Research On Privacy-preserving Association Rules Mining In Distributed Environment

With the rapid development of the era of big data,data mining is particularly important.As one of the common technical methods of data mining,association rule mining is widely used in various large data sets.Through the analysis of the mining results,the business model can be adjusted to maximize the benefits.However,excessive mining of user data will cause a large amount of private information to be leaked.This problem has caused people to worry about the practical application of data mining.Therefore,privacy-preserving association rule mining(PPARM)has become a significant research direction.In recent years,some researchers have proposed many PPARM schemes based on distributed data environment.However,the existing solutions still have shortcomings in terms of communication volume,privacy and practicality.This paper mainly does the following work for PPARM in a distributed data environment:1.A PPARM scheme of vertical data partition based on the intersection of multi-party sets is proposed.At present,existing solutions under the same model have a large number of communication rounds and security issues.In addition,the fact that there are multiple parties in the actual scene is not considered.To address the above problems,this paper improves on the multi-party set intersection scheme proposed by Freedman et al.,and proposes two protocols.Protocol 1 can achieve fewer communication rounds under the condition of achieving the same privacy.Protocol 2 can guarantee the same privacy without introducing a fully trusted third-party server.The improved association rule mining scheme will have the effect of smaller communication rounds and higher security.It is suitable for different outlets or different types of hospitals in the same community.Their customer sources are basically the same,but they serve the product types are different.2.A verifiable PPARM scheme based on weighted in the horizontal data partition is proposed.Existing solutions require multiple cloud servers for auxiliary mining,and do not consider data permissions and the problem of tampering during data transmission.Therefore,this solution is based on a cloud server and local support weights which can solve the problem of data permissions to get the rules of global association mining.At the same time,considering the insecurity of the communication channel,this scheme uses the elliptic curve Schnorr signature mechanism to verify the identity of the data sender in batches in order to ensure the accuracy of the data mining results.Our scheme can obtain the global association rule mining results through assistance by one cloud server on insecure channels.At the same time,the proposed scheme is suitable for the scenario of association rule mining on the electronic medical records of hospitals or clinics in different locations.