Utilização de aprendizado profundo para detecção de anomalias em redes de computadores

Abstract

The idea of an intelligent and independent learning machine has fascinated humans for decades, today several factors have come together to make machine learning a reality. In this context of artificial intelligence, deep learning has been one of the most widely used machine learning tools, but its use in network anomaly detection, a relevant area, is still not explored if we compare with areas such as image recognition and voice identification. In this work we study the use of deep learning in detecting anomalies in networks by implementing and analyzing three models of deep learning-based classifiers, which were tested in cross-validation runs and in data from a database parallel to the one of its training. As a result, it can be said that the binary classifier, had a reasonable performance in relation to the other classifiers proposed in the work.