Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Tabular Data Synthesis with Differential Privacy: A Survey

Mengmeng Yang, Chi-Hung Chi, Kwok-Yan Lam, Jie Feng, Taolin Guo, Wei Ni

TL;DR

This paper provides a comprehensive overview of existing differentially private tabular data synthesis methods, highlighting the unique challenges of each generation model for generating tabular data under differential privacy constraints and classify them into statistical and deep learning-based approaches based on their generation models.

Abstract

Data sharing is a prerequisite for collaborative innovation, enabling organizations to leverage diverse datasets for deeper insights. In real-world applications like FinTech and Smart Manufacturing, transactional data, often in tabular form, are generated and analyzed for insight generation. However, such datasets typically contain sensitive personal/business information, raising privacy concerns and regulatory risks. Data synthesis tackles this by generating artificial datasets that preserve the statistical characteristics of real data, removing direct links to individuals. However, attackers can still infer sensitive information using background knowledge. Differential privacy offers a solution by providing provable and quantifiable privacy protection. Consequently, differentially private data synthesis has emerged as a promising approach to privacy-aware data sharing. This paper provides a comprehensive overview of existing differentially private tabular data synthesis methods, highlighting the unique challenges of each generation model for generating tabular data under differential privacy constraints. We classify the methods into statistical and deep learning-based approaches based on their generation models, discussing them in both centralized and distributed environments. We evaluate and compare those methods within each category, highlighting their strengths and weaknesses in terms of utility, privacy, and computational complexity. Additionally, we present and discuss various evaluation methods for assessing the quality of the synthesized data, identify research gaps in the field and directions for future research.

Tabular Data Synthesis with Differential Privacy: A Survey

TL;DR

This paper provides a comprehensive overview of existing differentially private tabular data synthesis methods, highlighting the unique challenges of each generation model for generating tabular data under differential privacy constraints and classify them into statistical and deep learning-based approaches based on their generation models.

Abstract

Data sharing is a prerequisite for collaborative innovation, enabling organizations to leverage diverse datasets for deeper insights. In real-world applications like FinTech and Smart Manufacturing, transactional data, often in tabular form, are generated and analyzed for insight generation. However, such datasets typically contain sensitive personal/business information, raising privacy concerns and regulatory risks. Data synthesis tackles this by generating artificial datasets that preserve the statistical characteristics of real data, removing direct links to individuals. However, attackers can still infer sensitive information using background knowledge. Differential privacy offers a solution by providing provable and quantifiable privacy protection. Consequently, differentially private data synthesis has emerged as a promising approach to privacy-aware data sharing. This paper provides a comprehensive overview of existing differentially private tabular data synthesis methods, highlighting the unique challenges of each generation model for generating tabular data under differential privacy constraints. We classify the methods into statistical and deep learning-based approaches based on their generation models, discussing them in both centralized and distributed environments. We evaluate and compare those methods within each category, highlighting their strengths and weaknesses in terms of utility, privacy, and computational complexity. Additionally, we present and discuss various evaluation methods for assessing the quality of the synthesized data, identify research gaps in the field and directions for future research.

Paper Structure

This paper contains 51 sections, 8 equations, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Background knowledge
  3. Tabular data synthesis
  4. Concepts.
  5. Challenges.
  6. Privacy disclosure in synthetic data.
  7. Differential privacy
  8. Definition of differential privacy
  9. Differential privacy mechanisms
  10. Composition properties
  11. Relaxations of differential privacy
  12. Centralized data synthesis with DP
  13. Statistical method
  14. Copula
  15. Probabilistic graphical models
  16. ...and 36 more sections

Figures (3)

  • Figure 1: An example of the creation of the junction tree (The figure is modified from chen2015differentially)
  • Figure 2: Architecture of Distributed GAN. The left figure shows the multi-discriminator structure, where the server maintains a global generator and each client possesses a discriminator. The right figure shows the federated learning structure, where the server initializes both a global generator and discriminator and each client trains its local generator and discriminator on its dataset.
  • Figure 3: The taxonomy tree of the synthetic data evaluation methods

Theorems & Definitions (2)

  • Definition 1: $(\alpha,\epsilon)$-RDP mironov2017renyi
  • Definition 2: Zero-Concentrated Differential Privacy (zCDP) bun2016concentrated