TS-Inverse: A Gradient Inversion Attack Tailored for Federated Time Series Forecasting Models

2025-03-26

AI Summary

This work addresses the privacy risk of gradient inversion attacks (GIAs) against time-series forecasting (TSF) models in federated learning, where joint leakage of observed inputs and target outputs in time-series regression tasks has been largely unexplored. We propose the first GIA tailored to TSF, integrating quantile regression modeling, periodicity- and trend-aware regularization loss, and a quantile-based adaptive gradient regularization mechanism. Evaluated on four state-of-the-art TSF architectures and four real-world time-series datasets, our method reduces sMAPE reconstruction error by 2–10× compared to image-domain transferable GIA baselines, significantly improving time-series reconstruction fidelity. This study fills a critical gap in privacy threat analysis for time-series federated learning and establishes a new paradigm and benchmark toolkit for gradient security assessment in TSF scenarios.

Technology Category

• Machine Learning: Privacy
• Game Theory and Economic Paradigms: Adversarial Learning
• Computer Vision: Adversarial Attacks & Robustness

Application Category

• Security and Privacy: Large-scale security measurements
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Federated learning (FL) for time series forecasting (TSF) enables clients with privacy-sensitive time series (TS) data to collaboratively learn accurate forecasting models, for example, in energy load prediction. Unfortunately, privacy risks in FL persist, as servers can potentially reconstruct clients' training data through gradient inversion attacks (GIA). Although GIA is demonstrated for image classification tasks, little is known about time series regression tasks. In this paper, we first conduct an extensive empirical study on inverting TS data across 4 TSF models and 4 datasets, identifying the unique challenges of reconstructing both observations and targets of TS data. We then propose TS-Inverse, a novel GIA that improves the inversion of TS data by (i) learning a gradient inversion model that outputs quantile predictions, (ii) a unique loss function that incorporates periodicity and trend regularization, and (iii) regularization according to the quantile predictions. Our evaluations demonstrate a remarkable performance of TS-Inverse, achieving at least a 2x-10x improvement in terms of the sMAPE metric over existing GIA methods on TS data. Code repository: https://github.com/Capsar/ts-inverse

Problem

• Assessing gradient inversion attacks on federated time series models
• Overcoming challenges in reconstructing time series observations and targets
• Proposing TS-Inverse for improved time series data inversion

Innovation

• Gradient inversion model with quantile predictions
• Loss function with periodicity and trend regularization
• Regularization based on quantile predictions