Privacy-Preserving DBSCAN Clustering Over Vertically Partitioned Data

Data mining has been a popular research area for more than a decade because of its ability of efficiently extracting statistics and trends from large sets of data. However, in many applications, the data are originally collected at different sites owned by different users. The distributed data mining raises concerns about the privacy of individuals. This paper considers the problem of privacy preserving DBSCAN clustering over vertically partitioned data based on some results of SMC. Each site learns the final results about the clusters, but learns nothing about any other site ´s data. An efficient secure intersection protocol is first proposed to implement privacy preserving DBSCAN clustering. The security and complexity of the protocols are also analyzed. The results show that the protocols preserve the privacy of the data and the time complexity as well as the communication complexity is acceptable.