摘要：Most of the existing computing algorithms aredesigned for environment in which all the related data arestored at a single site, horizontally distributed or particularcase of vertically distributed where different nodes containdifferent attributes of a common set of entities. Moreover,existing algorithms that work for general vertically distributeddatabases have exponential messages and elapsed time, resultingin heavy network traffic. To minimize the network traffic, it isdesirable to maximize the amount of local computations on eachof the participating nodes. In this work, a new perspective onthese two competing demands is presented to achieve a scalableand secure solution on the most general vertical and horizontaldistribution of databases. The proposed algorithm efficientlycompute inter-tuple distance based clustering for data in thefederated databases. We show this by finding the closest pair.The computation is executed by reciprocity involving only highlevel of summaries, i.e., the actual tuple can never be detectedby the intruder, providing high level of privacy and securityof databases. The simulation results show that the proposedalgorithm efficiently find exact solution with less total messageexchanges and elapsed time than existing algorithm.
