Privacy-preserving clustering using C-means

Different data mining algorithms and techniques are developed for finding meaningful and important patterns from big databases. Nowadays databases are often divided between multiple organizations or users for the reason of geographical remoteness, but the most important limit is protecting privacy. Every participant involved in data analysis wants to keep its own data private and secure because of law regulations, reasons of economy or securing know-how. Secure multiparty computations are designed for data mining in a multiparty environment, where it is very important to keeping the privacy of the input (and possibly output) data. Data clustering is widely used in many fields of human activity like informatics, network traffic analysis, economics, biology and medicine. C-means is the data clustering method by which analytics can divide data objects into several fuzzy clusters; one object may be include into more than one cluster. This article presents protocols for preserving privacy in the process of cluster analysis using C-means. The protocols allow the implementation of C-means algorithm for several parties both to horizontally partitioned data and vertically partitioned data.