New privacy preserving clustering methods for secure multiparty computation

Many researches on privacy preserving data mining have been done. Privacy preserving data mining can be achieved in various ways by use of randomization techniques, cryptographic algorithms, anonymization methods, etc. Further, in order to increase the security of data mining, secure multiparty computation (SMC) has been introduced. Most of works in SMC are developed on applying the model of SMC on different data distributions such as vertically, horizontally and arbitrarily partitioned data. Another type of SMC with sharing data itself to each party attracts attention, and some studies have been done. A simple method to share data was proposed and it was applied to statistical computation. However, for SMC, complicated computation such as data mining has never been proposed. In the previous paper, we proposed a BP learning for SMC and showed the effectiveness of it. In this paper, we propose clustering methods such as k-means and NG for SMC and show the effectiveness in numerical simulation.