SDN-Net: A Novel Dataset for Intrusion Detection within SDN/NFV Network

This dataset, SDN-Net, is created specifically for AI-driven intrusion detection research within programmable SDN/NFV environments. SDN-Net offers the scale and attack diversity required to benchmark Machine Learning and Deep Learning models for intrusion and anomaly detection in centralized SDN/NFV network architectures. The dataset aggregates the original public datasets InSDN and SDN-Intrusion, followed by rigorous preprocessing, label harmonization, feature normalization, noise removal, and class balancing using SMOTE and SMOTE-Tomek links. SDN-Net includes 1,532,222 network flows, 79 engineered features, spanning 11 traffic categories: Normal, Dos, DDoS, Probe, Brute Force, XSS, BFA, Web-Attack, Botnet, SQL-Injection and U2R behaviors. The dataset preserves real-world imbalance to enable realistic ML/DL benchmarking, with synthetic balancing applied only in controlled research setups, not in this released raw file. Keywords: SDN, NFV, IDS, intrusion detection, dataset fusion, class imbalance.