Secure Computing Methods Over Encrypted Data In Cloud

With the prevalence of cloud computing,we can enjoy many benefits brought by cloud service.To enjoy the advantage of cloud service,huge data is increasingly outsourced to cloud.For preserving privacy,the outsouced data has to be encrypted by data owner before they are uploaded to cloud.Meanwhile,some users may want to take advantage of the powerful computation capability of cloud server to analyze the data stored in the cloud for extracting beneficial knowledge and patterns.Nevertheless,encryption will impede the functionality and performance of analyzing over the outsourced dataset in cloud.In this paper,we study the problem of how to complete the data analysis task on encrypted dataset without hampering data privacy.The main work of this paper is as follow:First,we investigate secure naive Bayesian classification on encrypted dataset in cloud and propose a secure scheme based on two non-colluding cloud server model,with additive homomorphic encryption and secure multi-party computation.Compared with exsiting works,our scheme outsources all the computation task of naive Bayesian classification to the cloud which cannot learn any useful information about data user's training dataset,parameters of Bayesian classifier and user's testing samples.Finally,we analyze its security and evaluate its computation complexity and communication through theoretical analysis and experiment.Second,we study secure comparison in encrypted form,which is one of the fundamental operations of many secure encrypted data analysis tasks,and propose a novel encrypted data comparison protocol based on a hybrid approach of Paillier cryptosystem and garbled circuits.Additionally,we design an efficient secure range query scheme based on the proposed secure comparison protocol.Our schemes reveal nothing about input and output result,and are provably secure under semi-honest model.And our proposed protocols can achieve higher efficiency,compared with existing schemes.Finally,theoretical analysis and experiment evaluations indicate the efficiency of our schemes.