The practical value of quantum machine learning for network anomaly detection remains unclear under controlled and comparable benchmarking conditions. This study evaluates a trash-qubit-based Quantum Autoencoder (QAE) against seven classical anomaly-detection methods on UNSW-NB15 using shared data partitions, preprocessing steps, model selection procedures, and threshold calibration protocols. The experimental design included five independent seeds, three training set sizes, two feature budgets, and 12 QAE architectures, together with low-FPR analysis, finite-shot and synthetic-noise evaluation, less-constrained classical reference settings, and a cross-dataset assessment on TON_IoT. The Angle-RY encoding with a Real-Amplitudes Ring ansatz at depth L = 2 achieved the highest mean validation-selection AUPRC. On UNSW-NB15, the QAE was competitive in AUPRC and AUROC but remained behind the strongest classical baselines in F1-score, and its attack recall declined sharply under low-FPR constraints. The TON_IoT results further showed that relative model performance depended on the dataset. Under finite-shot and synthetic-noise conditions, anomaly-score rankings were largely preserved, whereas threshold-based decisions were more sensitive. Overall, the QAE showed competitive continuous-score discrimination on UNSW-NB15, but achieving reliable detection at low false-alarm rates remains a key operational challenge.
The practical value of quantum machine learning for network anomaly detection remains unclear under controlled and comparable benchmarking conditions. This study evaluates a trash-qubit-based Quantum Autoencoder (QAE) against seven classical anomaly-detection methods on UNSW-NB15 using shared data partitions, preprocessing steps, model selection procedures, and threshold calibration protocols. The experimental design included five independent seeds, three training set sizes, two feature budgets, and 12 QAE architectures, together with low-FPR analysis, finite-shot and synthetic-noise evaluation, less-constrained classical reference settings, and a cross-dataset assessment on TON_IoT. The Angle-RY encoding with a Real-Amplitudes Ring ansatz at depth L = 2 achieved the highest mean validation-selection AUPRC. On UNSW-NB15, the QAE was competitive in AUPRC and AUROC but remained behind the strongest classical baselines in F1-score, and its attack recall declined sharply under low-FPR constraints. The TON_IoT results further showed that relative model performance depended on the dataset. Under finite-shot and synthetic-noise conditions, anomaly-score rankings were largely preserved, whereas threshold-based decisions were more sensitive. Overall, the QAE showed competitive continuous-score discrimination on UNSW-NB15, but achieving reliable detection at low false-alarm rates remains a key operational challenge.
Abstract Anomaly detection in cybersecurity is a challenging task, where normal events far outnumber anomalous ones with new anomalies occurring frequently. Classical autoencoders have been used for anomaly detection, but struggles in data-limited settings which quantum counterparts can potentially overcome. In this work, we apply Quantum Autoencoders (QAEs) for anomaly detection in cybersecurity, specifically on the BPF-extended tracking honeypot (BETH) dataset. QAEs are evaluated across multiple encoding techniques, ansatz types, repetitions, and feature selection strategies. Our results demonstrate that an 8-feature QAE using Dense-Angle encoding with a RealAmplitude ansatz can outperform Classical Autoencoders (CAEs), even when trained on substantially fewer samples. The effects of quantum encoding and feature selection for developing quantum models are demonstrated and discussed. In a data-limited setting, the best performing QAE model has a F1 score of 0.87, better than that of CAE (0.77). These findings suggest that QAEs may offer practical advantages for anomaly detection in data-limited scenarios.
Quantum autoencoder (QAE) is a key tool in quantum machine learning and can be applied to areas such as dimensionality reduction and feature extraction [1]. Anomaly detection identifies deviations from normal patterns, which is crucial yet challenging due to the need to detect risks while minimizing false positives [2]. We propose a QAE-based anomaly detection scheme using an optical quantum neural network. Figure 1 (A) shows the basic structure of the QAE: the encoder compresses the input quantum state into a low-dimensional latent space, and the decoder reconstructs the original quantum state from the latent space. QAE-based anomaly detection works by learning the low-dimensional features of normal data and minimizing its reconstruction error. Normal data can be well reconstructed, while anomalous data exhibits a larger reconstruction error. When the reconstruction error of test data exceeds a set threshold, it is classified as anomalous. We constructed the QAE using an optical quantum neural network consisting of 50:50 beam splitters and controllable phase shifters [3]. Due to the unitary matrix property, only the encoder needs to be trained, and the decoder is simply the conjugate transpose of the encoder. To compress an N-dimensional quantum stateρintoKdimensions, efficient compression can be achieved by minimizing the expectation value of the projection onto the trash state, given byC(θ,ϕ)=Tr(PU(θ,ϕ)ρU+(θ,ϕ))[4]. Here,P=(IK⊕∣0⟩⊗N−K⟨0∣⊗(N−K))represents the projection onto the trash state,IKis the K-dimensional identity matrix,∣0⟩⊗(N−K)isthe trash state,Trdenotes the trace,U(θ,ϕ)is the encoder with adjustable parameters, andU†(θ,ϕ)is the decoder. We only need to train the encoder during the training process, and the decoder is simply its conjugate transpose.
