Proposta de um sistema de detecção e classificação de intrusão em redes de computadores baseado em transformadas wavelets e redes neurais artificiais

Abstract

As the Internet has become an enormous interconnected network, the information security today is very important to guarantee confidentiality, integrity and availability of computing resources. Advanced Intrusion Detections Systems (IDS) should be capable of identifying malicious actions that may compromise these guarantees, as quickly as possible. In this work, we present a hybrid approach for the IDS, with two different techniques: wavelets and artificial neural network. The wavelet is used to indicate to detect anomalous behavior on the network, while the neural networks are used to classify the attacks. A prototype was developed and evaluated data from simulation tests, on laboratory network and the KDD99 database. Besides the analysis of results of the other approaches, was also carried out a comparison with the learning vector quantization. Good results were obtained in all experiments, this demonstrating that the approach is very promising.