The protocol of privacy-preserving clustering with distributed EM mixture modeling was proposed. However, it is not completely secure in the situation that something more than just the model parameters are revealed. Specially, when the dataset is horizontally partitioned into just two parts, this reveals extra information. The aim of this work is firstly to develop a more general protocol which allows the number of participating parties to be arbitrary and more secure. Secondly, we propose a

more »

... ter method for the case in which the dataset is horizontally partitioned into only two parts. This method allows computing covariance matrices and final results without revealing the private information and the clustering centers. Tóm tȃt. Mô . t số giao thú . c da ' m ba ' o tính riêng tu . trong bài toán phân cu . m dũ . liê . u du . . a trên thuâ . t toán EM dã du . o . . c dề xuất trong cô . ng dồng nghiên cú . u. Tuy nhiên, các giao thú . c này không hoàn toàn da ' m ba 'o tính riêng tu . , vì trong mô . t số tình huống mô . t vài tham số cu 'a mô hình có thê ' bi . lô . ,