Autoencoder Latent Space Influence on IoT MQTT Attack Classification
- Maite García-Ordás 11
- José Aveleira Mata 1
- José-Luis Casteleiro-Roca 2
- José Luis Calvo Rolle 2
- María del Carmen Benavides Cuéllar 12
- Héctor Alaiz Moretón 11
-
1
Universidad de León
info
-
2
Universidade da Coruña
info
- Cesar Analide (ed. lit.)
- Paulo Novais (ed. lit.)
- David Camacho Fernández (ed. lit.)
- Hujun Yin (ed. lit.)
Publisher: Springer International Publishing AG
ISBN: 978-3-030-62362-3, 978-3-030-62361-6, 978-3-030-62364-7, 978-3-030-62365-4
Year of publication: 2020
Volume Title: Part II
Volume: 2
Pages: 279-286
Congress: Intelligent Data Engineering and Automated Learning – IDEAL (21. 2020. Guimarães)
Type: Conference paper
Abstract
IoT (Internet of Things) alludes to many different devices and systems connected to Internet, being 5 billion the number of these devices working around the world actually. The security policies applied to this kind of systems can be improve due to their behaviour, usually associated to their low price and low computing capacity.This work addresses the behaviour and impact of latent space of an auto-encoder for creating a classification model based on decision trees, in order to include it in a IDS (Intrusion Detection System) specialized in IoT environments. A validate IoT dataset, based on MQTT (Message Queue Telemetry Transport), has been used for applied the techniques implemented for extracting an optimal model oriented to detect the attacks over this protocol with a suitable results.