Towards Artificial Neural Network hashing with strange attractors usage / Jacek TCHÓRZEWSKI, Agnieszka Jakóbik // W: ECMS 2020 : proceedings of the 34th international ECMS conference on Modelling and simulation : [Wildau, Germany, June 9-12, 2020 : canceled conference] / ed. by Mike Steglich, [et al.]. — [Germany : European Council for Modelling and Simulation], [cop. 2020]. — (Proceedings (European Council for Modelling and Simulation) ; ISSN 2522-2414). — ISBN: 978-3-937436-68-5; e-ISBN: 978-3-937436-69-2. — S. 354-360. — Bibliogr. s. 360, Abstr. — J. Tchórzewski - dod. afiliacja: Cracow University of Technology
- AGHTchórzewski Jacek
- Jakóbik Agnieszka
|Data dodania do BaDAP||2020-11-20|
|Typ publikacji||materiały konferencyjne (aut.)|
|Czasopismo/seria||Proceedings (European Council for Modelling and Simulation)|
A broad variety of methods ensuring the integrity of data in the mobile and IoT equipment is very important nowadays. Hash functions are used for detecting the unauthorized modification of data and for digital signatures generation. Traditional hash functions like SHA-2 or SHA-3 have relatively high computational power requirements, therefore are not always suitable (or optimal) for devices with limited computational capacity or battery capacity. Instead, light cryptography hash functions may be used. They are processing data strings of the shorter length and offers simpler mathematical models as the basis of hash calculation. In this paper Artificial Neural Network (ANN)-based model hashing is proposed. Instead of using s-boxes or complicated compression function, a simple two-layered non-recurrent ANNs are used for hash calculation. In order to provide a very high quality of the randomization of the output, several different chaotic attractors were incorporated into ANNs training phase. ANNs output was tested with appropriate statistical tests and compared with hashes returned by traditional hashing methods. Using shorter hash length enables implementing those methods in the mobile and IoT equipment. Our approach allows merging the low complexity of ANN processing with the high-quality standards of cryptography hash functions. © ECMS Mike Steglich, Christian Mueller, Gaby Neumann, Mathias Walther.