Comparação da eficiência de modelos de redes neurais artificiais na detecção de intrusões em redes de computadores

Abstract

Artificial Neural Networks (ANNs) are used in several areas, finding patterns, optimizing data, predicting actions, among other uses. In computer network security, the use of ANNs has been employed in order to detect intrusions, preventing network attacks. However, there are numerous models of ANNs, so that each type of attack can have a model that achieves better results than others. In this work, ANNs of 5 ANN models are implemented, which are used to detect 7 types of attacks. The results of this work indicate, for each ANN and for each attack, the amount of false positives and negatives, accuracy and execution time. Thus, the results help network security software designers in choosing the most appropriate ANN model.