Cybersecurity, artificial intelligence, machine learning, internet of things, MQTT
Abstract
This article presents an analysis of the performance of classification algorithms Decision Tree Machine Learning, Random Forest, Gradient Boosting, Naive Bayes, Sequential Neural Network and Multilayer Perceptron for Malicious Attack Detection in cyber-physical systems with Internet of Things (IoT) networks with communication-based on Message Queuing Telemetry Transport (MQTT). The dataset used in algorithm training is called MQTTset. The approaches used in this work were based on a study of the communication packages of that protocol, analysis of relevant features in that dataset and pre-processing techniques data. The results show that the algorithms classify well the network traffic that is under attack or not, which favours the security of a cyber-physical system.
