Network Flow Datasets Derived from ISCXVPN2016 and VNAT Using CICFlowMeter

This archive provides harmonized, labeled network flow datasets derived from two public benchmarks—ISCXVPN2016 and VNAT (MIT Lincoln Lab’s VPN/Non-VPN Network Application Traffic Dataset)—using CICFlowMeter for flow feature extraction.It includes: Original data files (ARFF, CSV, HDF5) as released by the dataset creators. Network flows extracted from PCAP files using CICFlowMeter, with consistent column naming and unified service category labels (e.g., Chat, Streaming, File Transfer, VoIP, Mail). Preprocessed data for direct use in machine learning workflows (NumPy .npz format with X_train and y_train). Jupyter notebooks implementing Decision Tree-based traffic classification and Bowerbird Courtship-inspired Wrapper Feature Selection (BBFS). The datasets cover both VPN and non-VPN traffic, annotated with service/application types, and are suitable for research in encrypted traffic analysis, feature selection, and benchmarking classification algorithms. Original sources: ISCXVPN2016: https://www.unb.ca/cic/datasets/vpn.html VNAT: https://www.ll.mit.edu/r-d/datasets/vpnnonvpn-network-application-traffic-dataset-vnat All code and data are provided for reproducibility and further research in network security and machine learning.