Diseño e implementación de IPS para redes WI-FI 802.11 ac
Cargando...
Autores
Mojica Sanchez, Julian Francisco
Director
Tipo de contenido
Document language:
Español
Fecha
Título de la revista
ISSN de la revista
Título del volumen
Documentos PDF
Resumen
Se presenta el diseño de un IPS (Intrusion Prevention System) para redes WI-FI 802.11 ac que emplea machine learning para determinar si una trama de datos pertenece o no a un ataque. Se usó Python y su librería Sklearn, se evaluó el rendimiento de los diferentes algoritmos de clasificación de dicha librería, para esto se empleó el AWID dataset y se midió exactitud y tiempo de entrenamiento requerido. Finalmente se escogió el algoritmo Decision Tree Classifier ya que se logró una exactitud de 99,99 % y un tiempo de entrenamiento de 20.8 segundos. Adicionalmente se planteó un modelo de teoría de juegos para la prevención de intrusiones en redes, a partir del análisis de dicho modelo se toman las medidas correspondientes para aumentar la seguridad del IPS. Mediante el modelo de teor´ıa de juegos se llegó a las ecuaciones que permiten determinar la probabilidad de ataque por parte de un atacante y la probabilidad de defensa por parte del IPS. Se realizó un análisis de diversos escenarios llegando a la conclusión de que el IPS debe tener un módulo de configuración de puertos capaz de evitar o entorpecer el descubrimiento de red y el análisis de vulnerabilidades.
Abstract: I present the design of an IPS (Intrusion Prevention System) for WI-FI 802.11 ac networks that uses machine learning to determine whether or not a data frame belongs to an attack. Python and its Sklearn library were used, the performance of the different classification algorithms of this library was evaluated, for this the AWID dataset was used and accuracy and training time was measured. The Decision Tree Classifier algorithm was finally chosen as it achieved an accuracy of 99.99 % and a training time of 20.8 seconds. Additionally, a game theory model is proposed for the prevention of intrusions in networks, based on the analysis of this model, the corresponding measures will be taken to increase the security of IPS. Using the game theory model, equations were arrived at to determine the probability of attack by an attacker and the probability of defense by IPS, and an analysis of various scenarios was performed, concluding that IPS must have a port configuration module capable of avoiding or hindering network discovery and vulnerability analysis.
Abstract: I present the design of an IPS (Intrusion Prevention System) for WI-FI 802.11 ac networks that uses machine learning to determine whether or not a data frame belongs to an attack. Python and its Sklearn library were used, the performance of the different classification algorithms of this library was evaluated, for this the AWID dataset was used and accuracy and training time was measured. The Decision Tree Classifier algorithm was finally chosen as it achieved an accuracy of 99.99 % and a training time of 20.8 seconds. Additionally, a game theory model is proposed for the prevention of intrusions in networks, based on the analysis of this model, the corresponding measures will be taken to increase the security of IPS. Using the game theory model, equations were arrived at to determine the probability of attack by an attacker and the probability of defense by IPS, and an analysis of various scenarios was performed, concluding that IPS must have a port configuration module capable of avoiding or hindering network discovery and vulnerability analysis.