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:2608.04047

Beyond the QBER Threshold: A Temporal QBER Based Machine Learning Framework for Multi Attack Detection in BB84 QKD

Isha, Deepak Singh, Devesh Kumar, S. K Pal, Praful Hambarde, Amit Shukla

Published Aug 4, 2026• 1 citation
arXivHTMLPDF

Abstract

Conventional BB84 Quantum Key Distribution (QKD) systems rely on a fixed 11% Quantum Bit Error Rate (QBER) threshold to detect eavesdropping. However, stealthy attacks can remain below this threshold while still compromising channel security. This paper proposes a temporal QBER based machine learning framework for detecting and classifying eavesdropping attacks in BB84 QKD systems. Rather than relying on average session level QBER, the framework extracts 63 physics-informed temporal features capturing burst behavior, temporal instability, basis dependent asymmetry, and QBER loss interactions. Random Forest, XGBoost, and Support Vector Machine with a Radial Basis Function kernel (SVM-RBF) classifiers are evaluated on seven eavesdropping attacks and a normal channel scenario under noisy and lossy conditions. Averaged over ten independent runs, XGBoost achieves the best performance with 88.01% (0.47%) accuracy and a macro F1 score of 0.8803, while SVM-RBF performs comparably, confirming the robustness of the proposed features. Evaluated as a binary attack-versus-normal detector for comparison with conventional monitoring, a fixed 11% QBER threshold achieves only 25.82% accuracy with a False Negative Rate (FNR) of 0.8477, whereas the proposed framework reduces the FNR to 0.0198, substantially improving detection of stealthy attacks that evade threshold-based monitoring. SHapley Additive exPlanations based (SHAP) explainability shows that physics-informed temporal and channel derived features are highly discriminative for identifying eavesdropping strategies. These results demonstrate that temporal QBER driven machine learning provides an accurate, explainable, and practical framework for multi attack security monitoring in BB84 QKD systems.

DOI
10.48550/arxiv.2608.04047

Related papers

Unsupervised ML Based Anomaly Detection for Securing BB84 QKD Against Intercept-Resend Attacks and Channel Noise: OC-SVM, iForest, and Ex-iForest

H A Sharath, Jayavrinda Vrindavanam V

Quantum Key Distribution (QKD) using the BB84 protocol offers the potential for information-theoretic security based upon the principles of quantum mechanics; however, eavesdropping mechanisms such as intercept-resend attacks and ambient noise sources that degrade quantum channel fidelity may render such systems exploitable. We present a oneclass machine learning approach to detect two categories of threats present within BB84 quantum channels: (i) interceptresend eavesdropping committed by an adversarial actor (Eve), and (ii) six types of channel noise. We have trained three unsupervised anomaly detectors (One-Class Support Vector Machine (OC-SVM), Isolation Forest (iForest), and Extended Isolation Forest (Ex-iForest)) on a dataset consisting of 1000 instances (940 clean and 60 anomalous) and tested each of them on separate test sets of 100 instances. With respect to the detection of adversarial attacks, Ex-iForest outperformed both OC-SVM and iForest; it achieved 95.00% accuracy, precision, recall, and F1-score, and when detecting channel noise, it achieved an overall accuracy, precision, recall, and F1-score of 92.00%. Feature importance analysis confirmed that the Quantum Bit Error Rate (QBER) and the fidelity of polarization measurement were the two most discriminative features for adversarial detection while fluctuations of photon detection counts and timing jitter were the two most informative features for channel noise discrimination. These results suggest that ExiForest is a viable, robust, and efficient solution for real-time anomaly detection systems in BB84-based quantum communication systems.

International Conference on Innovations in Cybersecurity and Data Science (ICICDS)•Jun 1, 2026•1 citation

Neural Network-Driven Statistical Analysis for Enhanced PNS Attack Detection in Practical BB84 Systems

