Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Log inSign up
Challenges
Datasets
Workspaces
Papers
Discussions
Leaderboard
Blog
Job Board
Q3AS

© 2026 Aqora Quantum S.A.S.

TermsPrivacyLegal Notice

arXiv:2610.25203

Simulation-Based Quantum Forensics for QKD: Machine Learning Attack Detection and Signature Extraction

Madhab Chandra Das, Sajid Ahmed, Uzair Hussain, Van Vung Pham

quant-phSubmitted Oct 18, 2026• Announced Oct 20, 2026• 0 citations
arXivHTMLPDF

Abstract

Quantum Key Distribution (QKD) supports secure key establishment under ideal protocol assumptions, but practical systems remain exposed to implementation-level attacks against optical channels, detectors, and controller behavior. This study presents Quantum Forensic Log (QF-LOG), a simulation-based forensic framework for analyzing aggregated QKD operational logs rather than raw quantum states. QF-LOG uses four observable features: quantum bit error rate (QBER), photon count, latency in milliseconds, and abort flag. The framework classifies normal operation and three simulated attack-related conditions: partial intercept-resend, detector-blind, and fiber-tap. A physics-inspired Generator-A dataset with 5 million balanced records was constructed using explicit mathematical generation models for all four features. A transparent threshold baseline, logistic regression, random forest, and XGBoost were evaluated on Generator-A. The full-feature learning models achieved macro F1-scores from 0.8211 to 0.8243 and outperformed the transparent two-feature comparator. Feature diagnostics showed that QBER contributed most, followed by latency in milliseconds and photon count, while no single feature reproduced the full feature set performance. Robustness was assessed on a separately generated Generator-B dataset of 5 million records that uses the same governing equations under shifted simulated operating parameters. Without retraining or tuning, their macro F1-scores decreased to values from 0.5645 to 0.5669. These findings show that QF-LOG can support interpretable forensic classification in a controlled synthetic setting, but stable internal evaluation does not establish reliability under changed simulated QKD operating conditions. The framework provides a documented basis for future evaluation across broader simulated settings and hardware-generated QKD logs.

AI Summary

Can aggregated QKD operational logs distinguish normal operation from attack-related conditions, and do classifiers remain reliable when operating conditions change? QF-LOG studies this question using four features: QBER, photon count, latency, and an abort flag. It considers normal operation, partial intercept-resend, a detector-blinding suppression proxy, and fiber tapping, without accessing raw quantum states. The evaluation uses two synthetic datasets of five million records each. Generator-A contains balanced classes generated from weak-coherent-pulse channel and detection equations, supplemented by configured attack, latency, and abort rules. Logistic regression, random forest, and XGBoost are evaluated using a stratified holdout, repeated five-fold validation, feature diagnostics, and comparison with a two-feature threshold baseline. Their Generator-A macro F1 scores are $0.8211$–$0.8243$, versus $0.7524$ for the baseline. QBER has the largest measured feature contribution, followed by latency and photon count; no individual feature matches full-feature performance. Generator-B retains the governing equations but shifts channel, detector, latency, and intervention parameters. Without retraining, classifier macro F1 falls to $0.5645$–$0.5669$. For random forest, the fraction of normal records classified as fiber tapping rises from $10.4\%$ to $51.1\%$; longer-fiber profiles yield macro F1 below $0.44$. The stated limitations include synthetic rather than hardware-calibrated data, four observables, three modeled attacks, independent frame classification without temporal modeling, and only one shifted generator. The results characterize classification within these simulations, not hardware attack detection or deployment readiness.

Some signs of AI-generated textIncrementalUnclear noveltyMissing comparisonStrong assumptions

Written by gpt-6.1-sol from the paper's full text. Automated: it can be wrong.

