Subspace Differential Privacy
Authors: Jie Gao, Ruobin Gong, Fang-Yi Yu3986-3995
AAAI 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We showcase the proposed mechanisms on the 2020 Census Disclosure Avoidance demonstration data, and a spatio-temporal dataset of mobile access point connections on a large university campus. |
| Researcher Affiliation | Academia | Jie Gao1, Ruobin Gong1, Fang-Yi Yu2 1 Rutgers University 2 Harvard University |
| Pseudocode | Yes | Algorithm 1: Gaussian induced subspace differentially private mechanism through extension |
| Open Source Code | No | The paper does not provide an explicit statement or a link indicating the availability of open-source code for the described methodology. |
| Open Datasets | Yes | We showcase the proposed mechanisms on the 2020 Census Disclosure Avoidance demonstration data, and a spatio-temporal dataset of mobile access point connections on a large university campus. For our demonstration, the privacy loss budget is set to accord exactly to the Census Bureau s specification, with ϵ = 0.192 = 4 (total) 0.16 (county level) 0.3 (population query). using the November 2020 vintage privacy-protected demonstration files curated by IPUMS NHGIS (Van Riper, Kugler, and Schroeder 2020). |
| Dataset Splits | No | The paper does not explicitly provide details about training/validation/test dataset splits for reproducibility. |
| Hardware Specification | No | The paper does not provide specific hardware details (e.g., CPU/GPU models, memory) used for running its experiments. |
| Software Dependencies | No | The paper does not provide specific software dependencies with version numbers needed to replicate the experiments. |
| Experiment Setup | Yes | For our demonstration, the privacy loss budget is set to accord exactly to the Census Bureau s specification, with ϵ = 0.192 = 4 (total) 0.16 (county level) 0.3 (population query). Right panel of Figure 1 showcases the county-level errors from ten runs of the projected Laplace (ϵ, 0)-induced subspace differentially private mechanism of Corollary 3.8, applied to the counties of Illinois arranged in increasing true population sizes. We apply the projection Gaussian mechanism in Corollary 3.5 with the scale of the elementwise Gaussian noise set to 1. |