Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Building a RAPPOR with the Unknown: Privacy-Preserving Learning of Associations and Data Dictionaries

Building a RAPPOR with the Unknown: Privacy-Preserving Learning of Associations and Data... Abstract Techniques based on randomized response enable the collection of potentially sensitive data from clients in a privacy-preserving manner with strong local differential privacy guarantees. A recent such technology, RAPPOR (12), enables estimation of the marginal frequencies of a set of strings via privacy-preserving crowdsourcing. However, this original estimation process relies on a known dictionary of possible strings; in practice, this dictionary can be extremely large and/or unknown. In this paper, we propose a novel decoding algorithm for the RAPPOR mechanism that enables the estimation of “unknown unknowns,” i.e., strings we do not know we should be estimating. To enable learning without explicit dictionary knowledge, we develop methodology for estimating the joint distribution of multiple variables collected with RAPPOR. Our contributions are not RAPPOR-specific, and can be generalized to other local differential privacy mechanisms for learning distributions of string-valued random variables. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Proceedings on Privacy Enhancing Technologies de Gruyter

Building a RAPPOR with the Unknown: Privacy-Preserving Learning of Associations and Data Dictionaries

Download PDF
21 pages

Loading next page...
 
/lp/de-gruyter/building-a-rappor-with-the-unknown-privacy-preserving-learning-of-0ps0Gc0j0b
Publisher
de Gruyter
Copyright
Copyright © 2016 by the
ISSN
2299-0984
eISSN
2299-0984
DOI
10.1515/popets-2016-0015
Publisher site
See Article on Publisher Site

Abstract

Abstract Techniques based on randomized response enable the collection of potentially sensitive data from clients in a privacy-preserving manner with strong local differential privacy guarantees. A recent such technology, RAPPOR (12), enables estimation of the marginal frequencies of a set of strings via privacy-preserving crowdsourcing. However, this original estimation process relies on a known dictionary of possible strings; in practice, this dictionary can be extremely large and/or unknown. In this paper, we propose a novel decoding algorithm for the RAPPOR mechanism that enables the estimation of “unknown unknowns,” i.e., strings we do not know we should be estimating. To enable learning without explicit dictionary knowledge, we develop methodology for estimating the joint distribution of multiple variables collected with RAPPOR. Our contributions are not RAPPOR-specific, and can be generalized to other local differential privacy mechanisms for learning distributions of string-valued random variables.

Journal

Proceedings on Privacy Enhancing Technologies – de Gruyter

Published: Jul 1, 2016

References

  • Differentially private aggregation of distributed time - series with transformation and encryption In SIGMOD International Conference on Management of data pages
    Vibhor Rastogi
  • Categorical Data Series in Probability and nd edition
    Agresti
  • Differentially private aggregation of distributed time - series with transformation and encryption In SIGMOD International Conference on Management of data pages
    Vibhor Rastogi
  • Local privacy and statistical minimax rates In pages
    John
  • Categorical Data Series in Probability and nd edition
    Agresti
  • Differentially private recommender systems : building privacy into the net In Proceedings of the th SIGKDD international conference on Knowledge discovery and data mining pages
    Frank McSherry
  • Maximum likelihood from incomplete data via the em algorithm of the Series
    Dempster
  • Extremal mechanisms for local differential privacy In NIPS
    Peter Kairouz
  • Adam note on differential privacy : Defining resistance to arbitrary side information
    Shiva Prasad Kasiviswanathan
  • Adam Local private efficient protocols for succinct histograms In to appear
    Raef Bassily
  • Differentially private recommender systems : building privacy into the net In Proceedings of the th SIGKDD international conference on Knowledge discovery and data mining pages
    Frank McSherry
  • Alessandro Differential privacy and the risk - utility tradeoff for multidimensional contingency tables In Privacy in Statistical Databases pages
    Stephen
  • Adam Local private efficient protocols for succinct histograms In to appear
    Raef Bassily
  • Fool gold an illustrated critique of differential privacy
    Jane
  • Alessandro Differential privacy and the risk - utility tradeoff for multidimensional contingency tables In Privacy in Statistical Databases pages
    Stephen
  • Probabilistic inference differential privacy In NIPS pages
    Oliver Williams
  • Dawn Privacy - preserving aggregation of time - series data In volume page
    Elaine Shi
  • Dawn Privacy - preserving aggregation of time - series data In volume page
    Elaine Shi
  • Google Inc Unwanted Software Policy http www google com about company unwanted software policy html
  • Differentially private online learning arXiv preprint arXiv
    Prateek Jain
  • Probabilistic inference differential privacy In NIPS pages
    Oliver Williams
  • der Asymptotic volume university press
    Van
  • Randomized aggregatable privacy - preserving ordinal response
    Úlfar Erlingsson
  • Maximum likelihood from incomplete data via the em algorithm of the Series
    Dempster
  • Differentially private continual monitoring of heavy hitters from distributed streams In pages
    Hubert Chan
  • Distributed private heavy hitters In Automata and pages
    Justin Hsu
  • Joe André Madeira Fast private norm estimation and heavy hitters In of pages
    Kilian
  • Joe André Madeira Fast private norm estimation and heavy hitters In of pages
    Kilian
  • The sites on the web http www alexa com topsites
    Alexa
  • Randomized response survey technique for eliminating evasive answer bias
    Stanley Warner
  • Adam note on differential privacy : Defining resistance to arbitrary side information
    Shiva Prasad Kasiviswanathan
  • Adam What can we learn privately In pages
    Shiva Prasad Kasiviswanathan
  • Maximum likelihood estimation for dependent observations of the
    Martin
  • Google Inc Unwanted Software Policy http www google com about company unwanted software policy html
  • Local privacy and statistical minimax rates In pages
    John
  • Distributed private heavy hitters In Automata and pages
    Justin Hsu
  • Fool gold an illustrated critique of differential privacy
    Jane
  • likelihood ratio and lagrange multiplier tests in econometrics Handbook of econometrics
    Robert
  • On finding cliques in partite graphs Optimization
    Mirghorbani
  • likelihood ratio and lagrange multiplier tests in econometrics Handbook of econometrics
    Robert
  • Differentially private online learning arXiv preprint arXiv
    Prateek Jain
  • Extremal mechanisms for local differential privacy In NIPS
    Peter Kairouz
  • der Asymptotic volume university press
    Van
  • On finding cliques in partite graphs Optimization
    Mirghorbani
  • The sites on the web http www alexa com topsites
    Alexa
  • Randomized aggregatable privacy - preserving ordinal response
    Úlfar Erlingsson
  • Differentially private continual monitoring of heavy hitters from distributed streams In pages
    Hubert Chan
  • Maximum likelihood estimation for dependent observations of the
    Martin
  • Adam What can we learn privately In pages
    Shiva Prasad Kasiviswanathan
  • Randomized response survey technique for eliminating evasive answer bias
    Stanley Warner