Towards Personalized Differentially Private Learning for Decentralized Local Graphs

2026-07-06

AI Summary

This work addresses a critical limitation in existing decentralized graph learning approaches that employ local differential privacy: the use of a uniform privacy budget across all users, which disregards heterogeneous privacy preferences and often leads to excessive perturbation and degraded model utility. To overcome this, we propose PPGNN, a novel framework that introduces personalized privacy budgets into decentralized graph learning for the first time. PPGNN incorporates a Personalized Perturbation Mechanism (PPM) and a Weighted Calibration Strategy (FlexProp), enabling user-defined privacy guarantees during local perturbation while mitigating noise-induced distortion through a two-stage optimization process that balances privacy and utility. Extensive experiments on six real-world graph datasets demonstrate that PPGNN significantly outperforms state-of-the-art methods, effectively accommodating diverse privacy requirements while enhancing model performance.

Abstract

Graph-structured data is increasingly generated and stored in decentralized environments, such as social platforms, mobile applications, and edge networks, where users maintain control over their local graph data. However, collecting and analyzing such decentralized graph data for downstream learning tasks raises significant privacy concerns, as nodes and their attributes often contain sensitive personal information. Local Differential Privacy (LDP) has emerged as a promising solution for privacy-preserving data collection without relying on trusted servers. Nevertheless, existing LDP-based graph learning methods typically assume uniform privacy requirements across users, ignoring the heterogeneous and personalized privacy preferences commonly observed in real-world systems. This uniform treatment leads to inflexible noise injection at the data collection stage, resulting in substantial distortion of graph data and degraded utility in subsequent analysis. To address this limitation, we propose PPGNN, a personalized differentially private framework for decentralized graph data. PPGNN enables user-specific privacy budgets during local perturbation while preserving analytical utility. To handle heterogeneous privacy levels and noise distortion, we design a two-stage solution consisting of a Personalized Perturbation Mechanism (PPM) and a weighted calibration strategy, FlexProp. Extensive experiments on six real-world graph datasets demonstrate that PPGNN effectively balances personalized privacy protection and data utility in decentralized graph learning scenarios.

Problem

• Personalized Privacy
• Local Differential Privacy
• Decentralized Graphs
• Privacy Heterogeneity
• Graph Data Utility

Innovation

• Personalized Differential Privacy
• Decentralized Graph Learning
• Local Differential Privacy
• Graph Data Perturbation
• Privacy-Utility Trade-off