Privacy Preserving Diffusion Models for Mixed-Type Tabular Data Generation

TL;DR

DP-FinDiff is a novel differentially private diffusion framework that synthesizes high-quality mixed-type tabular data by employing embedding-based representations and privacy-aware training strategies, improving utility while maintaining privacy.

Contribution

The paper introduces DP-FinDiff, a new differentially private diffusion model with embedding-based categorical features and two innovative privacy-aware training strategies.

Findings

Achieves 16-42% higher utility than DP baselines on financial and medical datasets.

Effectively scales to high-dimensional mixed-type tabular data.

Maintains strong privacy guarantees while improving data utility.

Abstract

We introduce DP-FinDiff, a differentially private diffusion framework for synthesizing mixed-type tabular data. DP-FinDiff employs embedding-based representations for categorical features, reducing encoding overhead and scaling to high-dimensional datasets. To adapt DP-training to the diffusion process, we propose two privacy-aware training strategies: an adaptive timestep sampler that aligns updates with diffusion dynamics, and a feature-aggregated loss that mitigates clipping-induced bias. Together, these enhancements improve fidelity and downstream utility without weakening privacy guarantees. On financial and medical datasets, DP-FinDiff achieves 16-42% higher utility than DP baselines at comparable privacy levels, demonstrating its promise for safe and effective data sharing in sensitive domains.