Avinash Upadhyay, Amit Singhal, Ritu Raj Singh +3 more

Quantum Key Distribution (QKD) techniques provide theoretically invulnerable security for cryptographic key exchange. However, hardware limitations, such as the use of weak coherent photon sources instead of a single-photon source, make it vulnerable to photon-number splitting (PNS) attacks. In a PNS attack Eve selectively intercepts multi-photon pulses while allowing single-photon pulses to pass through. PNS attacks do not affect the quantum bit error rate (QBER), rendering traditional QBER-based monitoring ineffective. Hence, to detect the attack in such cases, it becomes imperative to statistically monitor parameters such as photon counts at the detector, SKR, and shifted key length to identify the attack. In this paper, we propose two complementary deep learning-based frameworks for detection of the PNS attack in the QKD setup. The first operates directly on the raw photon sequence recorded at the receiver, and is realised as a 1-D convolutional network as well as a deeper residual 1-D CNN. The second operates on a compact 13-dimensional feature vector derived from photon statistics, secure-key length, and protocol scalars, and is realised as both a baseline feed-forward network and a deeper residual ANN with skip connections. Together, these four models let us study attack detection across two distinct input modalities. We further establish where each model is best suited by analysing accuracy alongside inference latency, throughput, and data-cost trade-offs, providing guidance for both offline auditing and real-time edge monitoring. Further, to train the model, we introduce a physics-based simulated dataset featuring both attack and non-attack scenarios. Proposed models achieve up-to 83.4% classification accuracy with minimal false positives and negatives, demonstrating robust performance in identifying compromised quantum channels. This work represents a significant advancement in machine learning-assisted quantum security, providing a practical tool for enhancing the security of real-world QKD systems against stealthy photon-level attacks.

•Jul 15, 2026•0 citations

Neural Network-Driven Statistical Analysis for Enhanced PNS Attack Detection in Practical BB84 Systems

Avinash Upadhyay, Amit Singhal, Ritu Raj Singh +3 more

Quantum Key Distribution (QKD) techniques provide theoretically invulnerable security for cryptographic key exchange. However, hardware limitations, such as the use of weak coherent photon sources instead of a single-photon source, make it vulnerable to photon-number splitting (PNS) attacks. In a PNS attack Eve selectively intercepts multi-photon pulses while allowing single-photon pulses to pass through. PNS attacks do not affect the quantum bit error rate (QBER), rendering traditional QBER-based monitoring ineffective. Hence, to detect the attack in such cases, it becomes imperative to statistically monitor parameters such as photon counts at the detector, SKR, and shifted key length to identify the attack. In this paper, we propose two complementary deep learning-based frameworks for detection of the PNS attack in the QKD setup. The first operates directly on the raw photon sequence recorded at the receiver, and is realised as a 1-D convolutional network as well as a deeper residual 1-D CNN. The second operates on a compact 13-dimensional feature vector derived from photon statistics, secure-key length, and protocol scalars, and is realised as both a baseline feed-forward network and a deeper residual ANN with skip connections. Together, these four models let us study attack detection across two distinct input modalities. We further establish where each model is best suited by analysing accuracy alongside inference latency, throughput, and data-cost trade-offs, providing guidance for both offline auditing and real-time edge monitoring. Further, to train the model, we introduce a physics-based simulated dataset featuring both attack and non-attack scenarios. Proposed models achieve up-to 83.4% classification accuracy with minimal false positives and negatives, demonstrating robust performance in identifying compromised quantum channels. This work represents a significant advancement in machine learning-assisted quantum security, providing a practical tool for enhancing the security of real-world QKD systems against stealthy photon-level attacks.

•Jul 15, 2026•0 citations

Adaptive BB84 Quantum Key Distribution Framework: A Comprehensive Implementation and Analysis Platform

Momin Shiraz Shafi, Vivek Parashar Pandey, Momin Rayyan +3 more

With the rapid development of quantum computing, traditional encryption techniques are expected to face serious security challenges in the near future. By depending on the principles of quantum physics rather than computing complexity, Quantum Key Distribution (QKD), and specifically the BB84 protocol, provides a radically new method of secure communication. In this project, an adaptive BB84-based framework is implemented and analyzed under realistic conditions, including channel noise and multiple eavesdropping strategies. The system integrates error correction, privacy amplification, finite-key analysis, and decoy-state mechanisms to reflect practical deployment scenarios. Three major attack models—intercept-resend, photon number splitting, and Trojan horse—are simulated to evaluate protocol robustness. The experimental results show that during normal communication the average Quantum Bit Error Rate (QBER) remains close to 3–4%, whereas the presence of active eavesdropping increases the QBER beyond the commonly accepted 11% threshold, resulting in protocol termination. Additional statistical testing under varying noise conditions confirms stable secret key generation and reliable detection of malicious interference. Overall, the proposed framework serves as a comprehensive simulation platform for studying the security behavior of BB84 in controlled experimental environments.

