P-STRIDE Dataset v2.0: A Live-Generated Multi-Threat Software-Defined Networking Dataset

P-STRIDE Dataset v2.0 is a live-generated, multi-class Software-Defined Networking (SDN) security dataset developed for machine-learning-based intrusion detection and mitigation research. The dataset was generated in a controlled Mininet–Ryu SDN environment through independently executed experimental runs. It contains 13,486 validated five-second observation windows collected from 1,189 independent runs across seven traffic classes: 0 – Normal1 – Denial of Service (DoS)2 – ARP Spoofing3 – Traffic Interception4 – Tampering5 – Port Scanning6 – High-Volume UDP Flooding Each record contains two metadata fields (run_id and window_id), 28 numerical traffic and behavioural features, and one ground-truth label. The features cover packet and byte statistics, traffic rates, flow behaviour, ARP activity, MAC–IP association changes, TCP flags, payload characteristics, and IPv4 Time-To-Live statistics. Version 2.0 is a comprehensively revised release. Exact duplicate records and invalid observations were removed, the feature-extraction pipeline was regenerated and validated, and the experimental runs were independently checked. Identical feature vectors originating from different independent runs were retained because they represent distinct validated observation windows. The dataset is designed for run-level machine-learning evaluation. Researchers are strongly advised to split the data using run_id rather than randomly splitting individual rows, in order to reduce information leakage between training and testing sets. This release supersedes the previous dataset version and represents the authoritative dataset used in the associated P-STRIDE study.