Obtención de un modelo de detección de ataques de denegación de servicios en redes SDN en mininet mediante aprendizaje de máquina

Abstract

This paper analyzes some characteristics of the architecture of software-defined networks that present security vulnerabilities and the most common attacks that they can suffer, paying special attention to denial-of-service attacks in the control layer, specifically those attacks that saturate the controller, which is the central element or brain of SDNs, and presents a model that allows predicting such attacks using a classification tree type algorithm. The generation of dataset for training is generated from a network topology designed and simulated in the mininet software and complemented with an external dataset; the evaluated controller is RYU, which is also the development framework where the model for attack detection is implemented. Evaluations on the accuracy and efficiency of the generated model against possible attacks and modifications of the proposed scenario are shown and finally, conclusions and future work are presented.