Research On Homomorphic Encryption Algorithm Based On Association Rule Mining

With big data becoming the basic strategic resource of the country,many enterprises and organizations hope to obtain economic benefits from massive data and provide convenience for users.For most enterprises and organizations,they do not have the ability to deal with massive data.Therefore,outsourcing data mining tasks to cloud computing service organizations can effectively solve the problems of insufficient computing and storage capacity,resource utilization and capital investment of these enterprises and organizations.However,there are new security risks.The core problem is that data owners do not want their sensitive information to be known by others.Therefore,privacy security is one of the main bottlenecks in the application of massive data mining technology.As one of the most important technologies of data mining,frequent itemsets and association rules mining are widely used in commodity recommendation system,financial industry prediction and medical data analysis.In these applications,the requirement of privacy protection is high.In this paper,the classic Paillier homomorphic encryption algorithm is enhanced to satisfy the homomorphic multiplication,so that the complex computation of ciphertext can be processed,and a complete privacy preserving association rule mining scheme is proposed.The main work is as follows:(1)Based on the original Paillier homomorphic encryption algorithm,we strengthen it.By using the properties of additive homomorphism and multiplicative homomorphism of Paillier,we realize the multiplicative homomorphism calculation to meet the requirements of fully homomorphic encryption,and design a security comparison scheme.The improved Paillier homomorphic encryption scheme is called FH Paillier homomorphic encryption scheme.Compared with the original Paillier homomorphic encryption algorithm,it can be applied in more complex computing scenarios while retaining the original security.(2)Based on FH Paillier encryption algorithm,a complete privacy protection outsourcing association rule mining scheme is proposed.The scheme can deal with frequency analysis attack by inserting false data to disturb the information of original database.FH Paillier homomorphic encryption algorithm is used to encrypt the data tag value to ensure that sensitive information is not leaked.Mreclat,a parallel algorithm based on Map Reduce framework,is used to mine frequent itemsets and association rules.The algorithm does not need communication between sites and has less I / O times.The proposed scheme is demonstrated and analyzed repeatedly through theory and experiment.The experiment shows that the scheme has high security and efficiency.(3)Aiming at the problem that Paillier encryption algorithm may be overloaded in encryption and decryption,this paper improves Paillier encryption algorithm based on Chinese remainder theorem,and proposes an improved encryption and decryption scheme.Specifically,the Chinese remainder theorem is used to improve the operation efficiency of encryption and decryption process,and the theoretical correctness and accuracy of the improved algorithm are proved.Experimental results show that the scheme has high decryption speed.