Comments
16 pages, 2 figures, 8 tables. Accepted and presented at the International Conference on Quantum Computing Infrastructure & Technologies (QUANCOM 2026), Trento, Italy, September 2026

Related papers

Forensics Investigation Framework for Advanced Threat Detection in Quantum-Era Networks

Owusu Nyarko-Boateng, Isaac Kofi Nti, Samuel T. Boateng +6 more

Objectives: To address the urgent need for forensic systems capable of detecting and analyzing advanced persistent threats in hybrid quantum-classical communication infrastructures, particularly those that may compromise quantum key distribution environments. Method: The study introduces a Quantum-Aware Forensics Investigation Framework, a multi-layered forensic architecture combining quantum telemetry, classical metadata analysis, and machine learning-driven threat classification. Experimental validation was conducted using a simulated testbed built with SimulaQron, Wireshark, and custom scripting tools. Various quantum attack scenarios were emulated, including intercept-resend, entanglement flooding, and control-plane hijacking. Machine learning models Random Forest, SVM, and Autoencoder were tested as standalone classifiers. A stacked ensemble model, with Random Forest and SVM as base learners and Logistic Regression as meta-classifier, was implemented for performance optimization. We used an experimentally generated, cross-layer dataset from a SimulaQron BB84 QKD emulation by combining quantum logs and classical control-plane captures under benign and scripted attacks such as intercept–resend, entanglement flooding, payload obfuscation, session hijacking, spoofing. Parameters studied were quantum - QBER, event inter-arrival jitter, event/count rate and classical - packet/flow statistics, inter-arrival mean/variance, latency proxy, TCP SYN/RST flags, byte-level Shannon entropy, with labels for benign vs. attack class. Findings: The standalone models achieved moderate performance on the held-out test set for Random Forest: ROC AUC = 0.93, F1 = 0.90, MCC = 0.86, Brier = 0.072; SVM (RBF): ROC AUC = 0.91, F1 = 0.88, MCC = 0.82, Brier = 0.081; Autoencoder (one-class): ROC AUC = 0.87, F1 = 0.83, MCC = 0.74, Brier = 0.094. By contrast, the stacked ensemble delivered perfect detection metrics for ROC AUC = 1.00, F1 = 1.00, MCC = 1.00, and Brier = 0.014. The study further emphasized the need for forensic systems to support explainability and continuous adaptability via Explainable AI and online learning with drift detection. Novelty: This study presents a cross-layer forensic framework for quantum–classical hybrid networks that fuses QKD telemetry with classical control-plane evidence and machine-learning analytics. Unlike prior work that treats these planes separately, our design unifies event-level QKD signals such QBER, arrival-time jitter with packet/flow features to produce timestamp-aligned, explainable alerts. In evaluation, the stacked-ensemble detector achieved perfect detection metrics for ROC AUC, F1, MCC and Brier on held-out data, which distinctly outperformed single-model baselines. The framework couples these gains with an XAI layer and an online, drift-aware learning loop, providing a scalable, auditable, and resilient foundation for forensic intelligence in the quantum era. Keywords: Quantum network forensics, QKD security, Advanced threat detection, Hybrid quantum-classical networks, Quantum-safe evidence, SimulaQron, Quantum cybersecurity

Indian Journal of Science and Technology•Dec 8, 2025•2 citations

Enhancing Cybersecurity using ML-Integrated Quantum Key Distribution Protocols

D Navyashree, P N Jayanthi

Quantum Key Distribution (QKD) leverages quantum mechanics to achieve information-theoretic secure key ex-change, resilient to both classical and quantum computational attacks. This paper proposes a novel framework that integrates QKD with machine learning-based intrusion detection in Software-Defined Networks (SDNs). Using BB84 and BB92 protocols, we evaluated resilience under intercept-resend and photon number splitting attacks. Furthermore, Random Forest and XGBoost classifiers were enhanced with QKD-generated keys to secure intrusion detection pipelines. The results show improved accuracy (99.8%) and F1-scores (85.12%) compared to classical IDS approaches, highlighting the novelty of combining QKD with ML-driven cyber defense. The study validates performance against benchmark survey works, emphasizing the role of QKD in advancing future-ready cybersecurity architectures.

