Design And Implementation Of Unified Framework Of Clustering Algorithmns Based On Secure Multiparty Computation

Clustering algorithms are widely used in many fields,but clustering algorithms can hardly apply to private data because of concerns about privacy leakage.It has important significance to research on running clustering algorithms in non-trusted environments.Secure Multiparty Computation(SMC)consists of several methods for parties to jointly compute functions over their private inputs while keeping their privacy secret.Therefore,secure multiparty computation is a suitable tool for applying clustering algorithms to private data.However,there many types of clustering methods,currently,there is no convenient and easy-to-use framework for deploying privacy-preserving clustering algorithms into private data.This thesis firstly improves two basic protocols of secure multiparty computation which are suitable for clustering algorithms.Then,it computes similarity of different private data using those basic protocols.Next,its studies network communication methods for secure multiparty computation,according to the characteristics of communication,this thesis introduces a configuration center for auxiliary establishing connections between participants.Finally,this thesis designs and implements a unified clustering framework based on secure multiparty computation and implements several privacy-preserving clustering algorithms based on the framework.The experimental results show that the system maintains the same accuracy in the results of clustering with normal algorithms.Furthermore,the system can be used to apply data mining into private data.