DPGraph is a web-based end-to-end benchmark platform built for researchers and practitioners to evaluate differentially private algorithms on graph data. This platform supports a rich set of tunable algorithms for popular graph statistics, including degree distribution and subgraph counting, with different differential privacy guarantees. We hope to support users to understand the trade-off between privacy, accuracy, and performance of existing work and discover suitable algorithms for their applications.
ACM Conference on Management of Data - SIGMOD 2021 Demo Paper
Theory and Practice of Differential Privacy Workshop - TPDP 2021
Feedback and contributions are welcomed. We hope visitors to this site will contribute to DPGraph by submitting new datasets, new algorithms, or suggestions. For now this is a manual process; please submit an Issue in the open source repository or email us at this address.
The differential privacy tools used to generate the results on this site are available in the DPGraph open source repository. The repository presents the algorithm implementations in Jupyter notebooks. Users can easily reproduce previous evaluations, evaluate provided algorithms on new data, or reuse or modify exisiting implementations. We welcome submissions of new algorithms.
DPGraph was inspired in part by previous websites for automatic and standarized execution of algorithms on datasets, such as MLcomp and DPComp.