The Design And Implementation Of Matrix Disguising Library Toward Privacy Preserving Computation

In recent years,the Internet of Things(IoT)and cloud computing technology are being used more and more widely,providing a great convenience,but also brings serious and urgent privacy protection problems.In order to avail themselves of high performance computing power,a large number of resource-limited devices,especially those in IoT,require outsourcing intractable computation problems to public clouds.However,the inputs and the results of the computation problem may contain sensitive information,like business secrets or privacy information,which will not want to be let out.Thus,before a secure way of outsourcing computing is available,users would not be willing to use the extremely powerful and effective computing services provided by the cloud.This thesis aims to solve the problem of privacy protection for users when they outsource expensive computations to un-trusted parties.Through studying the existing matrix disguising schemes,the efficient implementations of general matrix multiplication as well as analyzing the users' privacy preserving requirement when consuming cloud computing services,this thesis proposes a matrix disguising method named FMD(Fast Matrix Disguising),which has high time and spaceefficiency and can tune the trade-off between disguising speed and protecting strength with a parameter.FMD disguises a matrix by multiplying it with a semi-random non-singular matrix which is composed of many bar-shaped sub-matrices.By adjusting the bar size of the sub-matrices,FMD can smoothly tune between high disguising speed and high privacy protection strength.Taking FMD as the core algorithm,NumPy and SciPy as basic scientific computing package,this thesis designs and implements a matrix disguising library toward privacy preserving computation,which supports multiple formats of data files and major platforms.The matrix disguising library is proved to be more efficient and memory-friendly then other schemes by the tests,especially when the disguising coefficient grows.The library provides a convenient,safe and tunable matrix disguising scheme,which allows users to enjoy the high efficiency and convenience of cloud computing with their data privacy preserved.