Privacy-Preserving Distributed k-Nearest Neighbor Mining on Horizontally Partitioned Multi-Party Data

摘要

k-Nearest Neighbor (k-NN) mining aims to retrieve the k most similar objects to the query objects. It can be incorporated into many data mining algorithms, such as outlier detection, clustering, and k-NN classification. Privacy-preserving distributed k-NN is developed to address the issue while preserving the participants' privacy. Several two-party privacy-preserving k-NN mining protocols on horizontally partitioned data had been proposed, but they fail to deal with the privacy issue when the number of the participating parties is greater than two. This paper proposes a set of protocols that can address the privacy issue when there are more than two participants. The protocols are devised with the probabilistic public-key cryptosystem and the communicative cryptosystem as the core privacy-preserving infrastructure. The protocols' security is proved based on the Secure Multi-party Computation theory.