The research and testing of lightweight pseudorandom number generators for the Internet of Things

Abstract

The analysis of random sequences and random number generators is a rather specific task, but one or more
of the numerous test packages can be used to solve it. However, the analysis performed indicates that the existing testing
methods have a number of disadvantages, the solution of which can reduce the prerequisites for testing and improve
the accuracy of the results. The work is devoted to a rather urgent problem - the study of random number generators
operating on devices with limited resources, and short-length sequences for randomness. The paper considered the
construction of a physical model of a lightweight pseudo-random number generator. By using multivariate statistics
as a basis for testing, it is possible to better investigate a sequence for randomness by evaluating several characteristics
of the sequence simultaneously. Tests of multivariate statistics, which are based on the study of occurrences of several
patterns in a sequence, help to reveal hidden dependencies between data and low-quality generators. The main advantage of these tests is their effectiveness on short sequences, so they solve one of the problems of existing tests, facilitating the prerequisites for testing. The physical model of the IoT generator presented in the work, by its example, provides a broad overview of the factors and limitations that arise in the design of generators. The process of testing and optimizing the generator using tests of multivariate statistics illustrates the suitability of the software package for use and its integral role in creating a quality random number generator, especially for use in IoT devices. The software product that was created in this work can be used to solve a wide range of tasks, as has already been repeatedly noted. One of the most important, and indeed the one that can receive invaluable benefits in the field of application, is cryptography.