Work

Auditing Privacy Loss in a Low-Precision Noise Sampler

Year: 2026

Type: book

Abstract: This programme enumerates every possible output of a deliberately low-precision sampler for neighbouring counts of five and six. It identifies a violation of pure differential privacy, then repeats the analysis after rounding and clipping to calculate the remaining pairwise excess probability. The experiment demonstrates why ordinary post-processing does not automatically repair an incorrectly implemented privacy mechanism.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.22905885

Author: Majumdar, Partha

Institution: Kalinga University

Language: English

Topic: Privacy-Preserving Technologies in Data

Subfield: Artificial Intelligence

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7214039293

FWCI:

Cites:

Cited by:

Related to: 10