ATLAS: GAN-Based Differentially Private Multi-Party Data Sharing

In this article, we study the problem of differentially private multi-party data sharing, where the involved parties assisted by a semi-honest curator collectively generate a shared dataset while satisfying differential privacy. Inspired by the success of data synthesis with the generative adversarial network (GAN), we propose a novel GAN-based differentially private multi-party data sharing approach named ATLAS. In ATLAS, we extend the original GAN to multiple discriminators, and let each party hold a discriminator while the curator holds a generator. To update the generator without compromising each party's privacy, we decompose the calculation of the generator's gradient and selectively sanitize the discriminators' responses. Additionally, we propose two methods to improve the utility of shared data, i.e., the collaborative discriminator filtering (CDF) method and the adaptive gradient perturbation (AGP) method. Specifically, the CDF method utilizes trained discriminators to refine synthetic records, while the AGP method adaptively adjusts the noise scale during training to reduce the impact of deferentially private noise on the final shared data. Extensive experiments on real-world datasets validate the superiority of our ATLAS approach.

Automated summary

In this article, the authors propose a novel GAN-based approach named ATLAS for differentially private multi-party data sharing. The ATLAS approach extends the original GAN to multiple discriminators, with each party holding a discriminator and the curator holding a generator. The authors decompose the calculation of the generator's gradient and selectively sanitize the discriminators' responses to update the generator without compromising privacy. They also propose two methods, CDF and AGP, to improve the utility of shared data by refining synthetic records and adjusting the noise scale during training. Extensive experiments validate the superiority of the ATLAS approach.