IEEE International Conference on Advances in Computing Research on Science Engineering and Technology (ACROSET)•Sep 1, 2026•0 citations

Adversarial Reinforcement Learning for Adaptive Eavesdropping in BB84 Quantum Key Distribution

Ramon Jose C. Bagunu

BB84 quantum key distribution derives its security from a physical guarantee that any eavesdropper disturbs the channel in a statistically detectable way. Prior work evaluates this by assuming Eve attacks at a fixed, analytically optimized rate. We examine what happens when Eve is modeled instead as a learning agent. Classical reinforcement learning is used, specifically tabular Q-Learning, SARSA, and Double Q-Learning, to adaptive BB84 eavesdropping. This formulates the attacker's decision as a Markov Decision Process where the agent observes Quantum Bit Error Rate (QBER) feedback and decides, qubit by qubit, whether to intercept or pass. Experiments span three channel noise levels ($μ_{ch}\in\{1\%,3\%,5\%\}$) and are validated across five independent random seeds (45 training runs per condition, 10,000 episodes each). Against the best non-adaptive analytical baseline, Q-Learning reduces detection from $99.4\%$ to $0.28\%\pm0.27\%$ at $μ_{ch}=1\%$ while extracting approximately 10.5 correct bits per episode. This is a 355-fold reduction that is statistically significant ($p=0.020$, Mann-Whitney $U$ test). We also report the spontaneous emergence of an end-game burst, where agents independently learn to surge their attack rate at the final block. This exploit vanishes under randomized checkpoint intervals while stealth performance remains statistically indistinguishable. These results motivate the inclusion of adaptive adversary baselines in quantum cryptographic security evaluations.

quant-pharXiv:2606.22962•Quantum Information Processing•Jun 23, 2026•4 citations

Intelligent Security Monitoring and Secure Key Generation for Long-Distance QKD Using Hybrid Machine Learning and Chaotic Perturbation

Aditya Narayan, Ronaldo Mahabodhi, Yatendra Sahu +2 more

Abstract Quantum computing threatens classical security protocols such as RSA and ECC, motivating Quantum Key Distribution (QKD) schemes such as BB84 and E91. Over 100\,km optical links, however, channel-noise and selective-measurement attacks can mimic normal degradation, making single-metric monitoring (e.g., QBER alone) unreliable. We propose a hybrid framework that combines an SVM-RBF classifier -- driven by a trilateral feature vector of Quantum Bit Error Rate (QBER), Bell-CHSH $S$-parameter, and Temporal Shannon Entropy -- with a quantum-perturbed Logistic Map for post-verification session-key expansion. On a 50,000-sample digital-twin dataset (12,500 SAFE / 37,500 BREACHED) under extreme 100\,km channel perturbation, the classifier attains 87.82\% accuracy and an F1-score of 0.9187, outperforming nine baselines including a Compact CNN, under both batched and realistic single-query latency. All three features are individually significant (ANOVA, $p<0.001$). The expanded session key was evaluated with the NIST SP~800-22 suite at a $10^5$-bit stream length (nine tests, including Binary Matrix Rank): the raw Logistic-Map output passes only 3 of 9 tests, confirming that chaotic expansion alone preserves the residual bias of the quantum seed, whereas applying SHA-256-based cryptographic whitening (per NIST SP~800-90B) to the same material raises the pass rate to 9 of 9, showing that the whitened key meets NIST-grade randomness and is directly deployable for symmetric encryption.

