Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SoK: Practical Aspects of Releasing Differentially Private Graphs

Nicholas D'Silva, Surya Nepal, Salil S. Kanhere

Abstract

Graph data is increasingly prevalent across domains, offering analytical value but raising significant privacy concerns. Edges may encode sensitive relationships, while node attributes may contain sensitive entity or personal data. Differential Privacy (DP) has gained traction for its strong guarantees, yet applying DP to graphs is challenging because of their complex relational structure, leading to trade-offs between privacy and utility. Existing methods vary in privacy definitions, utility goals, and contextual settings, complicating comparison. For practitioners, this is compounded by DP's interpretability issues, contributing to misleading protection claims. To address this, we propose a novel systemisation of existing methods tailored to practical considerations and adaptable to varying practitioner objectives. Our contributions include: (i) a comprehensive survey of differentially private graph release methods; (ii) identification of key vulnerabilities; and (iii) a practitioner-oriented, objective-based framework to guide the selection, interpretation, and sound evaluation of existing methods. We demonstrate the use of our systemisation through two exemplary scenarios in which we assume the role of a social network analyst, apply it, and conduct evaluations in accordance with our framework. Together, these two illustrative instantiations ultimately provide a unified benchmark for state-of-the-art methods in the social networks domain.

SoK: Practical Aspects of Releasing Differentially Private Graphs

Abstract

Graph data is increasingly prevalent across domains, offering analytical value but raising significant privacy concerns. Edges may encode sensitive relationships, while node attributes may contain sensitive entity or personal data. Differential Privacy (DP) has gained traction for its strong guarantees, yet applying DP to graphs is challenging because of their complex relational structure, leading to trade-offs between privacy and utility. Existing methods vary in privacy definitions, utility goals, and contextual settings, complicating comparison. For practitioners, this is compounded by DP's interpretability issues, contributing to misleading protection claims. To address this, we propose a novel systemisation of existing methods tailored to practical considerations and adaptable to varying practitioner objectives. Our contributions include: (i) a comprehensive survey of differentially private graph release methods; (ii) identification of key vulnerabilities; and (iii) a practitioner-oriented, objective-based framework to guide the selection, interpretation, and sound evaluation of existing methods. We demonstrate the use of our systemisation through two exemplary scenarios in which we assume the role of a social network analyst, apply it, and conduct evaluations in accordance with our framework. Together, these two illustrative instantiations ultimately provide a unified benchmark for state-of-the-art methods in the social networks domain.
Paper Structure (68 sections, 6 equations, 4 figures, 9 tables)

This paper contains 68 sections, 6 equations, 4 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Contributions
  3. Organisation
  4. Related Work
  5. Background on Differential Privacy
  6. Systemisation
  7. Scope and Methodology
  8. Structure
  9. Approaches
  10. L1: Graph Models
  11. L2: Trust Models
  12. Centralised
  13. Localised
  14. L3: Privacy Mechanism
  15. (Pure) Differential Privacy
  16. ...and 53 more sections

Figures (4)

  • Figure 1: Systemisation of DP graph release methods (central, blue), with two practical aspects: vulnerabilities (right, red), and practitioner objectives (left, green). The reader is encouraged to parse this as a flowchart from top to bottom, along the central blue column from sensitive input to private output. Overlaid pink and purple boxes represent two exemplar scenarios respectively (Sec. \ref{['evaluations.scenario-1']}, \ref{['evaluations.scenario-2']}), to illustrate our systemisation's use. A clean version is provided in our supplementary repository.
  • Figure 2: Results for Scenario 1. Distribution errors are quantified with the Wasserstein distance; harmonic diameter with relative error; and remaining error-based metrics with absolute error. All values are averaged over 10 independent trials. Each metric is assigned an alphabetical identifier.
  • Figure 3: Link prediction accuracy against edge-DP methods, evaluated on the Facebook dataset.
  • Figure 4: Results for Scenario 2. Distribution errors are quantified with the Wasserstein distance; the number of nodes, number of edges, and harmonic diameter with relative error; and remaining error-based metrics with absolute error. All values are averaged over 10 independent trials.