Quantum Kernel Encoding Meets Deep Learning for Secure Software-Defined Networks.

Ahmed Elbialy, Ahmed Allam, Tarek Amr +6 more

Muhammad Afaq

Software-Defined Networks (SDNs) with their centralized control system and enhanced programmability requires a sophisticated approach to predict link states in complex network topologies. While predictive performance of traditional machine learning (ML) models remains high on static data, they struggle under dynamic and imbalanced network behavior. Although Quantum Machine Learning (QML) offers a promising solution by leveraging quantum-enhanced kernels to discover non-linear patterns, current quantum hardware constraints limit its standalone applicability. This paper presents an extended hybrid classical-quantum learning framework for SDN link state prediction under both benign (stable) and attack (compromised) conditions, integrating classical preprocessing with quantum kernel embedding via the ZZFeatureMap. A tunable parameter α0,1 enables dynamic interpolation between classical Radial Basis Function (RBF) and quantum kernels. On a stratified 5,000-sample subset of the InSDN dataset, which contains both normal and attack traffic flows, the hybrid Quantum Support Vector Machine (QSVM) achieves 85% accuracy, 0.92 ROC–AUC, and 0.73 Average Precision (AP). Scalability experiments on 20,000 samples confirm stable performance using Nyström-approximated kernels. Comparative evaluations with existing core classical, gradient boosting, and deep learning approaches highlight hybrid QSVM’s tunable expressivity, controlled computational scaling, and robustness to data distribution shifts, demonstrating its potential for future quantum-enabled SDN analytics.

Carlos Rubio García, Simon Rommel, Juan José Vegas Olmos +1 more

Pathan Shaheen Khan, Chandan Kumar

Software Requirement Risk Analysis is crucial for detecting potential failures early in the Software Development Life Cycle (SDLC). Existing machine learning (ML) and deep learning (DL) methods provide promising results, but they often struggle with capturing complex dependencies in requirement data. This paper introduces a novel hybrid Quantum-Classical Machine Learning (QML) framework to improve the accuracy and reliability of requirement risk classification. The dataset was preprocessed using label encoding, Synthetic Minority Oversampling Technique (SMOTE), ANOVA F-test (SelectKBest) for feature selection, and feature scaling to enable quantum circuit integration. Strongly entangling quantum circuits were implemented using the Qiskit framework to generate quantumenhanced features, which were fused with classical features. Four hybrid models-RF + QML, LGBM + QML, LR + QML, and Q-Kernel + MLP-were trained and evaluated. Experimental results show that the proposed Q-Kernel + MLP model achieved the best performance with $97.76 \%$ accuracy, $97.92 \%$ precision, $97.76 \%$ recall, and an F 1 -score of $97.77 \%$, outperforming classical baselines by up to $4 \%$. These findings highlight the potential of quantum-enhanced learning in software engineering, offering a scalable and robust solution for early risk detection and supporting more reliable project planning and software quality assurance.

Kouhei Nakaji, Hiroyuki Tezuka, Naoki Yamamoto

Abstract Recently, quantum neural networks or quantum–classical neural networks (qcNN) have been actively studied, as a possible alternative to the conventional classical neural network (cNN), but their practical and theoretically-guaranteed performance is still to be investigated. In contrast, cNNs and especially deep cNNs, have acquired several solid theoretical basis; one of those basis is the neural tangent kernel (NTK) theory, which can successfully explain the mechanism of various desirable properties of cNNs, particularly the global convergence in the training process. In this paper, we study a class of qcNN composed of a quantum data-encoder followed by a cNN. The quantum part is randomly initialized according to unitary 2-designs, which is an effective feature extraction process for quantum states, and the classical part is also randomly initialized according to Gaussian distributions; then, in the NTK regime where the number of nodes of the cNN becomes infinitely large, the output of the entire qcNN becomes a nonlinear function of the so-called projected quantum kernel. That is, the NTK theory is used to construct an effective quantum kernel, which is in general nontrivial to design. Moreover, NTK defined for the qcNN is identical to the covariance matrix of a Gaussian process, which allows us to analytically study the learning process. These properties are investigated in thorough numerical experiments; particularly, we demonstrate that the qcNN shows a clear advantage over fully classical NNs and qNNs for the problem of learning the quantum data-generating process.

Raied Ibrahim, Ibrahim Khider, Salaheldin Mohamed Ibrahim Edam