Pseudorandom Number Generators, Perfect Learning and Model Visualization with Neural Networks: Expanding on LFSRs and Geffe

Sara Boancă

2025

Abstract

The present paper explores the use of Artificial Neural Networks in the context of Pseudorandom Number Generators such as Linear Feedback Shift Registers and Geffe. Because of their hardware efficiency, variations of these generators may be used by IoT devices for security purposes. Testing to ensure security is essential, but it was observed that traditional test suites are too slow for the task. Machine Learning models, on the other hand, represent a faster alternative. While Artificial Neural Networks have been able to learn from these generators, improvements are still needed in terms of optimization and lowering domain knowledge. For that, the present paper focuses on the manner in which state of the art neural network approaches scale for a wider variety of Linear Feedback Shift Registers, including some of degree ≥ 100 and discusses the challenges that arise. Moreover, it proposes a novel Geffe learning approach that produces up to 100% testing accuracy and, based on that, promotes an additional optimization by capitalizing on model visualization and the ability of neural networks to learn deterministic functions to perfection. A comparative analysis is performed in order to show the superiority of the approach and an in-depth discussion is conducted on the possibility and implications of neural network perfect learning, particularly when coupled with model visualization. The obtained results can be regarded as incremental advances towards the creation of more robust neural network models to perform PRNG security evaluation for IoT devices.

Paper Citation

in Harvard Style

Boancă S. (2025). Pseudorandom Number Generators, Perfect Learning and Model Visualization with Neural Networks: Expanding on LFSRs and Geffe. In Proceedings of the 10th International Conference on Internet of Things, Big Data and Security - Volume 1: IoTBDS; ISBN 978-989-758-750-4, SciTePress, pages 117-128. DOI: 10.5220/0013298500003944

in Bibtex Style

@conference{iotbds25,
author={Sara Boancă},
title={Pseudorandom Number Generators, Perfect Learning and Model Visualization with Neural Networks: Expanding on LFSRs and Geffe},
booktitle={Proceedings of the 10th International Conference on Internet of Things, Big Data and Security - Volume 1: IoTBDS},
year={2025},
pages={117-128},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0013298500003944},
isbn={978-989-758-750-4},
}

in EndNote Style

TY - CONF

JO - Proceedings of the 10th International Conference on Internet of Things, Big Data and Security - Volume 1: IoTBDS
TI - Pseudorandom Number Generators, Perfect Learning and Model Visualization with Neural Networks: Expanding on LFSRs and Geffe
SN - 978-989-758-750-4
AU - Boancă S.
PY - 2025
SP - 117
EP - 128
DO - 10.5220/0013298500003944
PB - SciTePress