Research On Secure Outsourcing Algorithm Of Face Recognition And Compressed Sensing Reconstruction In Cloud Environmen

With the advent of the era of big data,new and reliable cloud computing technologies have been rapidly developed.As an important cloud computing service model,outsourcing computing enables individuals or companies with little computing power to lease cloud servers with rich computing and storage resources to perform large-scale computing tasks.However,with the popularity of outsourced computing models,many security issues have arisen.Since users’ computing tasks often contain private information,the leakage of these data will cause huge losses to users.The uncontrollable nature of cloud servers makes them not necessarily completely trustworthy.For example,in order to save resources and earn more profits,the cloud server will not faithfully calculate the assigned tasks,or even maliciously forge a calculation result and return it to the user.These potential security issues hinder the large-scale deployment of outsourced computing models.Therefore,it is of high value to research outsourced computing schemes that protect the privacy of user data and can verify the malicious behavior of cloud servers.This thesis studies the outsourcing design of face recognition(FR)algorithm based on sparse representation classification(SRC)and medical image compressed sensing reconstruction(CSR)algorithm in cloud environment,since both algorithms contain complex convex optimization problems,this thesis proposes two encryption methods for convex optimization problems according to the specific applications of these two algorithms.Finally,this thesis proposes a security outsourcing scheme for these two algorithms respectively.(1)A secure outsourcing algorithm of SRC-based FR is proposed.FR as a typical biometric authentication technique,has been extensively deployed in the real world.However,the large-scale matrix operations or complex optimization problems that emerged in commonly used FR algorithms are overloaded for resource-constrained FR client.Regarding this,there is an important practical need to study cloud-aided FR algorithms.Nevertheless,the sensitivity of the data and machine learning model of the FR client and the incredibility of the cloud server make this intriguing computing paradigm suffer from many security challenges.This thesis focus on the popular SRC-based FR,an efficient and secure outsourcing design of the algorithm is proposed for the first time.The key technique ingredient involved in this design is a new norm-preserving matrix transformation employed in outsourcing convex optimization problem(7)1-minimization problem).This novel technique can well protect the critical privacy information in SRC-based FR and discern the malicious behavior of cloud servers with the best probability 1.Furthermore,extensive experiments performed on publicly available databases further corroborate the theoretical arguments in this thesis.(2)A secure outsourcing algorithm for medical image CSR is proposed.The wellknown CSR uses the sparse characteristics of the signal to obtain discrete samples with the compression(i.e.measurement)algorithm,and then perfectly reconstructs the signal through the reconstruction algorithm.Benefiting from the storage savings,the CSR has been widely used in the field of large-scale image processing.However,the convex optimization problem in the reconstruction process is computationally overloaded for resource-constrained clients.Therefore,designing a cloud-aided CSR algorithm becomes a hot topic.This thesis investigate the existing secure CSR outsourcing algorithms within a cloud environment,and propose a new privacy-enhanced and verifiable CSR outsourcing algorithm for online medical image processing service.Compared with previous work,the new design combining linear transformation,permutation and restricted random padding techniques can efficiently achieve stronger security.Finally,this thesis demonstrates the security and efficiency of the algorithm with rigorous theoretical derivation and comprehensive experimental analysis.