Privacy-Preserving Bayesian Network Structure Learning On Distributed Heterogeneous Data

Data mining has long been an area of active database research. However, accompanying such benefits are concerns about information privacy. Because of these concerns, some people might decide to give false information in fear of privacy problem, or they might simply refuse to divulge any information at all. So privacy is an important issue in data mining and knowledge discovery.Privacy-preserving data mining provides a solution by creating distributed data mining algorithm in which the underlying data is not revealed.In this paper, we address a particular data mining problem to learn the structure of Bayesian network on distributed heterogeneous data. In this setting, three or more parties owning confidential databases wish to learn the structure on the combination of their databases without revealing anything about their data to each other.In this paper, the computational issue in the problem of learning Bayesian belief networks based on the minimum descrition length (MDL) principle. Two algorithms- exhaustive searches MDL (EX_MDL) and Branch&Bound MDL (B&B_MDL) have been implemented to find true network structure. Combined with a Private generalised homomorphic SSP protocol and homomorphic public key cryptosystems Bresson, I give two effective and privacy-preserving algorithms PP_EXMDL and PP_BBMDL to construct the structure of a Bayesian network for the parties' joint data. According to the output of the experiment, all of the two algorithem could generate accurate Bayes belief network, and in comparison with the previously known solution for this problemWY (Wright, Yang, 2004), these solutions provide better performance, full privacy and complete accuracy.