BROWSE
Detail
Conference Paper
A Study on Semi-Supervised Intrusion Detection for Private 5G Networks Using a Testbed-Generated Flow Dataset
Cited - time in 
Share 
Share
- Authors
- HaLin Jeon, Jonghoon Lee, DaeSung Moon
- Issue Date
- 2026-08
- Citation
- World Conference on Information Security Applications (WISA) 2026, pp.1-4
- Language
- English
- Type
- Conference Paper
- Abstract
- Intrusion detection for Private 5G (P5G) security needs datasets built from traffic on the 5G data plane. We build a P5G testbed with open-source software, collect benign and attack traffic against enterprise services behind the 5G core, turn it into flow records, and evaluate a deep-learning autoencoder for semi-supervised intrusion detection. A compact set of twenty-two behavioral features performs slightly better than the full forty-seven-feature set, achieving an AUROC of approximately 0.92. This paper presents preliminary results of ongoing work on AI-based P5G security.
- Keyword
- Private 5G, Intrusion detection, Flow-based analysis, Anomaly detection, Semi-supervised learning
- KSP Keywords
- 5G networks, Compact set, Data plane, Feature set, Flow Records, Flow-based(FB), Intrusion Detection, Open Source Software(OSS), Semi-Supervised Learning(SSL), anomaly detection, behavioral features