Synthetic Tabular Data: Methods, Attacks and Defenses

Lecture-Style Tutorial @ KDD 2025 | August 3, 1:00-4:00 PM (EDT) | Toronto, CA

Tutorial @ VLDB 2025 | TBD Date & Time | London, UK

Abstract

Synthetic data is often positioned as a solution to replace sensitive fixed-size data sets with a source of unlimited matching data, freed from privacy concerns. There has been much progress in synthetic data generation over the last decade, leveraging corresponding advances in machine learning and data analytics. In this tutorial, we survey the key developments and the main concepts in tabular synthetic data generation, including paradigms based on probabilistic graphical models and on deep learning. We provide background and motivation, before giving a technical deep-dive into the methodologies. We also address the limitations of synthetic data, by studying attacks that seek to retrieve information about the original sensitive data. Finally, we present extensions and open problems in this area

Schedule

We will have a 15 min coffee break and two short 5 min breaks. Approximate breakdown of material (details in survey paper):

Introduction to Synthetic Data, Background & Preliminaries
- Synthetic Data as a Privacy Enhancing Technology
- Desiderata for Synthetic Data
- Social and Legal Reasons for using Synthetic Data
Marginal Based Methods
- Early Statistical Approaches -- Data Swapping & SMOTE
- Probabilistic Graphical Models
- Early Marginal Based Approaches -- Naïve Bayes & MWEM.
- PrivBayes: A Middle Way
- Richer Models -- Private-PGM, MST, PrivMRF, PrivSyn
- State-of-the-Art -- Select-Measure-Generate Paradigm, RAP++ & AIM
Deep Learning Based Methods
- Generative Adversarial Networks -- CTGAN, TVAE etc.
- Extensions to GANs -- PATE-GAN, GEM
- Recent Advances -- Diffusion-based Models
- Deep Learning vs Marginal-based Methods
Privacy Attacks & Defenses on Synthetic Data
- Privacy Analysis Setup -- Threat Models & Privacy Risk Measurement
- Techniques for Privacy Attacks & Risk Analysis
Advanced Topics & Open Problems
- Additional Modalities of Synthetic Data -- Graphs, Text, Images & Videos
- Distributed Synthetic Data Generation
- Pragmatic Privacy Considerations in Differentially Private Synthetic Data
- Synthetic Data Generation Frameworks

Reference Materials

This webpage is a companion to the tutorial presented at KDD 2025 and VLDB 2025. We intend to share tutorial materials here including slides, videos (if any), survey paper, and links to references & libraries. More details will be posted in the coming weeks.