Empirical Evaluation on Synthetic Data Generation with Generative Adversarial Network

Data release has been proven to be impactful in scientific research and business innovation. Nevertheless, the valuable data often contains personal information so that the data release also leads to privacy leakage. Releasing a synthetic data may be a solution for the problem of private data release. In this paper, we consider a generative adversarial networks (GAN)-based synthetic data generation. Furthermore, we perform extensive experiments to evaluate the data utility and risk of re-identification of our GAN-based solution.