Quantum convolutional neural networks (QCNNs) have become increasingly popular in quantum machine learning (QML) due to their efficient parameterization and hierarchical representation of quantum information. Anomaly detection is an important machine learning task with applications across a wide range of domains, including scientific data analysis. In this work, we adapt a QCNN architecture into a quantum autoencoder (QAE) framework for reconstruction-based anomaly detection. The models are trained in a semi-supervised manner on normal samples to reconstruct feature-extracted and dimensionally reduced time-series data, with reconstruction error used as an anomaly score. We investigate two quantum convolutional autoencoder architectures that differ in their treatment of latent information: a hierarchical architecture in which information remains distributed across the circuit and a bottleneck-based architecture in which information is explicitly compressed and reconstructed using additional decoder qubits. The size of the quantum latent space is varied to study its influence on reconstruction accuracy and anomaly detection performance. The approaches are benchmarked against both a variational quantum circuit and a comparable classical baseline using a real-world exoplanet anomaly-detection dataset. Results indicate a trade-off between latent-space size and model capacity, while also suggesting that explicit latent-space compression through a quantum bottleneck can improve anomaly detection performance relative to architectures that retain information throughout the circuit.
There is a pressing need for anomaly detection algorithms that can handle large amounts of diverse traffic data due to the quick growth of networked systems and the sophistication of contemporary cyberattacks. Despite the widespread use of conventional machine learning techniques for network intrusion detection, feature representation constraints, dataset imbalance, and the computational expense of scaling to vast problem spaces sometimes restrict their efficacy. In the meantime, new developments in quantum-assisted computation have opened up exciting possibilities for using quantum state changes to improve the feature learning process. Using the NF-UNSWNB15-v2 benchmark, this research presents HybridQCNN, a unique quantum-classical computing framework for binary classification of network traffic.The proposed architecture combines standard preprocessing techniques with an 8-qubit Parameterized Quantum Circuit (PQC) featuring multiple variational layers, where IsingXX and IsingYY entangling operations are applied to encode interfeature correlations, and a subsequent fully connected classical layer handles the final binary decision. A rigorous leakage-free data partitioning protocol was enforced, ensuring that all preprocessing transformations—including feature encoding, normalization, and threshold computation—were derived exclusively from the training subset, thereby preventing any information leakage from held-out partitions during evaluation. Experimental validation on a class-balanced subset of 60,000 traffic records yielded a classification accuracy of 97.33%, an F1-score of 0.9730, and a malicious traffic detection precision of 98.38%. Furthermore, the model achieved an AUC-ROC of 0.9861, reflecting strong class-level discriminability. Despite this performance, the entire model contains only 489 trainable parameters, underscoring the parameter efficiency of hybrid architectures relative to conventional deep learning baselines. These results suggest that integrating variational quantum circuits with classical inference modules provides an effective and scalable solution for network anomaly detection.
This work investigates variational quantum autoencoders (QAE) for one-class anomaly detection under strict parametric constraints, shifting the evaluation focus from absolute performance to performance–capacity tradeoffs. We implement a pure QAE baseline and a Hybrid QAE with classical compression/decoding around a quantum latent block, comparing both against classical baselines - AE, VAE, Isolation Forest, and One-Class SVM - on the NSL-KDD and ECG5000 benchmarks. The central contribution is a fairnessoriented evaluation protocol combining standard detection metrics (AUC-ROC, AUC-PR, and F1) with a performance-density measure (AUC-ROC per 1,000 trainable parameters) and a fair-budget comparison against a compact classical AE. On NSL-KDD, the QAE baseline achieves 0.9612 AUC-ROC with only 476 parameters versus 0.9679 for a classical AE with 5,580 parameters, while attaining 2.0193 AUC-ROC/1k parameters against 0.1735 for the classical AE - an order-of-magnitude improvement in parametric efficiency. Under a fairbudget setting, the QAE baseline also surpasses the compact classical AE on NSL-KDD (0.9612 vs. 0.9405). In contrast, ECG5000 favors classical methods, indicating domain dependence rather than universal quantum advantage. Overall, quantum and hybrid autoencoders are not universally superior, but deliver competitive anomaly detection with remarkably high parametric efficiency.
Quantum machine learning (QML) for network intrusion detection (NIDS) is routinely reported to reach near-perfect accuracy, yet the most rigorous studies find that well-tuned classical models remain competitive, and that apparent quantum gains may be artefacts of classical dimensionality reduction and implicit regularisation rather than genuine quantum effects. We ask not whether a quantum model can post a high accuracy, but how quantum the advantage really is. We present a unified, reproducible QML-IDS benchmark evaluating hybrid variational quantum circuits and quantum-kernel SVMs against five honestly-tuned classical baselines across four standard NIDS datasets (NSL-KDD, UNSW-NB15, CICIDS2017, NF-ToN-IoT-v2) under one leakage-controlled protocol, with an equal-budget feature view, imbalance- and calibration-aware metrics with significance testing, and a simulated NISQ noise sweep. We introduce a quantum-attribution audit (parameter-matched classical controls, a random-feature kernel, and a regularisation sweep) that quantifies how much of any gain is genuinely attributable to the quantum component. Tuned classical models (Random Forest, XGBoost) match or exceed the quantum models on aggregate detection on every dataset, and the audit attributes this to classical preprocessing and regularisation rather than quantum effects. Two advantages survive false-discovery-rate correction: the quantum-kernel SVM out-ranks its direct classical surrogate (a random-feature kernel) on AUPRC and ROC-AUC, and a small four-qubit hybrid out-detects the best classical baseline at the 1% false-positive operating point on the distribution-shifted NSL-KDD task (p = 0.005, BH q = 0.030). Code, seeds, and splits are released; our contribution stands whether quantum wins, ties, or loses.