Privacy-Preserving Kernel k-Means Outsourcing with Randomized Kernels

Abstract

Kernel k-means is a useful way to identify clusters for nonlinearly separable data. Solving the kernel k-means problem is time consuming due to the quadratic computational complexity. Outsourcing the computations of solving kernel k-means to external cloud computing service providers benefits the data owner who has only limited computing resources. However, data privacy is a critical concern in outsourcing since the data may contain sensitive information. In this paper, we propose a method for privacy-preserving outsourcing of kernel k-means based on the randomized kernel matrix. The experimental results show that the clustering performance of the proposed randomized kernel k-means is similar to a normal kernel k-means algorithm.