[论文解读] Quantum AI for Cybersecurity: A hybrid Quantum-Classical models for attack path analysis

Jessica A. Sciammarelli, Waqas Ahmed|arXiv (Cornell University)|Jan 5, 2026

Quantum Computing Algorithms and Architecture被引用 0

一句话总结

该论文提出一个混合量子-经典管道，将经典网络安全特征映射到量子嵌入，通过变分电路实现，然后用经典SVM进行分类，并在UNSW-NB15数据集上与经典基线在数据丰富和数据匮乏条件下比较性能。

ABSTRACT

Modern cyberattacks are increasingly complex, posing significant challenges to classical machine learning methods, particularly when labeled data is limited and feature interactions are highly non-linear. In this study we investigates the potential of hybrid quantum-classical learning to enhance feature representations for intrusion detection and explore possible quantum advantages in cybersecurity analytics. Using the UNSW-NB15 dataset, network traffic is transformed into structured feature vectors through classical preprocessing and normalization. Classical models, including Logistic Regression and Support Vector Machines with linear and RBF kernels, are evaluated on the full dataset to establish baseline performance under large-sample conditions. Simultaneously, a quantum-enhanced pipeline maps classical features into variational quantum circuits via angle encoding and entangling layers, executed on a CPU-based quantum simulator, with resulting quantum embeddings classified using a classical SVM. Experiments show that while classical models achieve higher overall accuracy with large datasets, quantum-enhanced representations demonstrate superior attack recall and improved class separability when data is scarce, suggesting that quantum feature spaces capture complex correlations inaccessible to shallow classical models. These results highlight the potential of quantum embeddings to improve generalization and representation quality in cybersecurity tasks and provide a reproducible framework for evaluating quantum advantages as quantum hardware and simulators continue to advance.

研究动机与目标

• Motivate the use of hybrid quantum-classical learning for attack path analysis in cybersecurity.
• Evaluate whether quantum feature embeddings improve discrimination in security tasks, especially with limited data.
• Provide a reproducible workflow combining classical graph preprocessing with quantum encoding for intrusion detection.

提出的方法

• Feature-based graph abstraction of network flows into eight classical features mapped to eight qubits.
• Label-encode categorical features and Min–Max normalize for quantum encoding suitability.
• Encode features via angle encoding into variational quantum circuits with depth 2 and StronglyEntanglingLayers.
• Obtain quantum embeddings from Pauli-Z expectation values and classify with a classical SVM.
• Compare quantum-augmented pipeline against Logistic Regression, Linear SVM, and RBF SVM on full data and a 200-sample subset.
• Use PennyLane default.qubit simulator to ensure CPU-based reproducibility.

实验结果

研究问题

• RQ1 Can hybrid quantum-classical models outperform purely classical approaches for attack path analysis under limited training data?
• RQ2 Do quantum feature embeddings offer higher inter-class separability and better generalization in data-scarce cybersecurity settings?
• RQ3 How do quantum embeddings affect recall and precision for attack vs benign classes compared to classical baselines?

主要发现

• Classical models achieve around 69% accuracy on the full UNSW-NB15 dataset.
• In the 200-sample regime, Linear SVM and RBF SVM reach 80% accuracy, Logistic Regression 72.5%, while Quantum Embedding + SVM attains 64%.
• Quantum embeddings exhibit perfect recall (100%) for attack class in small data, but misclassify benign samples, yielding macro F1 of 0.39.
• Classical models show high attack recall (≈97–98%) but low benign recall (≈33–34%) on full data, indicating class imbalance in this feature space.
• Quantum embeddings provide strong separation of attack patterns with limited data, suggesting representational benefits but limited generalization due to shallow circuits and data imbalance.
• The study emphasizes a reproducible, CPU-based workflow using free tools and discusses future improvements in circuit depth, qubit count, and noise-aware evaluation.