International Conference on Computation System and Information Technology for Sustainable Solutions (CSITSS)•Nov 20, 2025•0 citations

Machine Learning Techniques for Enhancing Quantum Key Distribution

Ali Al-Kuwari, Safaa Alqrinawi, Lujayn Al-Amir +2 more

Quantum Key Distribution (QKD) offers theoretically unbreakable security by leveraging quantum mechanics. However, practical implementation is challenged by environmental vulnerabilities, noise, and hardware imperfections. Recently, Machine Learning (ML) has emerged as a powerful tool to address these limitations and enhance the real-world viability of QKD systems. In this survey, we review ML techniques applied to improve QKD security and performance across five applications. First, parameter optimization, covering signal calibration, polarization alignment, phase stabilization, modulation state tuning, and post-processing enhancements to maximize secure key generation and minimize error rates. Second, attack detection, where ML models identify and classify quantum threats such as photon-number-splitting and Trojan-horse attacks. Third, protocol selection, leveraging ML to dynamically choose QKD protocols based on operational conditions. Fourth, key performance prediction of core metrics such as Secret Key Rate (SKR) and Quantum Bit Error Rate (QBER). Finally, quantum network management, optimizing large-scale QKD deployments through intelligent routing, node management, and resource allocation. Performance improvements are evaluated using accuracy, reduced QBER, and increased SKR. While ML shows significant potential for finance, government, and defense applications, challenges remain in scalability, computational demands, and real-world testing. Ongoing work should focus on lightweight, generalizable models and standardized benchmarks for practical ML-enhanced QKD deployment.

quant-pharXiv:2603.07384•Mar 10, 2026•3 citations

AI-Augmented Quantum Key Distribution for Cloud Data Security: A Simulation Study of Aggregate Versus Block-Wise Eavesdropping Detection

Akheel Mohammad

Cloud infrastructures face a long-horizon threat from cryptographically relevant quantum computers, motivating both post-quantum cryptography (PQC) and quantum key distribution (QKD) as complementary defenses. QKD additionally offers a physical, information-theoretic signal of eavesdropping through the quantum bit error rate (QBER) of the sifted key. Prior work has applied machine learning across the QKD pipeline, chiefly for parameter optimization and physical-layer stabilization, but has given less systematic attention to whether learned, distribution-aware statistics improve on a simple aggregate-QBER threshold for detecting partial intercept-resend eavesdropping in a cloud key-rotation setting. We simulate BB84 key exchange under uniform and time-concentrated (“bursty”) partial interception across two monitoring-window lengths, and compare a static aggregate-QBER threshold against a logistic-regression detector that additionally uses block-wise QBER variance and maxima, at a matched false-positive rate. Across all four tested conditions, the aggregate threshold matched or exceeded the block-wise detector, while monitoring-window length had a substantially larger, quantifiable effect on detection sensitivity than the choice of detector. For this attack model, cloud QKD key-management pipelines therefore gain more from shortening or adaptively sizing the QBER monitoring window than from adding distributional machine-learning features to the detection statistic, which also helps explain why reported machine-learning gains in QKD concentrate in parameter optimization rather than eavesdropping detection Keywords: machine learning; Quantum Key Distribution; BB84; eavesdropping detection; quantum bit error rate; cloud security

International Journal of Technology & Emerging Research•Nov 15, 2026•0 citations

Dataset design and analysis for ML-assisted performance evaluation in QKD networks

Hassan Akbari, Behnam Shariati, Pooyan Safari +3 more

Progress in monitoring and performance analysis in quantum key distribution (QKD) networks has been influenced by the limited availability of well-structured and publicly accessible datasets. As a result, this limitation forces researchers to rely on specifically designed data collections, which not only makes reproducibility harder but also makes it difficult to compare methodologies and results across different studies. To address this gap, we present publicly available synthetic datasets specifically designed for the evaluation of QKD system performance at the optical network level. These datasets encompass a wide range of configurations for point-to-point QKD, including two operational scenarios: (i) dark fiber, where only QKD traffic is transmitted, and (ii) coexistence of classical and quantum channels over the same fiber in the C-band. Additionally, the datasets incorporate variations in detector technologies and utilize two distinct analytical models for performance evaluation. By establishing a consistent benchmark, these datasets aim to accelerate the development of machine learning (ML)-driven automation in QKD networks. To demonstrate their utility, we developed and assessed a regression-based ML model, focused on predicting QKD performance metrics. Our findings underscore the value of these datasets in enabling robust benchmarking of ML approaches, promoting reproducibility, and facilitating meaningful cross-study comparisons. Ultimately, this contribution is expected to support the broader adoption of ML-based solutions for the automation and optimization of QKD networks.

