Efficient secure arithmetic on floating point numbers

Abstract

Technology called cloud computing entrusts a huge amount of information processing that can not be done sufficiently with only its own computing resources by connecting internal and external computing resources. By using a method called secure computation, it is also possible to perform calculation processing while keeping the data concealed. In order to practically use secure computation, it is desirable that there is no load equal to or greater than the usual calculation processing time, independent of the data type to be handled. The data types dealt with in the existing secure computation studies are mainly integers, especially arithmetics on finite fields. But when using actual data such as statistics and geographic information, it is difficult to calculate them with integer arithmetic. Recently secure computation studies dealing with real numbers has increased. In this paper, based on Aliasgari’s method that achieve secure computation on floating point which can obtain sufficient precision, performance improvement of the methods is proposed and evaluated. As a result, improvement of performance is shown.