Synthetic Packet Traffic Generative Adversarial Networks in Multi Agents With Peer-to-Peer and Global Priority Queue Generation

Network management research requires both theoretical analysis and practical evaluations of network traffic data. Due to security and privacy issues, collecting large network data from the real world is difficult. These data are also extremely imbalanced. To address these problems, synthetic traffic generation is a critical technique. Researchers generate data for network simulations and performance evaluations, also protecting data privacy from real traffic. However, the synthetic traffic generated by modern generators cannot satisfy the network research since the quality is lower than that of realistic traffic. Therefore, we propose Synthetic PAcket Traffic Generative Adversarial Networks (SPATGAN) that can generate high-quality synthetic traffic. The framework includes server agents and client agents, which transmit synthetic packets to each side and take the opponent's synthetic packets as conditional labels for the built-in Timing Synthesis Generative Adversarial Networks (TSynGAN) and a Packet Synthesis Generative Adversarial Networks (PSynGAN) to generate synthetic traffic. Based on the SPATGAN, we further designed two traffic generating algorithms for multiple agents scenarios. To evaluate the quality of the synthetic traffic, we developed a new metric Fréchet Traffic Distance (FTD). The experiments demonstrate that the SPATGAN can generate traffic whose distribution resembles real traffic distribution.