DRNNA: Decomposable Reverse Nearest Neighbor Algorithm for Vertically Distributed Databases

Abstract

There are many applications for the Reverse Nearest Neighbor (RNN) problem, including continuous referral systems, resource allocation, decision support, location-based services, bioinformatics, profile-based marketing, and many others. Although there exist numerous studies on RNN problem, most of the existing algorithms for RNN works on a single database. In contrast to the existing approaches, we propose a Decomposable Reverse Nearest Neighbor Algorithm (DRNNA), which computes RNN in a high dimensional space across distributed databases. DRNNA helps in minimizing the amount of data transferred between sites and hence provides data privacy and security for each site. Our approach aims to achieve valid results through minimum information disclosure. The data privacy at individual sites are preserved as our approach requires only minimal transmission of information between sites. Only a minimal count of higher-level summaries is exchanged for performing computations and therefore an intruder cannot obtain the actual data tuples even if they try to capture the exchanged summaries. The simulation results prove that the algorithm can correctly find the RNN set for a given point.