Randomness of Spritz via DieHarder testing

TL;DR

This paper evaluates the randomness of the Spritz cipher's keystreams using DieHarder tests, demonstrating strong statistical randomness with minimal weak results that improve with more samples.

Contribution

It provides the first comprehensive statistical analysis of Spritz's keystreams using DieHarder, confirming its high randomness quality.

Findings

All DieHarder tests passed with the initial sample size.

Weak test results diminished as sample size increased.

Spritz's keystreams exhibit high statistical randomness.

Abstract

RC4 is a stream cipher included in the TLS protocol, and widely used for encrypting network traffic during the last decades. Spritz is a possible candidate for replacing RC4. Spritz is based on a sponge construction and preserves the byte-oriented behaviour existing in RC4, but introduces an interface that provides encryption, hashing or MAC-generation functionalities. We present here the results obtained after applying several statistical tests on the keystreams generated by Spritz when used in the cipher mode. Our methodology makes use of 1024 keystreams of 2^25 bits. The algorithm was tested against the DieHarder test suite. None of the tests failed. Few tests produced weak results that were corrected when the number of samples increased.