The Impact of Post-processing Functions on Random Number Sequences

Abstract

In this paper we study and compare several postprocessing functions in order to provide a better overview of their impact on random number sequences. Using a reference benchmark provided by the NIST statistical test suite, we explore the impact post-processing functions have on the output of random number generators, and empirically analyze methods to improve the randomness quality. Finally, we provide guidelines for applying post-processing functions based on the quality of the raw input bitstreams and the expected results regarding bias correction and throughput.