Detecção de anomalias no tráfego MQTT de redes IOT utilizando técnicas de aprendizado de máquina

Abstract

In view of the growth of internet of things networks and devices and together with them several associated security flaws, research on machine learning techniques and intrusion detection systems in order to make secure networks and devices that communicate using the MQTT protocol. Therefore, it is necessary to analyze and describe the communication process and protocol of these devices, describe aspects about the use and operation of intrusion detection systems in internet networks of things and describe techniques and algorithms of machine learning necessary for the detection of attacks on networks. This work proposes a security system for real-time IoT networks under MQTT protocol using machine learning techniques to detect anomalies and create rules for detected attacks and an intrusion detection system to use the rules and detect recurrences of attacks based on them. The system was developed under a modular infrastructure, allowing the modification and use of different intrusion detection systems and extitbrokers. As part of the solution, the national Dojot platform was used, which has the objective of integrating IoT solutions, allowing device management, data processing and alert visualization. The experiments were carried out in two distinct networks through data collected from them. In view of this, it turns out that internet of things networks and devices can communicate inanely and the integration of machine learning algorithms and security mechanisms such as intrusion detection systems can be made to create a layer of security for networks and devices.