True Randomness using the Randomness in Natural Phenomenon

Abstract

Patterns and ability to discover them, makes a system predictable. All of the modern day computers are finite state machines, which make them quite predictable in their nature. Yet, there are certain applications and cases which require a high level of unpredictability. This unpredictability, or sometimes referred to as non determinism, is characterised by Randomness of a system. Randomness attributes to the lack of knowledge on the causality behind a specific output, even if the system is known to fine details. Randomness is a key requirement in certain critical applications such as cryptography, simulations and so forth. Randomness at the most abstract level is bi-fold. They are namely true randomness and pseudo randomness. True randomness is the ideal form, which is existing in the surrounding as various phenomenon, such as lightning, thermal noise, Brownian motion of particles and so forth. Even though these are available and truly random, most of the times they are far from practicality within the environment of a computing device, due to various reasons such as difficulty to measure and feed to the device, very low rates of change, generated bit strings being inadequate in size and so on. Often, capturing true randomness that is existing in the surroundings requires expensive hardware devices which are not feasible in the context of personal computing. Therefore, true random sources are mostly used to provide the initial seeds to a pseudo random generator. Pseudo randomness on the other hand is the method of generating randomness by deterministically transforming an initial state called a seed). Often these systems generate randomness which has equal statistical qualities as true randomness, at much faster rates. Yet, output of the pseudo random generators almost all the times, repeat after a certain number of iterations. This is known as the period of a pseudo random generator and considered a weakness that is inherent in pseudo random generators. This research study focuses on using the random variables within a typical system environment to generate randomness which is void of the inherent weaknesses of pseudo randomness and close to true randomness in terms of statistical quality. In order to achieve the said, feasibility of using Floating point, an existing number representation scheme was evaluated. Initially, a conceptual model was composed to address the different issues in generating randomness that is close to true randomness. Then each of the stage in the conceptual model was addressed with different possible strategies. For the generation of bits, a new generator is proposed which uses the concepts of floating point representation at its core. Then, the performance of the proposed model was tested using the Statistical Test Suite for Randomness provided by National Institute for Standards and Technology (NIST) by bench marking the results against some commonly used and recent pseudo random generators.