基 于 扩 散 模 型 改 进 的 数 字 孪 生 网 络 流 量 生 成 技 术

Generating realistic network traffic for digital twin systems is difficult because traffic packets exhibit temporal dependence，periodic variation，and scenario-driven uncertainty. To address this problem，this paper proposes FlowDiff，a diffusion-based packet generation algorithm，and TDGM，a temporal diffusion generation model. FlowDiff formulates packet generation as a conditioned reverse-diffusion process that progressively denoises latent traffic features. TDGM integrates temporal-aware embedding，periodic features，convolutional neural network（CNN）and Transformer modules，and cross-attention to capture local，global，and long-range traffic patterns. Experiments on real network traffic datasets show that the generated samples are close to real traffic in MSE，KL divergence，FID，and statistical features，demonstrating the effectiveness of the method for digital twin network simulation. Highlights: 1. A large-scale network traffic packet generation algorithm, FlowDiff, is proposed for digital twin scenarios. It formulates packet generation as a reverse-diffusion process conditioned on temporal characteristics and external network scenarios, thereby generating traffic samples closer to real network behavior. 2. A temporal diffusion generation model (TDGM) is developed by introducing temporal-aware feature embedding and Fourier-based periodic feature modeling, enabling the model to capture short-term dependence, long-term trends, and periodic fluctuations in network traffic. 3. A hybrid generative architecture is designed by integrating latent-space compression, local feature extraction, and global context modeling. The combination of VAE, UNet/CNN, Transformer, and cross-attention improves feature representation and conditional generation capability for complex traffic data. 4. Experiments on the real MAWI network traffic dataset evaluate the proposed method using MSE, KL divergence, FID, and statistical feature distance, together with comparison.