•Sep 8, 2026•0 citations
Conventional BB84 Quantum Key Distribution (QKD) systems rely on a fixed 11% Quantum Bit Error Rate (QBER) threshold to detect eavesdropping. However, stealthy attacks can remain below this threshold while still compromising channel security. This paper proposes a temporal QBER based machine learning framework for detecting and classifying eavesdropping attacks in BB84 QKD systems. Rather than relying on average session level QBER, the framework extracts 63 physics-informed temporal features capturing burst behavior, temporal instability, basis dependent asymmetry, and QBER loss interactions. Random Forest, XGBoost, and Support Vector Machine with a Radial Basis Function kernel (SVM-RBF) classifiers are evaluated on seven eavesdropping attacks and a normal channel scenario under noisy and lossy conditions. Averaged over ten independent runs, XGBoost achieves the best performance with 88.01% (0.47%) accuracy and a macro F1 score of 0.8803, while SVM-RBF performs comparably, confirming the robustness of the proposed features. Evaluated as a binary attack-versus-normal detector for comparison with conventional monitoring, a fixed 11% QBER threshold achieves only 25.82% accuracy with a False Negative Rate (FNR) of 0.8477, whereas the proposed framework reduces the FNR to 0.0198, substantially improving detection of stealthy attacks that evade threshold-based monitoring. SHapley Additive exPlanations based (SHAP) explainability shows that physics-informed temporal and channel derived features are highly discriminative for identifying eavesdropping strategies. These results demonstrate that temporal QBER driven machine learning provides an accurate, explainable, and practical framework for multi attack security monitoring in BB84 QKD systems.
Quantum Key Distribution (QKD) using the BB84 protocol offers the potential for information-theoretic security based upon the principles of quantum mechanics; however, eavesdropping mechanisms such as intercept-resend attacks and ambient noise sources that degrade quantum channel fidelity may render such systems exploitable. We present a oneclass machine learning approach to detect two categories of threats present within BB84 quantum channels: (i) interceptresend eavesdropping committed by an adversarial actor (Eve), and (ii) six types of channel noise. We have trained three unsupervised anomaly detectors (One-Class Support Vector Machine (OC-SVM), Isolation Forest (iForest), and Extended Isolation Forest (Ex-iForest)) on a dataset consisting of 1000 instances (940 clean and 60 anomalous) and tested each of them on separate test sets of 100 instances. With respect to the detection of adversarial attacks, Ex-iForest outperformed both OC-SVM and iForest; it achieved 95.00% accuracy, precision, recall, and F1-score, and when detecting channel noise, it achieved an overall accuracy, precision, recall, and F1-score of 92.00%. Feature importance analysis confirmed that the Quantum Bit Error Rate (QBER) and the fidelity of polarization measurement were the two most discriminative features for adversarial detection while fluctuations of photon detection counts and timing jitter were the two most informative features for channel noise discrimination. These results suggest that ExiForest is a viable, robust, and efficient solution for real-time anomaly detection systems in BB84-based quantum communication systems.
Quantum Key Distribution (QKD) techniques provide theoretically invulnerable security for cryptographic key exchange. However, hardware limitations, such as the use of weak coherent photon sources instead of a single-photon source, make it vulnerable to photon-number splitting (PNS) attacks. In a PNS attack Eve selectively intercepts multi-photon pulses while allowing single-photon pulses to pass through. PNS attacks do not affect the quantum bit error rate (QBER), rendering traditional QBER-based monitoring ineffective. Hence, to detect the attack in such cases, it becomes imperative to statistically monitor parameters such as photon counts at the detector, SKR, and shifted key length to identify the attack. In this paper, we propose two complementary deep learning-based frameworks for detection of the PNS attack in the QKD setup. The first operates directly on the raw photon sequence recorded at the receiver, and is realised as a 1-D convolutional network as well as a deeper residual 1-D CNN. The second operates on a compact 13-dimensional feature vector derived from photon statistics, secure-key length, and protocol scalars, and is realised as both a baseline feed-forward network and a deeper residual ANN with skip connections. Together, these four models let us study attack detection across two distinct input modalities. We further establish where each model is best suited by analysing accuracy alongside inference latency, throughput, and data-cost trade-offs, providing guidance for both offline auditing and real-time edge monitoring. Further, to train the model, we introduce a physics-based simulated dataset featuring both attack and non-attack scenarios. Proposed models achieve up-to 83.4% classification accuracy with minimal false positives and negatives, demonstrating robust performance in identifying compromised quantum channels. This work represents a significant advancement in machine learning-assisted quantum security, providing a practical tool for enhancing the security of real-world QKD systems against stealthy photon-level attacks.
