项目名称: 基于安全多方计算的数据挖掘隐私保护研究
项目编号: No.60803146
项目类型: 青年科学基金项目
立项/批准年度: 2009
项目学科: 金属学与金属工艺
项目作者: 黄征
作者单位: 上海交通大学
项目金额: 18万元
中文摘要: 本项目研究使用安全多方计算协议来解决数据挖掘过程中的用户隐私保护问题。我们首先对数据挖掘隐私保护的需求进行了分析,然后将隐私保护的需求与安全多方计算协议的安全需求进行了对比。数据挖掘隐私保护需求是广义安全多方计算安全需求的一种特例,广义安全多方计算的形式化安全需求定义可以用于数据挖掘隐私保护的需求定义。为了提高安全多方计算协议用于解决隐私保护计算的效率,本项目研究了一种使用预计算的安全多方计算协议。协议参与者在预处理或计算空闲阶段分享大量满足乘法关系的随机数三元组,然后在数据挖掘计算阶段可以使用事先分享的随机数来掩盖用户的隐私数据,从而减少用户之间的交互,提高安全多方计算协议在计算阶段的效率。为了提高共享随机数算法的效率,本项目研究了一种批处理方式在多个协议参与者中分享随机数的协议,该协议与多次使用分享一个随机数的协议相比,具有交互轮数小,效率高的特点。对于具体的隐私数据保护问题,本项目基于安全多方计算和同态加密的思想研究了一种多服务器环境下隐私数据查询方案。
中文关键词: 安全多方计算;秘密分享;隐私保护数据挖掘
英文摘要: This project has studied how to use secure multiparty computation (SMC) protocol to protect user's privacy in data-mining applications. Firstly, we study the security requirements for user's privacy protecting data-mining (PPDM), and then we compare the requirement with the security requirement of secure multiparty computation. We realize that security requirement of SMC is a more generalized requirement and the security requirement of PPDM is a special case of that of SMC. We can use the security requirement for SMC to define the requirement for PPDM. We design a secure multiparty computation protocol that uses pre-computation to improve the efficiency of PPDM applications. In this protocol, users of PPDM can share a lot of random number among them; the random numbers will be used in the real computation phase to protect user's privacy. In this way, we reduce a lot of communication rounds which of course improve the efficiency of PPDM. We also design a secret sharing protocol that allows user to share random number in a batch mode which requires less communication bands than traditional secret sharing protocol. To the specific application of PPDM, we choose to study privacy information retrieve (PIR) protocol and design a PIR protocol for multi server environment using MPC and secret sharing technologies.
英文关键词: Secure Multiparty Computation; Secret Sharing; Privacy Preserving Data-mining