Statistical Characteristics Improvement of a Hardware Random Number Generator by a Software Method

Abstract

As a source of random numbers, hardware random number generators are often used, the operation of which is based on randomly changing parameters of various physical processes. The statistical characteristics of such generators do not always allow their use in the field of information security. To improve the statistical characteristics, various software tools for processing the output data of the generator are used. The purpose of this work is to study the possibility to improve the statistical characteristics of a hardware random number generator by software. The investigated hardware random number generator is based on the ND103L noise diode and has a random digital sequence of binary numbers at the output. To improve the statistical characteristics, the output stream of random numbers was processed using a software method based on the calculation of high-order finite differences. This method would allow one to get a more symmetrical distribution of random numbers, as well as increase the speed of their generation. After processing, the data from the generator under study have better statistical characteristics, which is confirmed by the NIST and Diehard tests, and the generation rate has also increased by more than 5 times. The results of this work may be useful to developers of hardware random number generators who need to improve the performance of the generator.