An IoT Attack Detection Framework Leveraging Graph Neural Networks

Onderzoeksoutput: Hoofdstuk in Boek/Rapport/CongresprocedureHoofdstukAcademicpeer review

Samenvatting

We propose an attack detection framework for Internet of Things (IoT) networks, which leverages Graph Neural Networks (GNN) to capture the inherent structure of IoT network data. Specifically, we utilize GraphSAGE and propose a framework to detect network intrusions by capturing the graph’s edge features and data flow information for IoT networks. To evaluate the effectiveness of our approach, we use the Kitsune and BoT-IoT datasets that include botnet, Man-in-the-Middle (MiTM), Reconnaissance, Denial of Service (DoS), Distributed Denial of Service (DDoS), and information theft attacks. To reduce time complexity and analyze the significance of removing extraneous features, we conduct feature selection experiments also. Our study highlights the effectiveness of GNN-based attack detection for IoT security with 89.3% accuracy for kitsune and 88.6% accuracy for BoT-IoT and underscores the importance of unbiased cross-validation to ensure model performance.
Originele taal-2Engels
TitelIntelligence of Things: Technologies and Applications
SubtitelThe Second International Conference on Intelligence of Things (ICIT 2023), Ho Chi Minh City, Vietnam, October 25-27, 2023, Proceedings
RedacteurenNhu-Ngoc Dao, Tran Ngoc Thinh, Ngoc Thanh Nguyen
UitgeverijSpringer
Pagina's225-236
Aantal pagina's12
Volume2
DOI's
StatusGepubliceerd - 20 okt. 2023
EvenementSecond International Conference on Intelligence of Things (ICIT 2023) - Ho Chi Minh City, Vietnam
Duur: 25 okt. 202327 okt. 2023

Publicatie series

NaamLecture Notes on Data Engineering and Communications Technologies
Volume188
ISSN van geprinte versie2367-4512
ISSN van elektronische versie2367-4520

Congres

CongresSecond International Conference on Intelligence of Things (ICIT 2023)
Land/RegioVietnam
StadHo Chi Minh City
Periode25/10/2327/10/23

Financiering

Part of this work was funded by the Dutch Research Council (NWO) in the context of its commitment to the Dutch Research Agenda (NWA) as part of the INTERSCT research program funded under grant NWA.1160.18.301.

FinanciersFinanciernummer
Dutch Research AgendaNWA.1160.18.301
Nederlandse Organisatie voor Wetenschappelijk Onderzoek

    Vingerafdruk

    Duik in de onderzoeksthema's van 'An IoT Attack Detection Framework Leveraging Graph Neural Networks'. Samen vormen ze een unieke vingerafdruk.

    Citeer dit