Journal of Optical Communications and Networking•May 19, 2026•1 citation

Machine-Learning based Detection of Quantum Attacks

L Rajasekar, Hansa, Riddhipan Manna +3 more

With advent of Quantum Computing, traditional cyber security mechanisms are failing. This paper discusses the effectiveness of certain machine learning models in detecting such quantum attacks on quantum systems. The dataset used is custom generated using qiskit and AerSimulator and it focuses on three states of system, normal state, side-channel attack state and intercept-resend attack state. Machine learning models like logistic regression, neural networks, SVM and XGBoost were trained on the dataset. The highest accuracy observed was 0.81 obtained using XGBoost. These findings highlight the challenges of quantum attack detection and suggest future research directions for improving classification accuracy.

International Conference on Intelligent Cyber Physical Systems and Internet of Things (ICoICI)•Sep 17, 2025•0 citations
Quantum Key Distribution (QKD) supports secure key establishment under ideal protocol assumptions, but practical systems remain exposed to implementation-level attacks against optical channels, detectors, and controller behavior. This study presents Quantum Forensic Log (QF-LOG), a simulation-based forensic framework for analyzing aggregated QKD operational logs rather than raw quantum states. QF-LOG uses four observable features: quantum bit error rate (QBER), photon count, latency in milliseconds, and abort flag. The framework classifies normal operation and three simulated attack-related conditions: partial intercept-resend, detector-blind, and fiber-tap. A physics-inspired Generator-A dataset with 5 million balanced records was constructed using explicit mathematical generation models for all four features. A transparent threshold baseline, logistic regression, random forest, and XGBoost were evaluated on Generator-A. The full-feature learning models achieved macro F1-scores from 0.8211 to 0.8243 and outperformed the transparent two-feature comparator. Feature diagnostics showed that QBER contributed most, followed by latency in milliseconds and photon count, while no single feature reproduced the full feature set performance. Robustness was assessed on a separately generated Generator-B dataset of 5 million records that uses the same governing equations under shifted simulated operating parameters. Without retraining or tuning, their macro F1-scores decreased to values from 0.5645 to 0.5669. These findings show that QF-LOG can support interpretable forensic classification in a controlled synthetic setting, but stable internal evaluation does not establish reliability under changed simulated QKD operating conditions. The framework provides a documented basis for future evaluation across broader simulated settings and hardware-generated QKD logs.
Can aggregated QKD operational logs distinguish normal operation from attack-related conditions, and do classifiers remain reliable when operating conditions change? QF-LOG studies this question using four features: QBER, photon count, latency, and an abort flag. It considers normal operation, partial intercept-resend, a detector-blinding suppression proxy, and fiber tapping, without accessing raw quantum states.
The evaluation uses two synthetic datasets of five million records each. Generator-A contains balanced classes generated from weak-coherent-pulse channel and detection equations, supplemented by configured attack, latency, and abort rules. Logistic regression, random forest, and XGBoost are evaluated using a stratified holdout, repeated five-fold validation, feature diagnostics, and comparison with a two-feature threshold baseline. Their Generator-A macro F1 scores are 0.82110.82110.8211–0.82430.82430.8243, versus 0.75240.75240.7524 for the baseline. QBER has the largest measured feature contribution, followed by latency and photon count; no individual feature matches full-feature performance.
Generator-B retains the governing equations but shifts channel, detector, latency, and intervention parameters. Without retraining, classifier macro F1 falls to 0.56450.56450.5645–0.56690.56690.5669. For random forest, the fraction of normal records classified as fiber tapping rises from 10.4%10.4\%10.4% to 51.1%51.1\%51.1%; longer-fiber profiles yield macro F1 below 0.440.440.44.
The stated limitations include synthetic rather than hardware-calibrated data, four observables, three modeled attacks, independent frame classification without temporal modeling, and only one shifted generator. The results characterize classification within these simulations, not hardware attack detection or deployment readiness.
