Abstract : "The advent of Software-Defined Networking (SDN) and Network Function Virtualization (NFV) represents significant advances in networking and telecommunications. Their emergence reveals a change in networking paradigms that are more programmable and adaptable. Nevertheless, these technologies present new difficulties, especially regarding security. Conventional security measures are less effective in these complex, dynamic environments. In fact, through using Artificial Intelligence (AI) abilities, we can develop more robust and adaptive intrusion detection systems that can effectively identify and overcome challenging issues. Combining AI with SDN/NFV may assure a more secure network architecture by significantly enhancing security measures.In this paper, we propose the integration of Deep Learning (DL) and Machine Learning-based (ML) detection algorithms with the concepts of SDN and NFV, to enhance intrusion detection. DL/ML-based techniques have notable success in detecting novel and emerging types of network intrusions when supplied with sufficient and relevant training data. Anomaly detection methods rely heavily on data, which includes features that are representative of system behavior.Data fusion is essential in this context, as it combines information from multiple sources to provide a more comprehensive and accurate understanding of network activities.To address the challenge of finding relevant training data, this study introduces the SDN-Net dataset, which is the result of combining two pre-existing SDN-oriented datasets.SDN-Net has 79 features, 11 categories of traffic, and more than one and a half million rows of observations. By providing a dataset including a wide range of normal and abnormal network behaviors, we facilitate the training and testing of DL/ML models capable of detecting network threats and improving the security and resilience of SDN/NFV networks. The results demonstrate how, in terms of accuracy and efficiency, our AI-based intrusion detection model surpasses more conventional methods. We reached 99.99\% precision and 99.9\% accuracy when using different DL and ML models."
Index terms : Artificial Intelligence, Deep Learning, Security, Intrusion Detection, Software Defined Network, Dataset Fusion