Privacy-Preserving Regression Modeling and Attack Analysis in Sensor Network

With the advancements of sensing technologies, participatory data fusion and analysis have raised more and more attention as it provides a promising way enabling the public benefit from this process. However, it is increasingly becoming a challenging problem about how to construct a statistical model for this kind of particular phenomenon under the premise of protecting the data privacy. In this paper, we present a method to build a linear regression model describing a phenomenon observed in the distributed network. At the same time, the private data of each node is preserved. We propose a data aggregation algorithm to fuse the indispensable data for constructing linear regression equation even though the data is private and scattered in the whole network. We also point out that the aggregate node can not only conduct regression modeling, but can conduct some complex statistical analyses based on the aggregation result as well. In addition, we investigate the ability and the degree of the algorithm to preserve the data privacy. We mainly focus on the ability to protect the aggregation result of a community composed of sensor nodes. The experiment shows that the aggregation result can not be disclosed to people unless all of the sensor nodes in the community collude with each other.