Benchmarking Time Series Generation Methods for Privacy-Preserving Forecasting

2026-08-11

AI Summary

This study addresses the challenge of balancing predictive utility and privacy preservation in sensitive scenarios by leveraging synthetic time series as substitutes for real data. The authors propose a “train-on-synthetic, test-on-real” evaluation protocol and systematically compare multiple generative methods against noise-based anonymization baselines across seven datasets. They introduce Grasynda-P, an extension of graph-based generation that integrates matrix ensembling with kernel density estimation. Their experiments provide the first comprehensive characterization of the trade-off between prediction accuracy and distance-based privacy risk, revealing that no synthetic method fully replaces real data; noise-based anonymization offers the strongest privacy but poorest utility; simple transformations consistently outperform complex deep generative models; and Grasynda-P achieves a superior privacy-utility balance, residing on the Pareto frontier.

Abstract

Time series forecasting in privacy-sensitive domains often requires training models on released data rather than original observations. Synthetic time series generation has been developed primarily for data augmentation, where generated series supplement the original training set. How well these methods perform when fully replacing the original data - and how much privacy risk the released series carry - remains underexplored. We address this gap through a benchmark evaluating synthetic generation methods and noise-based anonymization baselines under a Train on Synthetic, Test on Real (TSTR) protocol. We jointly assess forecasting performance and distance-based empirical privacy risk across seven datasets, characterizing the trade-off between these objectives. We also introduce Grasynda-P, a privacy-motivated extension of the graph-based generator Grasynda, incorporating matrix ensembling and kernel density estimation. Our results show that: (1) no generation method fully substitutes for original training data; (2) noise-based anonymization yields the strongest privacy but the worst forecasting performance; (3) simple transformation-based generators outperform deep generative models for forecasting in this setting; and (4) Grasynda-P lies on the Pareto frontier, achieving competitive forecasting with stronger privacy separation than other generators. This benchmark establishes a reference point for evaluating and developing new privacy-aware synthetic time series generation methods.

Problem

• time series generation
• privacy-preserving forecasting
• synthetic data
• empirical privacy risk
• TSTR protocol

Innovation

• synthetic time series generation
• privacy-preserving forecasting
• Grasynda-P
• TSTR benchmark
• empirical privacy risk