Research On Floating Point Oriented Privacy Protection Clustering Outsourcing Computing Scheme

Cluster analysis is a main task in data mining and data analysis.It is widely used in life,such as biological information processing,pattern recognition,digital forensics,information retrieval and target marketing.In recent years,with the rapid development of cloud computing,mobile terminal devices with limited storage /computing capacity often send large private data sets and local applications(such as clustering)to the cloud for outsourcing storage and computing.The existing ECS usually runs in a semi trusted or malicious environment,that is,the cloud is not trusted yet.Therefore,how to perform secure and efficient proxy operation on the encryption domain in the process of clustering has become one of the most important problems.Homomorphic encryption technology supports the specified operation of ciphertext without decryption,which is suitable for solving the privacy protection problem of outsourcing operation.K-means clustering is a classical clustering algorithm,which is widely used because of its simplicity and efficiency.In the existing homomorphic encryption K-means clustering schemes: 1)there is no effective method to deal with the storage and calculation of floating-point outsourcing ciphertext;2)There is no effective method to compare homomorphic ciphertext without decryption;3)It does not minimize the computing overhead of the client.In order to solve the above problems,this thesis proposes a clustering scheme based Privacy-Preserving Outsourced Calculation on Floating Point Numbers,which makes the following three contributions:1.Aiming at the problem that the previous homomorphic encryption K-means clustering algorithm does not support decimal calculation,this thesis focuses on studying and proposing a homomorphic encryption outsourcing computing clustering scheme for floating point numbers,and designs three security sub protocols according to the specific steps of k-means algorithm.In this thesis,a secure floating-point number storage method is introduced into the k-means algorithm.2.Aiming at the problem that the previous homomorphic encryption k-means schemes do not support ciphertext comparison,an SSD sub protocol supporting complete ciphertext distance comparison is designed by using the SEQ security comparison operation in PCPD sub protocol and SFPC floating-point security comparison operation in POCF sub protocol of Paillier algorithm supporting partial decryption,which can solve the minimum distance of k distances in ciphertext.3.According to the semi trusted characteristics of ECS,the security of the protocol is proved by using the UC model.Our privacy protection clustering algorithm is comprehensively evaluated on four data sets.The experimental data show that with the increase of the number of samples,the overhead of the cloud side and computer server side accounts for more than 99%,and the cloud allocates more computing tasks,bringing the smallest homomorphic load to the user side lacking computing resources.