Privacy-Preserving Association Rule Mining Scheme Based On Homomorphic Encryption

With the advent of the era of big data,data mining techniques are widely used to obtain the potential value in data.As one of the most important data mining techniques,association rule mining is used to discover the implicit correlation information between transactions,which brings great business opportunities to enterprises.Due to the shortage of storage and computing resources,enterprises choose to outsource the association rule mining task to cloud service providers,however,cloud service providers are not fully trusted,and there is a risk that the raw data and the sensitive information in the mining results will be leaked.It will not only harm the interests of enterprises,but also violate the privacy of users.Therefore,it is necessary to formulate a secure and reliable privacy-preserving association rule mining scheme.In order to achieve the correct mining results while protecting data privacy,the main work of this article is as follows:1.We propose a privacy-preserving association rule mining scheme in a multi-key environment.Currently,there are few studies on privacy-preserving association rule mining in a multi-key environment,and most of the existing schemes increase the risk of message interception due to the strong decryption ability of the auxiliary cloud server.To address this problem,we adopt DT-PKC homomorphic cryptography scheme to encrypt the data,and split the strong private key into two partial private keys for the primary cloud server and the secondary cloud server respectively.It weakens the decryption capability of the auxiliary cloud server and effectively resists active interception attacks.In addition,in most of the association rule mining schemes based on homomorphic algorithms,it is necessary to add additional spurious transactions to prevent the support from being leaked to the cloud server.However,it not only introduces additional computational overhead,but also leaks the upper limit of support degree and some mining results.So we propose a secure support calculation method that can achieve a higher security level without introducing additional spurious transactions,and it will not disclose the sensitive information of support and all mining results.Our scheme also gives an effective solution to improve the computational efficiency for sparse data sets,which can be used by data miners to adjust the balance between user privacy and computational efficiency with higher flexibility and applicability.Finally,we analyze the correctness,security,and performance of the scheme to prove that the scheme is safe and feasible.2.We propose a privacy-preserving association rule mining scheme based on proxy reencryption.The scheme is also applicable to the multi-key environment.We realize the conversion of ciphertexts under different keys by proxy re-encryption technique,since only two cloud servers can collaborate to decrypt all ciphertexts,the decryption capability of one cloud server is weakened.So our scheme achieves higher security than other schemes.Considering from the perspective of practical application,we reduces the hard requirement without reducing security.The generation of keys in the scheme no longer relies on trusted institutions,which helps reduce the difficulty of application.In addition,our scheme adopts a special data storage method,and designs a secure cross-key equality detection protocol based on homomorphic and proxy re-encryption algorithms in order to judge the equality of plaintexts under ciphertexts.Then,we implement privacy-preserving frequent itemset mining and association rule mining based on the protocol.Finally,we specifically analyze the correctness as well as the security of the scheme and conduct experiments on the transaction dataset of the hypermarket to verify that our scheme is secure and reliable.