A modified secure dot product in the privacy-preserving k-means clustering with the vertical partitioning data

The authors consider the actual problems in the design and implementation of the privacy-preserving computations in the multiparty data clustering. There is a basic description of the secure multiparty computations and the definition of the adversary model in it. This article presents the flaws of existing solutions of the privacy-preserving k-means clustering over the vertically partitioned data, and proposed ways to eliminate these disadvantages by the correct using the cryptographic primitive realized secure dot product. The mathematical apparatus given in this article describes all aspects of the original protocol and the details of its modernization. The article has a comparative analysis of the advanced cryptographic primitive with its analog - secure sum. There are conclusions about the effectiveness of improved cryptographic primitive of the secure dot product and recommendations on application of the developed model.