ANALYSIS OF THE GENERALISABILITY OF THE PARKING TEST

Abstract

In 1995, George Marsaglia published a collection of randomness tests , "DIE HARD". One of thetests included there was the ,,Parking Lot Test", which was designed around the problem ofrandomly placing hyperspheres in d-dimensional space. Designed in this way, the randomness testworked very well for small amounts of data and was able to detect randomness where other testshad failed. However, due to its statistical properties, it could not cope with longer bit strings.The main objective of the present study was to parameterise the test under investigation so that itcould be used to test longer pseudorandom strings. The statistical properties of the test withalready different parameters were established empirically by performing a series of tests onstrings generated by pseudorandom generators constructed from Trivium and LFSR ciphersencrypted with the AES-128-ECB algorithm.The work also included an analysis of the possibility of optimizing the computational andmemory performance of the test. As a result of the research, the parameters of the test wereselected to allow an increase in the amount of analyzed data, while maintaining the computationaland memory complexity that allows its practical application.This article is an abstract of a Master's thesis of the same title9, which received a distinction in the2nd edition of the competition for the Marian Rejewski Award for the best Master's thesis in thefield of cyber security and cryptology.