Host-Based Intrusion Detection System for IoT using Convolutional Neural Networks

Research output: Chapter in Book/Report/Conference proceedingsChapterpeer-review

Abstract

This paper proposes and analyses a lightweight Convolutional Neural Network (CNN) based anomaly detection framework for Internet of Things (IoT) devices. IoT security has become a massive concern in recent years. IoT devices form the backbone of much of the critical infrastructure we have today. From power stations to biomedical devices, there is the potential of heavy financial damage and loss of human life if they become compromised. As IoT adoption accelerates, the amount of cyberattacks on IoT devices increases substantially. Due to the resource constrained nature of IoT devices, no security solution addresses all concerns in the IoT field. By training models based on normal power consumption behaviour, a wide range of anomalies can be detected in the power time series data of the IoT device. The methodology proposed in this paper is generic in nature, making it applicable to every IoT device on the market. The work in this paper is implemented at the edge, on an ultra-low-power microcontroller.

Original languageEnglish
Title of host publication2022 33rd Irish Signals and Systems Conference, ISSC 2022
PublisherInstitute of Electrical and Electronics Engineers Inc.
ISBN (Electronic)9781665452274
DOIs
Publication statusPublished - 2022
Externally publishedYes
Event33rd Irish Signals and Systems Conference, ISSC 2022 - Cork, Ireland
Duration: 9 Jun 202210 Jun 2022

Publication series

Name2022 33rd Irish Signals and Systems Conference, ISSC 2022

Conference

Conference33rd Irish Signals and Systems Conference, ISSC 2022
Country/TerritoryIreland
CityCork
Period9/06/2210/06/22

Keywords

  • anomaly detection
  • CNN
  • HIDS
  • IDS
  • IoT
  • low-power
  • machine learning
  • sustainable security

Fingerprint

Dive into the research topics of 'Host-Based Intrusion Detection System for IoT using Convolutional Neural Networks'. Together they form a unique fingerprint.

Cite this