Hamming Net and LVQ Neural Networks for Classification of Computer Network Attacks: A Comparative Analysis

Abstract

This paper presents a comparative analysis of results obtained when applying Hamming Net and LVQ (Learning Vector Quantization) classifiers neural networks to recognize attack signatures in datasets. Strings similar to those located on payload field in computer networks packets are inserted in these neural networks for pattern classification. Since 2004, when it was presented for the first time, ANNIDA system (Artificial Neural Network for Intrusion Detection Application) has been improved. Although the very sufficient results presented by the application of Hamming Net neural network in this system, researches have continued to find other classification and data modeling methods in order to compare new results with those obtained from Hamming Net usage. As the LVQ neural network also uses basedcompetition techniques and presents architecture more simple than the Hamming Net architecture, it was decided to implement the LVQ to do the comparative tests. Tests results and analysis are presented in this paper, as well some proposals for future researches.