Abstract:

Privacy-preserving data mining aims to accurately mine knowledge while unrevealing sensitive data. For solving the privacy-preserving clustering problem in vertical distribution, a privacy-preserving clustering model based on full homomorphous encryption protocols and data perturbation technology is proposed. The model protects original data effectively by using secure comparison protocols to compute the nearest cluster and estimate the updating of the cluster center, which are two key steps in clustering process. Theory argument demonstrates the security of the privacy-preserving clustering model and analyzes computation complexity and communication costs. Experiment results prove that the privacy-preserving clustering model is secure and effective.