Privacy-preserving Distributed Data Fusion Based on Attribute Protection

Privacy-preserving distributed data fusion is a pretreatment process in data mining involving security models. In this paper, we present a method of implementing multi-party data fusion, wherein redundant attributes of a same set of individuals are stored by multiple parties. In particular, the merged data does not suffer from background attacks or other reasoning attacks, and individual attributes are not leaked. To achieve this, we present three algorithms that satisfy K-Anonymous and differential privacy. Experimental results on real data sets suggest that the proposed algorithm can effectively preserve information in data mining tasks.