Kappa-PSO-ARTMAP Fuzzy: uma metodologia para detecção de intrusos baseado em seleção de atributos e otimização de parâmetros numa rede neural ARTMAP Fuzzy

In the last years have seen a strong increase in the 802.11 wireless local area network (WLAN) technologies use, and the security mechanisms implemented by amendments IEEE 802.11i and IEEE 802.11w have proven not very effective in combating attacks against availability of WLAN services. Intrusion detection systems emerge as a way to help computer networks in this combat against the deny of services. In this thesis it's proposed a model of intrusion detection called Kappa-PSO-Fuzzy ARTMAP, where initially the original database is pre-processed through a feature selection technique based on ARTMAP Fuzzy neural network and Kappa coefficient for reduce the amount of attributes, leaving only the most representative features. Then, apply the particle swarm optimization (PSO) technique in searching a set of criteria (choice parameter, ARTa module vigilance parameter, training rate and increase in the ARTa module vigilance paramater) employees in training attacks classifier, in order to maximize the accurate identification of classified samples. The intrusion detection algorithm used in the attacks classifier is the ARTMAP Fuzzy neural network. The performance of this new strategy is evaluated over three colleted databases respectively in a simulated wired network, infrastructured wireless network with WEP (Wired Equivalent Privacy) and WPA (WiFi Protected Access) encryption enabled and infrastructured wireless network with WPA2 (WiFi Protected Access version 2) encryption enabled. The obtained results in the Kappa-PSO-ARTMAP Fuzzy methodology demonstrate the IDS computational cost reduction without causing deterioration in the recognize capacity and have a scope of application over wireless and wired local area networks.