Core Concepts
Quantifying information disclosure in secure computation is crucial for protecting private inputs.
Abstract
The content delves into the importance of quantifying information disclosure in secure multi-party computation, focusing on the computation of average salaries. It discusses the impact of different distributions on information leakage and provides recommendations for minimizing disclosure. The analysis covers discrete and continuous distributions, highlighting the behavior of entropy loss relative to the original entropy before execution.
Introduction to Secure Multi-Party Computation and Data Analysis Practices.
Fundamental Questions in Secure Computing Practices.
Quantitative Information Flow and Measures of Information Leakage.
Function Information Disclosure in Secure Multi-Party Computations.
Preliminaries on Differential Entropy and Summation of Random Variables.
Analysis of Single Execution with Discrete Distributions (Uniform, Poisson).
Analysis of Single Execution with Continuous Distributions (Normal, Log-Normal).
Recommendations for Lowering Information Disclosure in Practice.
Stats
"Our goal is to develop mechanisms that lower information disclosure about participants’ inputs to a desired level."
"The absolute entropy loss remains very close for different distributions as we vary the number of participants/spectators."
"Engaging in the computation the second time with an overlapping set of 50% participants whose inputs do not change results in only 30% entropy loss."