Exploring the potential of deep learning and machine learning techniques for randomness analysis to enhance security on IoT

dc.authoridINCE, Kenan/0000-0003-4709-9557
dc.authorwosidINCE, Kenan/ABH-4111-2020
dc.contributor.authorInce, Kenan
dc.date.accessioned2024-08-04T20:54:51Z
dc.date.available2024-08-04T20:54:51Z
dc.date.issued2024
dc.departmentİnönü Üniversitesien_US
dc.description.abstractThe Internet of Things (IoT) is an incredibly growing technology. However, due to hardware inadequacy, IoT security is not improving to the same extent. For this reason, lightweight encryption algorithms have begun to be developed. This paper presents a method for assessing the security of Pseudorandom Number Generator (PRNG) generated binary sequences in a reasonable time using a pre-trained deep learning (DL) model. Due to their long execution times, Randomness Test Standards (RTSs) that include statistical tests that examine whether the sequences generated by PRNGs contain any patterns that cause cryptographic vulnerabilities are not suitable for running on edge devices with low processing capacities such as the IoT. We argue that every random sequence, even generated by a PRNGs that are classified as cryptographically secure, utilized in cryptographic applications should be used after successful results obtained from RTSs in every time. Therefore, an alternative method based on machine learning has been proposed to overcome the processing time problem of these test suites. The most utilized RTSs are NIST 800-22 Rev.1a, GB/T 32915-2016 and AIS 20/31. The 800-22 Rev.1a, which NIST has designated as a standard, has been observed to be the most referenced test standard in the literature. With this implementation, we sought to show that 15 statistical tests of the NIST 800-22 rev.1a environment can be modeled using DL. The application findings indicate that this modeling can serve as an alternative to the existing test environments. The average accuracy recorded throughout 15 tests was 98.64 percent. As a result, the trained model can be implemented even in edge computing devices with limited capability including IoTs.en_US
dc.description.sponsorshipIdot;nonu University Scientific Research Projects Department (SRPD) [FBG-2020-2143]en_US
dc.description.sponsorshipThis work was supported by the projects of the & Idot;nonu University Scientific Research Projects Department (SRPD) numbered FBG-2020-2143. The author would like to thank & Idot;noenue University SRPD for their valuable feedback.en_US
dc.identifier.doi10.1007/s10207-023-00783-y
dc.identifier.endpage1130en_US
dc.identifier.issn1615-5262
dc.identifier.issn1615-5270
dc.identifier.issue2en_US
dc.identifier.scopus2-s2.0-85177558293en_US
dc.identifier.scopusqualityQ1en_US
dc.identifier.startpage1117en_US
dc.identifier.urihttps://doi.org/10.1007/s10207-023-00783-y
dc.identifier.urihttps://hdl.handle.net/11616/101682
dc.identifier.volume23en_US
dc.identifier.wosWOS:001106715600001en_US
dc.identifier.wosqualityQ2en_US
dc.indekslendigikaynakWeb of Scienceen_US
dc.indekslendigikaynakScopusen_US
dc.language.isoenen_US
dc.publisherSpringeren_US
dc.relation.ispartofInternational Journal of Information Securityen_US
dc.relation.publicationcategoryMakale - Uluslararası Hakemli Dergi - Kurum Öğretim Elemanıen_US
dc.rightsinfo:eu-repo/semantics/closedAccessen_US
dc.subjectEdge computingen_US
dc.subjectRandomness testsen_US
dc.subjectIoT securityen_US
dc.subjectDeep learningen_US
dc.subjectMachine learningen_US
dc.titleExploring the potential of deep learning and machine learning techniques for randomness analysis to enhance security on IoTen_US
dc.typeArticleen_US

Dosyalar