Paper title: Understanding Privacy and Quality Tradeoffs in Synthetic Network Data
Requested Badge(s):
- Available
- Functional
- Reproduced
Paper Title: Understanding Privacy and Quality Tradeoffs in Synthetic Network Data Authors: Andrew Chu (University of Chicago), Kyle MacMillan (University of Chicago), Paul Schmitt (Cal Poly), Nick Feamster (University of Chicago) Year: 2026
This repository contains the source code for the evaluation in our paper. Specifically, in folder membership_inference exists the code for the binary classifier-based MIA attack models for each of the three models we evaluate in Section 3 (NetShare, NetDiffusion, NetSSM), and in folder fidelity_utility exists the code for evaluating the fidelity and utility of synthetic data (Section 4).
The code in this artifact does not raise any security/privacy issues or ethical concerns. The models we train using this code are trained on pre-existing, publicly released datasets, and are not a novel contribution of this paper.
- Python3 >= 3.10
The file REQUIREMENTS.txt contains the Python libraries needed to run the scripts in this repo.
The artifact repo can be accessed at the following URL: https://github.com/noise-lab/networking-mia-pets/tree/main