On the Utility Recovery Incapability of Neural Net-based Differential Private Tabular Training Data Synthesizer under Privacy Deregulation

2022/11/28 by Yucong Liu, Liu, Yucong, Chi-Hua Wang +3 · 1 citation
Computer Science · #Cryptography and Data Security #FOS: Computer and information sciences #Machine Learning (cs.LG) #Privacy-Preserving Technologies in Data #Stochastic Gradient Optimization Techniques

paper · pdf · doi:10.48550/arxiv.2211.15809

openalex publication_date 2022/11/28 · openalex created_date 2022/12/11 · openalex updated_date 2026/07/28

Abstract

Devising procedures for auditing generative model privacy-utility tradeoff is an important yet unresolved problem in practice. Existing works concentrates on investigating the privacy constraint side effect in terms of utility degradation of the train on synthetic, test on real paradigm of synthetic data training. We push such understanding on privacy-utility tradeoff to next level by observing the privacy deregulation side effect on synthetic training data utility. Surprisingly, we discover the Utility Recovery Incapability of DP-CTGAN and PATE-CTGAN under privacy deregulation, raising concerns on their practical applications. The main message is Privacy Deregulation does NOT always imply Utility Recovery.

Cited by

• Synthetic Tabular Data: Methods, Attacks and Defenses

Related

• Privacy Regularization: Joint Privacy-Utility Optimization in Language Models
• Graphical vs. Deep Generative Models: Measuring the Impact of Differentially Private Mechanisms and Budgets on Utility
• Differentially Private Neural Network Training Under the Hidden State Assumption
• Does Training with Synthetic Data Truly Protect Privacy?
• Differentially Private Synthetic Data: Applied Evaluations and Enhancements