TITLE:
Ensuring Quality of Random Numbers from TRNG: Design and Evaluation of Post-Processing Using Genetic Algorithm
AUTHORS:
Jose J. Mijares Chan, Parimala Thulasiraman, Gabriel Thomas, Ruppa Thulasiram
KEYWORDS:
True Random Number Generators, Genetic Algorithms, Auto-Correlation, Entropy, Power Spectral Density
JOURNAL NAME:
Journal of Computer and Communications,
Vol.4 No.4,
April
8,
2016
ABSTRACT: Random numbers generated by pseudo-random
and true random number generators (TRNG) are used in a wide variety of
important applications. A TRNG relies on a non-deterministic source to sample
random numbers. In this paper, we improve the post-processing stage of TRNGs
using a heuristic evolutionary algorithm. Our post-processing algorithm
decomposes the problem of improving the quality of random numbers into two
phases: (i) Exact Histogram Equalization: it modifies the random numbers
distribution with a specified output distribution; (ii) Stationarity Enforcement:
using genetic algorithms, the output of (ii) is permuted until the random
numbers meet wide-sense stationarity. We ensure that the quality of the numbers
generated from the genetic algorithm is within a specified level of error
defined by the user. We parallelize the genetic algorithm for improved
performance. The post-processing is based on the power spectral density of the
generated numbers used as a metric. We propose guideline parameters for the
evolutionary algorithm to ensure fast convergence, within the first 100
generations, with a standard deviation over the specified quality level of less
than 0.45. We also include a TestU01 evaluation over the random numbers
generated.