Privacy Preserving Data Mining in Terms of DBSCAN Clustering Algorithm in Distributed Systems

We raise an issue of information security in distributed data mining systems. Authors compare two approaches aimed to protect security of processed data in distributed systems: with the assistance of trusted third party and without it. An in-depth description of external and internal adversary capabilities to violate privacy, availability and entirety of processing data shows major drawbacks of approach based on attraction of trusted third party. The opposite approach is founded on the organization of secure channel which is based on homomorfic and asymmetric encryption. A privacy-preserving DBSCAN clustering algorithm over vertically partitioned data is demonstrated.