The Target-Charging Technique for Privacy Analysis across Interactive Computations

Authors: Edith Cohen, Xin Lyu

NeurIPS 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We demonstrate numerically the benefits of our TCT-based Between Thresholds (Section 2.6) compared with the SVT-based prior work of [3]. Figure 1 shows the dependence of the number of between responses (y axis) on the number of data points n (x axis) for TCT and for the provided and optimistic bounds for [3].
Researcher Affiliation Collaboration Edith Cohen Google Research and Tel Aviv University edith@cohenwang.com Xin Lyu UC Berkeley lyuxin1999@gmail.com
Pseudocode Yes Algorithm 1: Target Charging ... Algorithm 2: Conditional Release and Revise Calls ... Algorithm 3: One-Shot Top-k Selection ... Algorithm 4: Boundary Wrapper ... Algorithm 5: SVT with Individual Privacy Charging ... Algorithm 6: Target Charging with Approximate DP ... Algorithm 7: Simulation of Target Charging ... Algorithm 8: Multi-Target Charging ... Algorithm 9: Private Selection: A Simulation
Open Source Code No The paper does not contain any explicit statements or links indicating that open-source code for the described methodology is provided.
Open Datasets No The paper mentions a generic 'sensitive dataset D = {x1, . . . , xn}' in Section H for numerical comparison. It does not refer to a specific, publicly available dataset with concrete access information (e.g., URL, DOI, or formal citation).
Dataset Splits No The paper does not provide specific details about training, validation, or test dataset splits. The numerical comparison in Section H uses a generic 'dataset D = {x1, . . . , xn}' without specifying any splits.
Hardware Specification No The paper does not provide any specific hardware details (e.g., GPU/CPU models, memory amounts) used for running its experiments, including the numerical comparison in Section H.
Software Dependencies No The paper does not provide specific ancillary software details, such as library or solver names with version numbers, needed to replicate any numerical results or theoretical derivations.
Experiment Setup No The paper does not contain specific experimental setup details, such as concrete hyperparameter values or training configurations, as its primary focus is on theoretical privacy analysis. The parameters mentioned in Section H (ε, δ, α) relate to the theoretical bounds rather than an empirical training setup.