New Statistically Secure Pseudorandom Sequence Generation Methods

Abstract

This paper compares the statistical properties of several modern pseudorandom number generation algorithms. A method is also proposed for using various hash functions to improve the statistical properties of the considered pseudorandom sequence generations. Using cryptographic hash functions MD5, SHA3, Streebog, Blake3 to create pseudorandom number generators allows combining high generation speed with excellent statistical properties of output sequences. Due to the ability of hash functions to effectively “mix” data, eliminating any patterns, even if there are patterns in the internal state of the generator, the output is a statistically independent sequence. Modern optimized hash algorithms demonstrate data processing speed at the level of gigabytes per second even on conventional processors for example, GE + Streebog – up to 8.03 MB/s per core, which makes them suitable for modeling, rendering, statistical research and cryptography. At the same time, they often require fewer resources than traditional cryptographic generators based on block ciphers, providing comparable data quality. This approach is optimal for modern computing systems, where the balance between speed, randomness quality and energy efficiency is important.