ABSTRACT
The rapid expansion of the Internet of Things along with the recent 5G wireless communications technology has provided an excellent breeding ground for the proliferation of botnets. The extensive use of network traffic encryption makes it impossible to analyze packet payloads. This chapter proposes an architecture that combines Software Defined Network and network function virtualization to detect and mitigate cyberthreats in heterogeneous 5G network by using artificial intelligence applied to network flows. Machine learning algorithms, and particularly those based on deep learning, have achieved state-of-the-art in a wide range of difficult domains. These algorithms have proved to be suitable for complex pattern identification, as well as able to deal with unseen samples, thanks to their generalization capability. Due to the usual imbalance between anomalous and normal classes in anomaly detection datasets, accuracy is not considered the most adequate metric for detection performance. The chapter presents the necessary background with references to the main relative work. It describes the methodology followed in design of the experiments.
