Local Pan-privacy for Federated Analytics

Authors: Vitaly Feldman, Audra Mcmillan, Guy N. Rothblum, Kunal Talwar

ICML 2025 | Venue PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Theoretical We show that local pan-privacy is severely limited if we insist on providing (information-theoretic) differential privacy in the face of intrusions. Our lower bound shows that for the COUNTNONZERO task, the error of any algorithm must be T times larger than that needed for local differential privacy alone. This lower bound holds even when the event occurs at most once on each device. Theorem 1 (Informal version of Theorem 9)... We then show that under standard cryptographic assumptions, local pan-privacy can be ensured without this overhead. We present algorithms for all of the aforementioned problems... Theorem 2 (Informal version of Theorem 10)... Theorem 3 (Informal version of Theorem 6.1). Suppose that we have a locally pan-private algorithm for COUNTNONZERO that has additive error less than n/4 with high probability. Then we can build a public key encryption scheme.
Researcher Affiliation Industry 1Apple, Cupertino, CA, USA. Correspondence to: Kunal Talwar <EMAIL>.
Pseudocode Yes Algorithm 1 f D Algorithm 2 h Algorithm 3 COUNTNONZERO, Client Algorithm Algorithm 4 Averaging, Client Algorithm Algorithm 5 Counter, Client Algorithm Algorithm 6 COUNTNONZERO, Client Algorithm, Two-server model
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 is theoretical and does not describe experiments that utilize specific datasets. Therefore, no information about open datasets is provided.
Dataset Splits No The paper does not describe experiments or the use of datasets, and therefore does not provide any information about dataset splits.
Hardware Specification No The paper is theoretical and does not describe any experimental setup or specific hardware used for running experiments.
Software Dependencies No The paper is theoretical and focuses on algorithm design and proofs, without detailing any software dependencies with version numbers for practical implementation or experimentation.
Experiment Setup No The paper is theoretical, presenting algorithms and proofs without conducting empirical experiments. Therefore, no specific experimental setup details, hyperparameters, or training configurations are provided.