Quantum Key Distribution (QKD) techniques provide theoretically invulnerable security for cryptographic key exchange. However, hardware limitations, such as the use of weak coherent photon sources instead of a single-photon source, make it vulnerable to photon-number splitting (PNS) attacks. In a PNS attack Eve selectively intercepts multi-photon pulses while allowing single-photon pulses to pass through. PNS attacks do not affect the quantum bit error rate (QBER), rendering traditional QBER-based monitoring ineffective. Hence, to detect the attack in such cases, it becomes imperative to statistically monitor parameters such as photon counts at the detector, SKR, and shifted key length to identify the attack. In this paper, we propose two complementary deep learning-based frameworks for detection of the PNS attack in the QKD setup. The first operates directly on the raw photon sequence recorded at the receiver, and is realised as a 1-D convolutional network as well as a deeper residual 1-D CNN. The second operates on a compact 13-dimensional feature vector derived from photon statistics, secure-key length, and protocol scalars, and is realised as both a baseline feed-forward network and a deeper residual ANN with skip connections. Together, these four models let us study attack detection across two distinct input modalities. We further establish where each model is best suited by analysing accuracy alongside inference latency, throughput, and data-cost trade-offs, providing guidance for both offline auditing and real-time edge monitoring. Further, to train the model, we introduce a physics-based simulated dataset featuring both attack and non-attack scenarios. Proposed models achieve up-to 83.4% classification accuracy with minimal false positives and negatives, demonstrating robust performance in identifying compromised quantum channels. This work represents a significant advancement in machine learning-assisted quantum security, providing a practical tool for enhancing the security of real-world QKD systems against stealthy photon-level attacks.
With the rapid development of quantum computing, traditional encryption techniques are expected to face serious security challenges in the near future. By depending on the principles of quantum physics rather than computing complexity, Quantum Key Distribution (QKD), and specifically the BB84 protocol, provides a radically new method of secure communication. In this project, an adaptive BB84-based framework is implemented and analyzed under realistic conditions, including channel noise and multiple eavesdropping strategies. The system integrates error correction, privacy amplification, finite-key analysis, and decoy-state mechanisms to reflect practical deployment scenarios. Three major attack models—intercept-resend, photon number splitting, and Trojan horse—are simulated to evaluate protocol robustness. The experimental results show that during normal communication the average Quantum Bit Error Rate (QBER) remains close to 3–4%, whereas the presence of active eavesdropping increases the QBER beyond the commonly accepted 11% threshold, resulting in protocol termination. Additional statistical testing under varying noise conditions confirms stable secret key generation and reliable detection of malicious interference. Overall, the proposed framework serves as a comprehensive simulation platform for studying the security behavior of BB84 in controlled experimental environments.
BB84 quantum key distribution derives its security from a physical guarantee that any eavesdropper disturbs the channel in a statistically detectable way. Prior work evaluates this by assuming Eve attacks at a fixed, analytically optimized rate. We examine what happens when Eve is modeled instead as a learning agent. Classical reinforcement learning is used, specifically tabular Q-Learning, SARSA, and Double Q-Learning, to adaptive BB84 eavesdropping. This formulates the attacker's decision as a Markov Decision Process where the agent observes Quantum Bit Error Rate (QBER) feedback and decides, qubit by qubit, whether to intercept or pass. Experiments span three channel noise levels (μch∈{1%,3%,5%}μ_{ch}\in\{1\%,3\%,5\%\}μch​∈{1%,3%,5%}) and are validated across five independent random seeds (45 training runs per condition, 10,000 episodes each). Against the best non-adaptive analytical baseline, Q-Learning reduces detection from 99.4%99.4\%99.4% to 0.28%±0.27%0.28\%\pm0.27\%0.28%±0.27% at μch=1%μ_{ch}=1\%μch​=1% while extracting approximately 10.5 correct bits per episode. This is a 355-fold reduction that is statistically significant (p=0.020p=0.020p=0.020, Mann-Whitney UUU test). We also report the spontaneous emergence of an end-game burst, where agents independently learn to surge their attack rate at the final block. This exploit vanishes under randomized checkpoint intervals while stealth performance remains statistically indistinguishable. These results motivate the inclusion of adaptive adversary baselines in quantum cryptographic security evaluations.
Abstract Quantum computing threatens classical security protocols such as RSA and ECC, motivating Quantum Key Distribution (QKD) schemes such as BB84 and E91. Over 100\,km optical links, however, channel-noise and selective-measurement attacks can mimic normal degradation, making single-metric monitoring (e.g., QBER alone) unreliable. We propose a hybrid framework that combines an SVM-RBF classifier -- driven by a trilateral feature vector of Quantum Bit Error Rate (QBER), Bell-CHSH SSS-parameter, and Temporal Shannon Entropy -- with a quantum-perturbed Logistic Map for post-verification session-key expansion. On a 50,000-sample digital-twin dataset (12,500 SAFE / 37,500 BREACHED) under extreme 100\,km channel perturbation, the classifier attains 87.82\% accuracy and an F1-score of 0.9187, outperforming nine baselines including a Compact CNN, under both batched and realistic single-query latency. All three features are individually significant (ANOVA, p<0.001p<0.001p<0.001). The expanded session key was evaluated with the NIST SP~800-22 suite at a 10510^5105-bit stream length (nine tests, including Binary Matrix Rank): the raw Logistic-Map output passes only 3 of 9 tests, confirming that chaotic expansion alone preserves the residual bias of the quantum seed, whereas applying SHA-256-based cryptographic whitening (per NIST SP~800-90B) to the same material raises the pass rate to 9 of 9, showing that the whitened key meets NIST-grade randomness and is directly deployable for symmetric encryption.