Abordagem comparativa entre técnicas de reconhecimento de padrões para classificação de pacotes de dados TCP/IP em uma rede de computadores ethernet

Abstract

This work represents a comparative study on the use of techniques of pattern recognition and classification of computer networking protocols into data packets through artificial neural networks (ANN) and support vector machines (SVM). We were approached aspects related to the architecture of neural networks, such as training algorithms, data samples, activation functions, concepts, and so on. Another point taken as the object of study of this work were the support vector machines, which is another pattern recognition technique whose characteristics such as machine learning algorithms, concepts, among others, were addressed. The study also developed explored the characteristics of the data packets analyzed, showing the structural differences between the packets on the basis of computer network protocols contained therein. Finally, it was developed some examples of artificial neural networks and SVMs, which were used for the pattern recognition. After testing, the performance of the two techniques were compared in order to present a conclusion as to which of these features more satisfactory performance in the case presented.