PGB (Private Graph Benchmark) is a comprehensive benchmark designed to enable researchers to compare differentially private graph generation algorithms fairly.
PGB Paper: https://arxiv.org/abs/2408.02928#
PGB Code: https://github.com/dooohow/PGB
| Algorithm | Paper | Code |
|---|---|---|
| DP1K | Preserving differential privacy in degree-correlation based graph generation,2013 | https://github.com/dooohow/PGB |
| Tmf | Differentially private publication of social graphs at linear cost,2015 | https://github.com/dooohow/PGB |
| PrivSKG | A differentially private estimator for the stochastic kronecker graph model,2012 | https://github.com/dooohow/PGB |
| PrivHRG | Differentially private network data release via structural inference,2014 | https://github.com/kaseyxiao/privHRG |
| PrivGraph | {PrivGraph}: Differentially Private Graph Data Publication by Exploiting Community Information,2023 | https://github.com/Privacy-Graph/PrivGraph |
| DGG | Generating synthetic decentralized social graphs with local differential privacy,2017 | https://github.com/dooohow/PGB |
1. Each number shows how often the algorithm performs best across 15 queries,given a privacy budget ε and a graph data set. For example,
the first number "5" means that DP-dK out performs others in 5 queries (i.e.,Q5,Q6,Q9,Q12,Q13) for the Minnesota graph with ε=0.1.
2. The highest frequency in each case is highlighted in gray.
