Abstract: The privacy preserving computation research area is considered. The problem appear when one party have confidential data and need to do intense computations over that data, and computations will be done by the second party, which may be supposed being untrusted. So the content of the raw data should be kept private from the second party during the computations. Therefore these data are to be encrypted before sending them to the second party. Two possible solution scenarios are considered – one in physical and the second in theoretical levels. Physical level solution assumes some hardware integration and reorganizations. Theoretical level solution is based on cryptographic approach (homomorphic encryption).The main idea is to encrypt data in such a way that the owner, after getting the results of computations over the encrypted data, will be able to get the results on original data only by decrypting the received results. The paper brings description and analyzes of such known schemas. The final outcome is that practical cryptographic tools today are really not ready to be applied on privacy preserving computations, so that the way of solution is the use of heuristic data analyses models and algorithms that replace original data with synthesized data. Considering preparatory, this article is followed by the base research part where synthetic data generation is considered on base of CART algorithm and clustering type computational algorithms.

Keywords: Privacy preserving computations, homomorphic encryption, synthetic data generation.

ACM Classification Keywords: H.1 Information Systems – Models and principles, I.2.0 Artificial intelligence.

Link:

BRIEF ANALYZIS OF TECHNIQUE FOR PRIVACY PRESERVING COMPUTATION1

Levon Aslanyan, Vardan Topchyan, Haykaz Danoyan

http://www.foibg.com/ijicp/vol01/ijicp01-01-p01.pdf