Causally Constrained Data Synthesis for Private Data Release

TL;DR

This paper introduces a method that integrates causal information into generative models to enhance privacy guarantees and utility in synthetic data generation, addressing privacy-utility trade-offs in data release.

Contribution

It proposes a novel approach that incorporates causal knowledge into generative models, improving privacy guarantees and utility in synthetic data release.

Findings

Causal information improves resilience to membership inference attacks.

The method enhances downstream utility of synthetic data.

Theoretical proof of stronger differential privacy guarantees.

Abstract

Making evidence based decisions requires data. However for real-world applications, the privacy of data is critical. Using synthetic data which reflects certain statistical properties of the original data preserves the privacy of the original data. To this end, prior works utilize differentially private data release mechanisms to provide formal privacy guarantees. However, such mechanisms have unacceptable privacy vs. utility trade-offs. We propose incorporating causal information into the training process to favorably modify the aforementioned trade-off. We theoretically prove that generative models trained with additional causal knowledge provide stronger differential privacy guarantees. Empirically, we evaluate our solution comparing different models based on variational auto-encoders (VAEs), and show that causal information improves resilience to membership inference, with improvements in downstream utility.