Objectives: To address the urgent need for forensic systems capable of detecting and analyzing advanced persistent threats in hybrid quantum-classical communication infrastructures, particularly those that may compromise quantum key distribution environments. Method: The study introduces a Quantum-Aware Forensics Investigation Framework, a multi-layered forensic architecture combining quantum telemetry, classical metadata analysis, and machine learning-driven threat classification. Experimental validation was conducted using a simulated testbed built with SimulaQron, Wireshark, and custom scripting tools. Various quantum attack scenarios were emulated, including intercept-resend, entanglement flooding, and control-plane hijacking. Machine learning models Random Forest, SVM, and Autoencoder were tested as standalone classifiers. A stacked ensemble model, with Random Forest and SVM as base learners and Logistic Regression as meta-classifier, was implemented for performance optimization. We used an experimentally generated, cross-layer dataset from a SimulaQron BB84 QKD emulation by combining quantum logs and classical control-plane captures under benign and scripted attacks such as intercept–resend, entanglement flooding, payload obfuscation, session hijacking, spoofing. Parameters studied were quantum - QBER, event inter-arrival jitter, event/count rate and classical - packet/flow statistics, inter-arrival mean/variance, latency proxy, TCP SYN/RST flags, byte-level Shannon entropy, with labels for benign vs. attack class. Findings: The standalone models achieved moderate performance on the held-out test set for Random Forest: ROC AUC = 0.93, F1 = 0.90, MCC = 0.86, Brier = 0.072; SVM (RBF): ROC AUC = 0.91, F1 = 0.88, MCC = 0.82, Brier = 0.081; Autoencoder (one-class): ROC AUC = 0.87, F1 = 0.83, MCC = 0.74, Brier = 0.094. By contrast, the stacked ensemble delivered perfect detection metrics for ROC AUC = 1.00, F1 = 1.00, MCC = 1.00, and Brier = 0.014. The study further emphasized the need for forensic systems to support explainability and continuous adaptability via Explainable AI and online learning with drift detection. Novelty: This study presents a cross-layer forensic framework for quantum–classical hybrid networks that fuses QKD telemetry with classical control-plane evidence and machine-learning analytics. Unlike prior work that treats these planes separately, our design unifies event-level QKD signals such QBER, arrival-time jitter with packet/flow features to produce timestamp-aligned, explainable alerts. In evaluation, the stacked-ensemble detector achieved perfect detection metrics for ROC AUC, F1, MCC and Brier on held-out data, which distinctly outperformed single-model baselines. The framework couples these gains with an XAI layer and an online, drift-aware learning loop, providing a scalable, auditable, and resilient foundation for forensic intelligence in the quantum era. Keywords: Quantum network forensics, QKD security, Advanced threat detection, Hybrid quantum-classical networks, Quantum-safe evidence, SimulaQron, Quantum cybersecurity
Quantum Key Distribution (QKD) leverages quantum mechanics to achieve information-theoretic secure key ex-change, resilient to both classical and quantum computational attacks. This paper proposes a novel framework that integrates QKD with machine learning-based intrusion detection in Software-Defined Networks (SDNs). Using BB84 and BB92 protocols, we evaluated resilience under intercept-resend and photon number splitting attacks. Furthermore, Random Forest and XGBoost classifiers were enhanced with QKD-generated keys to secure intrusion detection pipelines. The results show improved accuracy (99.8%) and F1-scores (85.12%) compared to classical IDS approaches, highlighting the novelty of combining QKD with ML-driven cyber defense. The study validates performance against benchmark survey works, emphasizing the role of QKD in advancing future-ready cybersecurity architectures.
Quantum Key Distribution (QKD) offers theoretically unbreakable security by leveraging quantum mechanics. However, practical implementation is challenged by environmental vulnerabilities, noise, and hardware imperfections. Recently, Machine Learning (ML) has emerged as a powerful tool to address these limitations and enhance the real-world viability of QKD systems. In this survey, we review ML techniques applied to improve QKD security and performance across five applications. First, parameter optimization, covering signal calibration, polarization alignment, phase stabilization, modulation state tuning, and post-processing enhancements to maximize secure key generation and minimize error rates. Second, attack detection, where ML models identify and classify quantum threats such as photon-number-splitting and Trojan-horse attacks. Third, protocol selection, leveraging ML