IOT Devices Using Supervised Machine Learning Models for Anomaly Based Intrusion Detection

Abstract

Identifying dangers and irregularities in any infrastructure is a growing problem in the Internet of Things (IoT) industry. IoT infrastructure is utilised more frequently across a wide spectrum of organisations, which increases the risks and attack methods. Attacks and anomalies that could lead an IoT system to malfunction include denial of service attacks, data type probing, malicious control, malicious operation, scans, surveillance, and improper configuration. This article studies the ability of several machine learning models to predict attacks and abnormalities on IoT devices. The f1 score, area under the receiver operating characteristic curve, accuracy, precision, recall, and precision are among the metrics used to assess performance. ANNs, decision trees, and random forests all shown performance with a 99.4% accuracy rate in the system's tests.