Research On Several Issues Of Confidential Scientific Computing And Statistical Analysis

With the rapid development of the Internet and the Internet of Things,human beings have entered into the information age.In the information age,data has become a driving force to promote industrial development,and deeply affects the development of various industries.People depend heavily on data to make various decisions.Although it is easier and easier to obtain data,no entity can obtain all the data it needs,so data sharing becomes an inevitable choice.The data to be shared may contain many privacy data.Sharing these data without proper protection will lead to privacy disclosure and result in serious consequences and losses.Therefore,privacy preserving data sharing is an urgent problem to be solved.Privacy preserving data sharing is highly valued by cryptographic and information security community and has become a focus in the international cryptographic community.Secure multiparty computation（Secure multiparty computation,SMC）is the key privacy preserving data sharing technology.This paper studies several important security multi-party computation problems.Our main work is as follows:（i）.We study SMC of the intersection,union and threshold sum set of private multiple sets.Aiming at the above problems,we design corresponding coding schemes and use the Paillier encryption system to design SMC protocols for them.And the protocols are applied to confidential medical information sharing and other practical problems.（ii）.In the aspect of vector computation,we study how to privately compute equal component number and related threshold problem of multiple privacy vectors.We use the threshold ElGamal cryptosystem and a component-matrix coding scheme to construct SMC protocols on vector equal component number and related threshold problem.We further solve the privacy record links and other practical problems.（iii）.In the aspect of privacy preserving statistical analysis,using the Paillier encryption system and ElGamal threshold cryptosystem,we design two protocols to privately determine whether the confidence interval of a normal distribution is in a known interval and whether the confidence interval of the difference of means of two normal distributions is in a known interval,respectively.（iv）.In our research,we prove that by using simulation paradigm all protocols we designed are secure in the semi-honest model.We analyze the efficiency of the protocols,implement the protocols on a personal computer and compare our protocols with related works.Theoretical analysis and the experimental results show that the protocols are efficient.The protocols can be used as building blocks to solve other SMC problems.