Consensus-Based Privacy-Preserving Algorithm

In this paper, we use secure multi-party computation to protect privacy. Based on consensus-based distributed support vector machines, we present a new consensus-based privacy-preserving algorithm to conduct secure multi-party computation. The proposed algorithm run in parallel at each iteration, which reduce the running time. Furthermore, what needed to be communicated at each iteration is only a coefficient vector, therefore privacy is protected to the uttermost. The algorithm is proved to be convergent globally. Numerical experiments demonstrate the feasibility and efficiency of the new algorithm.