Modelo de detección de intrusiones en sistemas computacionales, realizando selección de características con chi square, entrenamiento y clasificación con ghsom

For organizations, information has become one of their most valuable assets, that is whyit is necessary to safeguard it,  by using different strategies of protection,  in order to avoid intruders access or any incident caused by data damage and misuse.  Then, this paper aims to assess the efficiency of a proposed model related to network intruders detection, using metrics of sensitivity, specificity, precision and accuracy.  This model uses DATASETNSL-KD DARPA, by selecting the most relevant featureswith CHI SQUARE and training a neural network,through a simulation process which use a non-supervised learning algorithm based on hierarchical organization maps,  in order to classify BI-CLASSnetwork automatically.  As a result, it was showed that using the GHSOM classifier with CHI SQUARE as features selection criteria, generate its best result : 15 features with precision, sensitivity, specificity and accuracy.