattack Detection, Graph Convolutional Network, Intrusion Detection System, Network Traffic Analysis.
Subjects:	H Social Sciences
Depositing User:	admin eprints
Date Deposited:	01 Aug 2026 09:23
Last Modified:	01 Aug 2026 09:23
URI:	http://eprints.umsida.ac.id/id/eprint/16881

Actions (login required)

View Item