Pain-Free Random Differential Privacy with Sensitivity Sampling
Authors: Benjamin I. P. Rubinstein, Francesco Aldà
ICML 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details
| Reproducibility Variable | Result | LLM Response |
|---|---|---|
| Research Type | Experimental | We now demonstrate the practical value of SENSITIVITYSAMPLER. First in Section 7.1 we illustrate how SENSITIVITYSAMPLER sensitivity quickly approaches analytical high-probability sensitivity, and how it can be significantly lower than worst-case global sensitivity in Section 7.2. Running privatising mechanisms with lower sensitivity parameters can mitigate utility loss, while maintaining (a weaker form of) differential privacy. We present experimental evidence of this utility savings in Section 7.3. |
| Researcher Affiliation | Academia | 1School of Computing and Information Systems, University of Melbourne, Australia 2Horst G ortz Institute for IT Security and Faculty of Mathematics, Ruhr-Universit at Bochum, Germany. |
| Pseudocode | Yes | Algorithm 1 SENSITIVITYSAMPLER |
| Open Source Code | No | The paper cites a technical report on ArXiv (Rubinstein & Ald a, 2017) for full proofs, but it does not explicitly state that the source code for the methodology described in the paper is released or provide a direct link to a code repository. |
| Open Datasets | No | The paper describes synthetic datasets for its experiments ('We synthesise a dataset of n = 1000 points...' and 'mixture of two normal distributions...'), but does not provide concrete access information (link, DOI, repository, or citation) for a publicly available dataset. |
| Dataset Splits | No | The paper describes synthetic datasets and their sizes, but does not provide specific details on how these datasets were split into training, validation, or test sets, nor does it mention cross-validation. |
| Hardware Specification | No | The paper mentions 'on a stock machine' for sorting, but it does not provide specific hardware details (e.g., CPU/GPU models, memory, or cloud instance types) used for running the main experiments. |
| Software Dependencies | No | The paper does not provide specific version numbers for any software dependencies or libraries used (e.g., 'Python 3.8, PyTorch 1.9' or 'CPLEX 12.4'). |
| Experiment Setup | Yes | The SVMs are run with C = 3, SENSITIVITYSAMPLER with m = 1500 & varying γ. [...] Bernstein mechanism (with lattice size k = 10 and Bernstein order h = 3) |