Seguridad para redes IOT usando machine learning : análisis y comparación de distintos algoritmos de machine learning para la detección de intrusión en redes IOT.

Abstract

IoT devices acquire a large amount of personal information, which is why this information must be protected from possible threats. The security of IoT devices can be compromised due to their low computational power. This security can be achieved by processing information through Machine Learning algorithms. The analysis ranges from the preprocessing of the information and continues with the training and evaluation, with the purpose of complying with a Machine Learning model capable of identifying good and malicious frames without sacrificing the computational resources of the IoT devices. In this project, a data set containing network frame information will be used. These plots, for the most part, are good and a certain amount are malicious. The information will be purified, balanced, normalized and the dimensionality reduced, in order to obtain an efficient data set to train the Machine Learning algorithms. The already trained Machine Learning algorithms will be evaluated with performance and resource consumption metrics. These metrics will allow us to determine which Machine Learning model is feasible for IoT systems.