IoT Device Identification Using a Meta-Ensemble Multi-Class Classifier

Gregory Davrazos, Theodor Panagiotakopoulos, Sotiris Kotsiantis, Achilles Kameas

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Device identification in the Internet of Things, is a hot research topic nowadays, offering advantages that enable the widespread adoption of IoT systems. Device identification can be done through various methods such as hardware IDs, finger-printing, and technical features. Machine Learning techniques through big data analysis offer not only an alternative but also an efficient way for device detection and identification. This paper applies a wide set of existing machine learning classifiers for IoT device identification, using a public dataset of IoT devices for multi-class classification. Evaluation results show that a meta-enseble classifier that utilizes the soft voting technique of the three top performing classifiers outperforms all the other models that were employed in our research.

Original languageEnglish
Title of host publication14th International Conference on Information, Intelligence, Systems and Applications, IISA 2023
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9798350318067
DOIs
Publication statusPublished - 2023
Externally publishedYes
Event14th International Conference on Information, Intelligence, Systems and Applications, IISA 2023 - Volos, Greece
Duration: 10 Jul 202312 Jul 2023

Publication series

Name14th International Conference on Information, Intelligence, Systems and Applications, IISA 2023

Conference

Conference14th International Conference on Information, Intelligence, Systems and Applications, IISA 2023
Country/TerritoryGreece
CityVolos
Period10/07/2312/07/23

Keywords

  • Device Identification
  • Internet of Things
  • Machine Learning Algorithms
  • Meta Ensemble
  • PyCaret
  • Soft Voting

Fingerprint

Dive into the research topics of 'IoT Device Identification Using a Meta-Ensemble Multi-Class Classifier'. Together they form a unique fingerprint.

Cite this