to dynamically choose QKD protocols based on operational conditions. Fourth, key performance prediction of core metrics such as Secret Key Rate (SKR) and Quantum Bit Error Rate (QBER). Finally, quantum network management, optimizing large-scale QKD deployments through intelligent routing, node management, and resource allocation. Performance improvements are evaluated using accuracy, reduced QBER, and increased SKR. While ML shows significant potential for finance, government, and defense applications, challenges remain in scalability, computational demands, and real-world testing. Ongoing work should focus on lightweight, generalizable models and standardized benchmarks for practical ML-enhanced QKD deployment.
Cloud infrastructures face a long-horizon threat from cryptographically relevant quantum computers, motivating both post-quantum cryptography (PQC) and quantum key distribution (QKD) as complementary defenses. QKD additionally offers a physical, information-theoretic signal of eavesdropping through the quantum bit error rate (QBER) of the sifted key. Prior work has applied machine learning across the QKD pipeline, chiefly for parameter optimization and physical-layer stabilization, but has given less systematic attention to whether learned, distribution-aware statistics improve on a simple aggregate-QBER threshold for detecting partial intercept-resend eavesdropping in a cloud key-rotation setting. We simulate BB84 key exchange under uniform and time-concentrated (“bursty”) partial interception across two monitoring-window lengths, and compare a static aggregate-QBER threshold against a logistic-regression detector that additionally uses block-wise QBER variance and maxima, at a matched false-positive rate. Across all four tested conditions, the aggregate threshold matched or exceeded the block-wise detector, while monitoring-window length had a substantially larger, quantifiable effect on detection sensitivity than the choice of detector. For this attack model, cloud QKD key-management pipelines therefore gain more from shortening or adaptively sizing the QBER monitoring window than from adding distributional machine-learning features to the detection statistic, which also helps explain why reported machine-learning gains in QKD concentrate in parameter optimization rather than eavesdropping detection Keywords: machine learning; Quantum Key Distribution; BB84; eavesdropping detection; quantum bit error rate; cloud security
Progress in monitoring and performance analysis in quantum key distribution (QKD) networks has been influenced by the limited availability of well-structured and publicly accessible datasets. As a result, this limitation forces researchers to rely on specifically designed data collections, which not only makes reproducibility harder but also makes it difficult to compare methodologies and results across different studies. To address this gap, we present publicly available synthetic datasets specifically designed for the evaluation of QKD system performance at the optical network level. These datasets encompass a wide range of configurations for point-to-point QKD, including two operational scenarios: (i) dark fiber, where only QKD traffic is transmitted, and (ii) coexistence of classical and quantum channels over the same fiber in the C-band. Additionally, the datasets incorporate variations in detector technologies and utilize two distinct analytical models for performance evaluation. By establishing a consistent benchmark, these datasets aim to accelerate the development of machine learning (ML)-driven automation in QKD networks. To demonstrate their utility, we developed and assessed a regression-based ML model, focused on predicting QKD performance metrics. Our findings underscore the value of these datasets in enabling robust benchmarking of ML approaches, promoting reproducibility, and facilitating meaningful cross-study comparisons. Ultimately, this contribution is expected to support the broader adoption of ML-based solutions for the automation and optimization of QKD networks.
With advent of Quantum Computing, traditional cyber security mechanisms are failing. This paper discusses the effectiveness of certain machine learning models in detecting such quantum attacks on quantum systems. The dataset used is custom generated using qiskit and AerSimulator and it focuses on three states of system, normal state, side-channel attack state and intercept-resend attack state. Machine learning models like logistic regression, neural networks, SVM and XGBoost were trained on the dataset. The highest accuracy observed was 0.81 obtained using XGBoost. These findings highlight the challenges of quantum attack detection and suggest future research directions for improving classification accuracy.