A Hybrid Quantum-Classical Network Intrusion Detection Dataset

The increasing scale and complexity of modern networks have intensified the need for robust and generalizable network intrusion detection systems (NIDS). While machine learning–based security solutions have shown promise, their effectiveness is often constrained by limitations in existing benchmark datasets, including severe class imbalance, outdated attack representations, and the absence of security-aware features that reflect emerging cryptographic and communication paradigms. In parallel, advances in quantum technologies are reshaping the cybersecurity landscape. Quantum Key Distribution (QKD) introduces fundamentally new security properties such as measurement disturbance, quantum bit error rate (QBER), and eavesdropping detectability—that are not captured by classical network traffic features. However, the lack of publicly available datasets integrating quantum-inspired descriptors has hindered research into hybrid quantum-classical cybersecurity models.

To address this gap, this dataset presents HQC-2026 (Hybrid Quantum-Classical 2026), the first publicly available network intrusion detection dataset enriched with quantum-inspired features. The dataset was generated in a controlled cyber-range environment at the Military Institute of Science and Technology (MIST), Bangladesh, and contains 14.69 million samples with perfectly balanced classes, covering seven attack categories and benign traffic. HQC-2026 combines 89 classical network flow features with 7 quantum-inspired features derived from simulated BB84 QKD protocols, resulting in a 96-dimensional feature space. The dataset includes SMOTE-balanced labels and reproducible OpenQASM 2.0 quantum circuits for transparent quantum feature generation. HQC-2026 is intended to support research in intrusion detection, hybrid quantum-classical machine learning, and quantum-inspired cybersecurity, providing a scalable and reproducible benchmark for future studies.