Uma arquitetura de referência para detecção de anomalias em SDN utilizando inteligência computacional

Abstract

Emerging technologies such as Cloud, 5G, Internet of Things (IoT) and edge computing it is necessary to control and connect on network millions of devices every day. Managing traditional networks with millions of devices is a complex task as it requires configuring routes on each device on the network. Software Defined Networking (SDN) helps simplify the configuration and management of a network with this number of devices as it has a centralized network controller. Although SDN is promising, it has challenges mainly related to security and fine analysis of network indicators to detect problems. Many studies have researched the use of computational intelligence (CI) to detect anomalies in SDN. This work defines a reference architecture to validate, promote and explain (using Explainable AI) any IC technique that best suits each of the different types of anomalies. This architecture is based on hexagonal microservices, with a unique information model based on the application and information framework and processes of Open Digital Architecture, from TM Forum. In the prototype were used two differents datasets to train seven machine learning algorithms and we proved the need to have a flexible architecture where it is possible that differents IC models can be added (or removed) for each specific scenario.