AI-Based Network Traffic Modeling & Synthetic Data Generation
Year: 2026
Type: conference-paper
Abstract: Realistic network traffic data is a fundamental requirement for the design, evaluation, and optimization of modern communication systems. However, access to real-world traffic traces is often limited due to privacy, security, and operational constraints, particularly in large-scale and critical communication infrastructures. This limitation poses a significant challenge for data-driven network analysis and AI-based optimization methods. This paper presents an AI-based framework for network traffic modeling and synthetic data generation that aims to capture the statistical and temporal characteristics of real communication networks. Network traffic is represented as multivariate time-series data, and generative learning models are employed to produce realistic synthetic traffic patterns under different operational conditions, such as normal load and burst traffic scenarios. The quality of the generated synthetic traffic is evaluated using statistical similarity metrics and temporal correlation analyses. In addition, the usefulness of the synthetic data is assessed through downstream network analysis tasks. Experimental results indicate that the proposed approach can generate high-fidelity traffic patterns suitable for network simulation, performance evaluation, and AI model training, without requiring access to sensitive real traffic data. The proposed framework provides a practical and reproducible solution for AI-driven traffic modeling in modern communication networks.
Author: Ali Berkol
Institution: Aselsan (Turkey)
Language: English
Subfield: Computer Networks and Communications
Field: Computer Science
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7169514331