@article{Trilles2024a,
title = {Anomaly detection based on Artificial Intelligence of Things: A Systematic Literature Mapping},
author = {Sergio Trilles-Oliver and Sahibzada Saadoon Hammad and Ditsuhi Iskandaryan},
doi = {10.1016/j.iot.2024.101063},
issn = {2542-6605},
year = {2024},
date = {2024-04-01},
journal = {Internet of Things},
volume = {25},
pages = {101063},
abstract = {Advanced Machine Learning (ML) algorithms can be applied using Edge Computing (EC) to detect anomalies, which is the basis of Artificial Intelligence of Things (AIoT). EC has emerged as a solution for processing and analysing information on IoT devices. This field aims to allow the implementation of Machine/Deep Learning (DL) models on MicroController Units (MCUs). Integrating anomaly detection analysis on Internet of Things (IoT) devices produces clear benefits as it ensures the use of accurate data from the initial stage. However, this process poses a challenge due to the unique characteristics of IoT. This article presents a Systematic Literature Mapping of scientific research on the application of anomaly detection techniques in EC using MCUs. A total of 18 papers published over the period 2021–2023 were selected from a total of 162 in four databases of scientific papers. The results of this paper provide a comprehensive overview of anomaly detection using TinyML and MCUs. The main contributions of this survey are the fact that it aims to: (a) study techniques for anomaly detection in ML/DL and validation metrics used in the AIoT; (b) analyse data used in the estimation of models; (c) show how ML is applied in EC using hardware or software; (d) investigate the main microcontrollers, types of power supply, and communication technology; and (e) develop a taxonomy of ML/DL algorithms used to detect anomalies in TinyML. Finally, the benefits and challenges of this kind of TinyML analysis are described.},
keywords = {Anomaly detection, Edge computing, Internet of things, TidyML},
pubstate = {published},
tppubtype = {article}
}