Privacy Preserving Distributed Data Mining

摘要: OF DISSERTATION PRIVACY PRESERVING DISTRIBUTED DATA MINING Privacy preserving distributed data mining aims to design secure protocols which allow multiple parties conduct collaborative while protecting the privacy. My research focuses on and implementation of privacy two-party based homomorphic encryption. I present new results in this area, including for basic operations two fundamental protocols. propose a number additive secretsharing scheme derive relationship between secret its shares, with we develop e cient comparison division public divisor also inverse square root protocol Newton's iterative method hence solution problem. In addition, exponential Taylor series expansions. All these are implemented using multiplication can be used particular, tasks: linear regression EM clustering. Both work arbitrarily partitioned datasets. The is provably semi-honest model, clustering discloses only iterations. provide proof-of-concept C++, Paillier cryptosystem.