Universally Optimal Privacy Mechanisms for Minimax Agents - Computer Science > Cryptography and SecurityReport as inadecuate




Universally Optimal Privacy Mechanisms for Minimax Agents - Computer Science > Cryptography and Security - Download this document for free, or read online. Document in PDF available to download.

Abstract: A scheme that publishes aggregate information about sensitive data mustresolve the trade-off between utility to information consumers and privacy ofthe database participants. Differential privacy is a well-establisheddefinition of privacy-this is a universal guarantee against all attackers,whatever their side-information or intent. In this paper, we present auniversal treatment of utility based on the standard minimax rule from decisiontheory in contrast to the utility model in, which is Bayesian. In our model,information consumers are minimax risk-averse agents, each possessing someside-information about the query, and each endowed with a loss-function whichmodels their tolerance to inaccuracies. Further, information consumers arerational in the sense that they actively combine information from the mechanismwith their side-information in a way that minimizes their loss. Under thisassumption of rational behavior, we show that for every fixed count query, acertain geometric mechanism is universally optimal for all minimax informationconsumers. Additionally, our solution makes it possible to release queryresults at multiple levels of privacy in a collusion-resistant manner.



Author: Mangesh Gupte, Mukund Sundararajan

Source: https://arxiv.org/



DOWNLOAD PDF




Related documents