BROWSE
Detail
Conference Paper
Diffusion Model-Based Network Traffic Synthesis for Private 5G Network
Cited - time in 
Share 
Share
- Authors
- Noh-Sam Park, Mikyong Han, Jonghoon Lee
- Issue Date
- 2026-07
- Citation
- International Conference on Ubiquitous and Future Networks (ICUFN) 2026, pp.1-3
- Publisher
- IEEE
- Language
- English
- Type
- Conference Paper
- Abstract
- Private 5G (P5G) networks introduce heterogeneous traffic patterns and novel attack vectors that demand intelli- gent intrusion detection. Self-Supervised Learning (SSL)-based Intrusion Detection Systems (IDS) address the critical bottleneck of labeled data scarcity by learning latent representations from unlabeled traffic, yet their performance remains tightly coupled to the quality of data augmentation (DA). This paper proposes a Diffusion model-based network traffic synthesis framework as a DA component for SSL-based P5G IDS. Network flow records are encoded as grayscale images and processed by a diffusion model to generate class-conditioned synthetic samples. We further analyze the role of Large Language Models (LLMs) as a complementary synthesis paradigm capable of encoding protocol semantics and domain knowledge. Experiments on CIC-IDS 2017 validate high synthesis fidelity and demonstrate that synthetic data can support downstream IDS classifiers with high detection performance.
- Keyword
- Private 5G, Intrusion Detection, Diffusion Model, Network Traffic Synthesis
- KSP Keywords
- 5G networks, Data Augmentation, Data scarcity, Grayscale Images, Intrusion Detection Systems(IDS), Labeled data, Language Models, Latent representations, Network Traffic, Network flow, Quality of data