高效保密的参数区间估计

Efficient Secure Parameter Interval Estimation

信息技术的飞速发展,一方面使我们获取信息的手段越来越多,另一方面也极易导致隐私信息泄露,使隐私保护面临严峻的挑战.安全多方计算已经成为联合计算中保护隐私的核心技术.在保护数据隐私的条件下,对隐私数据进行准确的统计分析具有重要的实际意义.因此保密计算隐私数据的均值、方差等参数的估计区间并保密判定这些估计区间是否包含于已知区间已经成为安全多方计算的一个关键问题,这些问题目前还没有见到相关研究成果.本文首先利用Paillier加密算法设计判定单个正态总体的方差的置信区间是否含于一个已知区间的保密计算协议,进一步应用Lifted ElGamal门限加密算法,设计了判定两个正态总体的均值之差的置信区间是否含于一个已知区间的保密计算协议.并应用模拟范例方法严格证明了协议的安全性.效率分析以及实验都表明本文所设计的协议是简单高效的.利用本文的方案,其他相关参数区间估计保密判定问题也能够得到有效解决.

The rapid development of information technology provides us with more and more channels to get information,while it also brings security threat of private information leakage,hence privacypreserving is an important technique.Secure multi-party computation provides an effective way of privacy-preserving.Privacy-preserving statistical analysis is an important problem and is of important practical significance.Privately computing the parameters such as mean,variance and their confidence intervals and privately determining whether their confidence interval is within a given interval are important problems of secure multiparty computation.This paper makes use of Paillier encryption algorithm to design a protocol which can privately determine whether the confidence intervals of variance of a single normal population is within a given interval.Then the lifted ElGamal encryption algorithm is used to design a protocol to privately determine whether the confidence intervals of the difference of the means of two normal populations is within a given interval.The well accepted simulation paradigm is used in this paper to prove that these two protocols are secure in the semihonest model.The efficiency of the proposed protocols is analyzed,and the protocols are implemented on a personal computer.Theoretical analysis and test show that the proposed protocols are simple and efficient,and can be used as building blocks to address other privacy-preserving parameter estimation problems.