【摘要】 针对分布式数据共享及计算中的隐私保护问题,提出了一种适用于大规模分布式环境的隐私保护计算模型(PPCMLS),该模型的核心为隐私安全模块,其将计算划分为本地计算和全局计算.通过综合运用同态加密、安全点积协议、数据随机扰乱算法等多种安全技术,在实现了多个节点在一个互不信任的分布式环境下合作计算的同时,任何节点无法获取其他节点的隐私信息及敏感中间计算结果.据此,又给出了基于该模型的分布式隐私保护方差计算、分布式隐私保护数据聚类算法.安全及动态性分析结果表明,该模型及其应用算法既可保证隐私数据的安全性,又避免了繁琐的一对多的交互加密过程,并在节点变化时,恢复计算仅涉及到变化的节点和构成隐私安全模块的3个节点,从而满足了大规模分布式环境所要求的高效性和良好的动态适应性.更多还原

【Abstract】 For protecting the privacy in the large-scale distributed database sharing and computation,a generic privacy-protecting computation model supporting the privacy-protecting data computation and analysis is proposed.The key of this model is the ’privacy security module’ which divides the distributed computation into ’local computation’ and ’global computation’.By using homomorphic encryption,secure product protocol,random permutation and some other secure technologies comprehensively,the model achieves distributed data computations without compromising the privacy of both data record and sensitive intermediate computation result.Two privacy-protecting distributed algorithms based on this model,namely,the privacy-protecting distributed variance algorithm and privacy-preserving distributed k-means clustering algorithm,have also been introduced.The security and dynamic behavior analysis show that the proposed model and algorithms doesn’t require any one-to-all interactions encryption,and the recovery computation only involves the changed nodes and the three nodes composing privacy security module when nodes change.So it’s suitable for large-scale and dynamic distributed environments.更多还原