Research On Privacy-preserving Division And Vector Dot Product Computation

The Internet of Things achieves the exchange of information between physical objects and the network by connecting a large number of physical devices to the Internet.Thus,it is widely used in many areas to provide various services such as smart home,smart healthcare and smart cities.The emergence of Io T technology has not only stimulated revolutionary developments in such areas as industry and scientific research,but also greatly benefits our daily lives.However,a large amount of data generated by the Io T systems need be analyzed and processed based on a cloud platform due to the limited resources of a single physical device.Nevertheless,cloud servers are honest but curious,and are not fully trusted in many application scenarios.They may leak data privacy and threaten data security.In order to avoid data privacy leakage,users are inclined to upload encrypted data to a cloud server,and expect the cloud server to complete the correct processing of the encrypted data without obtaining any information about the original data and middle results.Eventually the cloud server returns the correct encrypted data processing and analysis results,which can only be decrypted by data requesters who satisfy the access policy.Lots of schemes based on secure multiparty computing and homomorphic encryption have been proposed to achieve secure data processing at the cloud.However,most of them can only meet the basic operational requirements of ciphertext such as addition and multiplication over encrypted data due to the high computation cost and low efficiency.And there still lacks research on special algebraic operations such as the division computation of encrypted integers and the dot product operation of encrypted vectors.To tackle the above issues,this thesis focuses on following research contents.First,this thesis proposes two privacy-preserving division schemes with flexible access control based on a dual server model,which applies homomorphic re-encryption system to complete data processing and attribute-based encryption algorithm to realize fine-grained access control.And we extend the proposed division schemes of encrypted integers to support division of multiple data types including fixed-point numbers and fractions.Moreover,we propose two secure greatest common divisor schemes to achieve the simplification of encrypted fractional arithmetic results.In addition,this thesis illustrates the protection of data privacy in proposed division schemes through security analysis.Finally,we further elaborate the advantages of our division schemes in big data processing by simulation and comparison with existing work.Second,this thesis also proposes two privacy-preserving dot product schemes based on homomorphic re-encryption system and attribute-based encryption algorithm,which realizes the secure outsourced computing of two encrypted vectors.Similarly,our proposed dot product schemes in this thesis can not only calculate the correct encrypted dot product with encrypted input vectors,but also achieve flexible access control to the encrypted dot product.To show the superiority of our dot product schemes in privacy protection and flexibility,we analyze security by referring to the dual-server system model and the corresponding security model,and conduct simulation and performance comparison with the existing related work.