Research Of Key Technologies On Privacy Protection For Secure Outsourcing Of Matrix Numerical Computation

Large-scale matrix numerical computation is one of the most common and fundamental problems in scientific and engineering.However,such problems are often too expensive to solve for resource-limited users.Cloud computing provides a new way to efficiently and economically solve such computing tasks.The clients can economically share the massive computational and storage resources,and the clients' computing is not restricted to their resources.They can outsource their matrix numerical computing tasks to the cloud,relieving the clients from computation costs.Although this new paradigm of computing brings much convenience and a pay-per-use manner for the clients,it also brings about many new security concerns and challenges.In the whole process of outsource computing,the clients have to preserve the input and output data privacy from the cloud server.The computational problems and their results usually contain sensitive information.To protect the privacy of sensitive data,the client should encrypt the sensitive data before outsourcing and decrypt the returned results from the cloud after outsourcing.The clients should have the ability to detect the correctness of the returned results from the cloud server.The cloud is not fully trusted.The cloud hides the detail operations instead of informing these operations for commercial purpose.The cloud cannot provide the guarantee on the quality of the computed results.In the outsourcing process,the quantity of calculation on the client side must be substantially smaller than performing the original computational problem on its own.There should not exist some other complicated computations for the client in the outsourcing process.To tackle these concerns,the main contributions of this thesis are as follows:1)We propose a technology on privacy protection for secure outsourcing matrix multiplication computation.Previous works for secure outsourcing matrix multiplication computation are mainly based on cryptography and disguise-based techniques.The existing schemes based on cryptography provide high cost of protocol computation and communication overhead.The existing schemes with disguise-based techniques cannot protect the privacy of data.To solve these problems,a technology on privacy protection for outsourcing matrix multiplication computation based on computational indistinguishability has been proposed in this thesis.The cloud server cannot deprive any sensitive information in polynomial time.We provide extensive theoretical analysis and experimental evaluation to show its high-efficiency and security.2)We propose a technology on privacy protection for secure outsourcing matrix determinant computation.Previous works for secure outsourcing matrix determinant computation are mainly based on cryptography and disguise-based techniques.The existing schemes based on cryptography provide high cost of protocol computation and communication overhead.The existing schemes with disguise-based techniques cannot protect the privacy of data.To solve these problems,a technology on privacy protection for outsourcing matrix determinant computation based on computational indistinguishability has been proposed in this thesis.In this technology,the original matrix expands a new matrix and then the new matrix is encrypted by privacy-preserving transformation matrix based on computational indistinguishability technique.It can protect the privacy of the input data.3)We propose an efficient,secure and non-iterative outsourcing technology of systems of linear equations.Previous works for secure outsourcing systems of linear equations are mainly based on iterative methods.The computation overhead of these schemes are linear to the number of iterations.To solve these problems,an efficient,secure and non-iterative outsourcing technology of systems of linear equations has been proposed in this thesis.In this technology,the privacy of the input data is protect by the matrix computation which is outsourced by the matrix multiplication outsourcing technology in this thesis.It can preserve the data privacy while efficiently outsourcing of systems of linear equations.In conclusion,to protect the data privacy,the thesis aims to secure outsourcing of matrix numerical computation in cloud computing.Based on computational indistinguishability,we propose a series of technologies on privacy protection for outsourcing matrix multiplication computation.The proposed technologies offer a higher level of privacy protection and achieve higher efficiency than the the previous works.