ESTUDIO COMPARATIVO DE TÉCNICAS DE ENTRENAMIENTO Y CLASIFICACIÓN EN SISTEMAS DE DETECCIÓN DE INSTRUSOS (IDS), BASADOS EN ANOMALIAS DE RED.

The main motivation of this investigation was the implementation of the Draper method applied to intrusion detection systems in different training and classification techniques in order to identify the best intrusion detection model with the objective of improving detection rates of attacks in computer network systems, using a procedure of selection of characteristics and different methods of algorithms of unsupervised trainings, in this case was used the technique INFO.GAIN identifying that the number of optimal characteristics is 15. Consequently, a neural network using a non-supervised learning algorithm (GHSOM, RANDOM FOREST, BAYESIAN NETWORKS, NAIVE BAYES, C4.5, LOGISTIC, PART AND NBTREE) for the purpose of classifying bi-class traffic automatically. obtained the best technique of training and classification using the selection technique In INFO.GAIN with 15 characteristics and cross validation 10 pligues, was the RANDOM FOREST technique.

• Autores: IBAÑEZ, KEVIN
• Tipo de recurso: Trabajo de grado de pregrado
• Fecha de publicación: 2016
• Institución: Corporación Universidad de la Costa
• Repositorio: REDICUC - Repositorio CUC
• Idioma: spa
• OAI Identifier: oai:repositorio.cuc.edu.co:11323/170
• Acceso en línea: https://hdl.handle.net/11323/170; https://repositorio.cuc.edu.co/
• Palabra clave: Dataset DARPA NSL-KDD; Sistema de Detección de Intrusiones; IDS; Técnicas de entrenamiento y clasificación; Técnicas de selección de características
• Rights: openAccess
• License: Atribución – No comercial – Compartir igual