ResearchTrend.AI
  • Papers
  • Communities
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 1911.04842
67
8
v1v2v3 (latest)

Developing Non-Stochastic Privacy-Preserving Policies Using Agglomerative Clustering

12 November 2019
Ni Ding
F. Farokhi
ArXiv (abs)PDFHTML
Abstract

We consider a non-stochastic privacy-preserving problem in which an adversary aims to infer sensitive information SSS from publicly accessible data XXX without using statistics. We consider the problem of generating and releasing a quantization X^\hat{X}X^ of XXX to minimize the privacy leakage of SSS to X^\hat{X}X^ while maintaining a certain level of utility (or, inversely, the quantization loss). The variables SSS and SSS are treated as bounded and non-probabilistic, but are otherwise general. We consider two existing non-stochastic privacy measures, namely the maximum uncertainty reduction L0(S→X^)L_0(S \rightarrow \hat{X})L0​(S→X^) and the refined information I∗(S;X^)I_*(S; \hat{X})I∗​(S;X^) (also called the maximin information) of SSS. For each privacy measure, we propose a corresponding agglomerative clustering algorithm that converges to a locally optimal quantization solution X^\hat{X}X^ by iteratively merging elements in the alphabet of XXX. To instantiate the solution to this problem, we consider two specific utility measures, the worst-case resolution of XXX by observing X^\hat{X}X^ and the maximal distortion of the released data X^\hat{X}X^. We show that the value of the maximin information I∗(S;X^)I_*(S; \hat{X})I∗​(S;X^) can be determined by dividing the confusability graph into connected subgraphs. Hence, I∗(S;X^)I_*(S; \hat{X})I∗​(S;X^) can be reduced by merging nodes connecting subgraphs. The relation to the probabilistic information-theoretic privacy is also studied by noting that the G{\'a}cs-K{\"o}rner common information is the stochastic version of I∗I_*I∗​ and indicates the attainability of statistical indistinguishability.

View on arXiv
Comments on this paper