Network Flow Generation Based on Reinforcement Learning Powered Generative Adversarial Network

Abstract

Mining anomalies and special events from massive network flows based on machine learning and deep learning is a promising approach for network management. However, it is difficult to build labeled network flow data sets for training machine learning and deep learning models. In this paper, we propose a novel reinforcement learning (RL) powered generative adversarial network (GAN) model named NF-GAN for network flow generation. The generator of NF-GAN is designed as a stochastic policy model to generate labeled network flow data. In terms of the discriminator, a check reward is integrated into the network reward to capture the correlations among attributes. Experiment results demonstrate that the majority of the generated flows conform to the strict network protocols of the standard OSI stack, and the success rate of network flows generation achieves 99.96%. To the best of our knowledge, this is the first time of applying RL powered GAN on network flow generation tasks.