Privacy Preserving Data Mining Using Time Series Data Aggregation

Abstract

This article proposes a mechanism to provide privacy to mined results by assuming that the data is distributed across many nodes. The first objective includes mining the query results by the node in a cluster, communicating it to the cluster head, aggregating the data collected from all the cluster nodes and then communicating it to the group controller. The second objective is to incorporate privacy at each level of the clusters node: cluster head and the group controller level. The final objective is to provide a dynamic network feature, where the nodes can join or leave the distributed network without disturbing the network functionality. The proposed algorithm was implemented and validated in Java for its performance in terms of communication costs computational complexity.