A Federated Learning Anomaly Detection Approach for IoT Environments

Basem Suleiman, Ali Anaissi, Wenbo Yan, Abubakar Bello, Sophie Zou, Ling Nga Meric Tong

Research output: Chapter in Book/Report/Conference proceedingConference proceeding (ISBN)peer-review

Abstract

The fast-growing development of smart home environments and the popularity of IoT devices have increasingly raised security concerns about anomalous behaviour and events. Traditional methods employ machine learning (ML) models that require sharing data collected from many IoT devices from smart homes so anomalous behaviour can be learned and detected. Such approaches introduce many problems in terms of privacy of sensitive IoT data to be shared over the network, models performance overhead, and scalability. We propose a novel Federated Learning Anomaly Detection (FLAD) approach for IoT smart home environments that address these problems. Our FLAD approach maintains the privacy of IoT data and faster training and detection performance by training several local models, instead of sharing it with a global model, on data collected from local devices. The local models require sharing only the learning parameter values with a global model which aggregates these values and sent them back to the local models to update their learning accordingly. Our experimental analysis of our FLAD approach on real IoT smart devices demonstrated very high accuracy (reaching over 99%) which is very comparable with the accuracy of the non-FL approach. Furthermore, our FLAD approach maintains the highest level of IoT privacy and faster model training time as it retains the IoT data within its local models and reduces network communication overhead, which has the potential to scale to a very large number of IoT devices.
Original languageEnglish
Title of host publicationProceedings of the 2nd International Conference on Advances in Computing Research, ACR 2024
EditorsKevin Daimi, Abeer Al Sadoon
PublisherSpringer Cham
Pages206-218
Number of pages13
ISBN (Electronic)978-3-031-56950-0
ISBN (Print)978-3-031-56949-4
DOIs
Publication statusPublished - 29 Mar 2024

Publication series

NameLecture Notes in Networks and Systems
Volume956 LNNS
ISSN (Print)2367-3370
ISSN (Electronic)2367-3389

Keywords

  • Federated Learning
  • IoT
  • Smart Home
  • Anomaly Detection
  • Deep Learning

Fingerprint

Dive into the research topics of 'A Federated Learning Anomaly Detection Approach for IoT Environments'. Together they form a unique fingerprint.

Cite this