The Method and Criterion for Quality Assessment of Random Number Sequences

Abstract

The authors analyze the dependence of the uniformity of distribution of signs of empirical autocorrelation function with respect to the number of overlapping symbols of intervals into which a sequence of random numbers is divided. A feasible “threshold” of overlap is established, below which the signs of the autocorrelation function are uniformly distributed. The concept of barrier function is defined and used to develop a criterion for the quality assessment of random number generators. The technique of its application and its implementation for several well-known generators are presented.