Research On Key Technologies Of Big Data Privacy Protection Based On Matrix Transformation

With the rapid development of information technology industry,especially the innovation of big-data technology,the global society has formally entered the data-driven era.However,novel technologies such as cloud-computing and Internet-of-Things not only promotes the development of the big-data technology,but also brings endless security risks to data resources.In recent years,data leakage incidents occur frequently and on a large scale,and the situation of data security and privacy protection is not optimistic,which attracts more and more attention from all walks of life.The era of big-data is inseparable from the open sharing of data.How to realize the availability and invisibility of data has become an urgent problem to be solved.Based on the technical line of matrix transformation,this paper studies the key technologies of big-data privacy protection,and main contributions are as follows:1.We propose a privacy-preserving outsourcing matrix multiplication computation scheme based on the subtly designed invertible matrix called SDIM.As one of the most important basic operations,matrix multiplication computation(MMC)has been widely used in scientific and engineering community.And there are many big-data analysis algorithms can be converted into MMC tasks,such as linear regression and principal component analysis.So in this paper,we first take MMC as example and put forward a privacy-preserving outsourcing scheme based on the subtly designed invertible matrix called SDIM.This scheme consists of both multiplicative and additive perturbations,which corrects the inherent security weakness(i.e.the statistical information of zero elements in the original data is disclosed and not satisfying the computational indistinguishability)of RPM,an outsourcing scheme based on random permutation matrix.So the privacy security of data is enhanced.Moreover,SDIM scheme employs optimal matrix-chain multiplication so that it has the comparable efficiency with the RPM scheme,which meets the requirements of practical application.2.We give out privacy-preserving outsourced linear regression and principal component analysis protocols based on SDIM.Linear regression(LR)and principal component analysis(PCA)are typical machine learning algorithms in the field of big-data analysis.They have wide range of outsourcing computing needs in the cloud-computing environment.With the SDIM-based MMC outsourcing scheme,we propose privacy-preserving outsourced LR and PCA protocols,and they have comparable efficiency to the state-of-the-art privacy-preserving schemes based on matrix transformation,which confirmes that SDIM scheme has broad applicability in the secure outsourcing computing of complex MMC-based machine learning tasks.has high enough efficiency.3.We propose an efficient biometric identification in cloud computing with enhanced privacy security.As an important technology to ensure the access security of big-data system,biometric identification has been widely used in recent years.However,biometric identification not only provides a secure and convenient identification method,but also poses a severe challenge to the privacy security of individuals: due to the irreversibility of biometric features,the disclosure of biometric templates will cause irreparable damage to individuals.For the privacy protection in the biometric identification process,we propose a biometric identification scheme based on efficient matrix transformation.Based on the situation where the template matching can be regarded as comparing inner product of the encryption vector,the security level of the scheme is apparently improved with more randomness introduced into the matrix transformation process.Further,the properties of orthogonal transformation and vector inner product are employed to reduce the computational complexity of the scheme and makes it more suitable for the practical application of large-scale biometric database.As the comparision of security and efficiency shown,our scheme is more secure and more efficient than the state-of-the-art schemes based on matrix transformation.