Privacy-preserving DBSCAN on horizontally partitioned data

Privacy preserving data mining of distributed data is an important direction for data mining, and privacy preserving clustering is one of the main researches. At present, most privacy preserving clustering algorithms are concentrated on k-means and based on two parties and a trusted third party, clustering results are uncertain and hard to find complex shape clusters, and the protocols are inefficient because of using encryption, so we propose a algorithm called HPPDBSCAN based on semi-honest models for horizontally partitioned databases using some secure protocols such as secure sum computation, scalar product computation, standardization, and comparison by means of a semi-honest third party. The algorithm resolves the problem of privacy preserving under semi-honest circumstance for multi-party. Theoretic argument and example analysis demonstrate that the scheme is secure and complete with good efficiency.