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

Science China Information Sciences

Dynamically quantum attack detection for quantum key distribution based on deep representations

Minjie Liu, Ye Chen, Xiaodong Fan, Jinquan Huang, Tonglin Mu, Junran Guo, Shi‐Hai Sun

Published Aug 13, 2026• 0 citations
Publisher
DOI
10.1007/s11432-025-4878-3

Related papers

Detecting quantum hacking attacks for continuous-variable quantum key distribution using quantum neural network

Junhao Li, Yiyu Mao, Qin Liao +3 more

Chaos Solitons & Fractals•Oct 29, 2025•4 citations

UDMTL-QKD: An Ultra-Deep Multi-Task Learning Approach for Quantum Key Distribution Secured Multi-Bank Transactions in 6G Networks

Bibek Kumar, Jashvant Kumar, Aparna Tripathi +2 more

Abstract As quantum computing starts taking over the world, traditional cryptography schemes become less and less secure, in need of quantum-resistant cryptography solutions. In theory, quantum Key Distribution provides a theoretically unconditional level of security; nevertheless, in practice, noise sensitivity, dynamic network behavior, limited transmission range and complexity of the system pose challenges to the practical implementation of quantum key distribution. To overcome these shortcomings, the suggested framework combines deep learning and quantum communication, developing a multi-task problem which simultaneously detects the security of the quantum communication, predicts it, and also classifies attack types in 6G networks.It optimizes the heterogeneous, non-linear and stochastic data to include feature engineering (load ratio, total cryptographic time), normalization, encoding and imputation, to increase model robustness and learning efficiency.. Baseline XGBoost models (27.5%–37% accuracy) and hybrid ensembles (~ 38%) exhibit limited efficacy, motivating the proposed UDMTL architecture leveraging deep fully connected layers (512–256–128–64) with batch normalization and dropout for hierarchical feature abstraction and enhanced generalization. The proposed model achieves ~ 87% overall accuracy (89% QKD detection, 85% attack classification, R² = 0.91), where shared representation learning captures cross-task dependencies, significantly outperforming conventional approaches.

•May 25, 2026•0 citations

A survey of known attacks against Quantum Key Distribution

Oliver Kaisti

•Apr 18, 2017•0 citations

Design of an integrated model using deep reinforcement learning and Variational Autoencoders for enhanced quantum security

Harshala Shingne, Diptee Vishwanath Chikmurge, Priya Parkhi +1 more

The need for secure communication systems has driven extensive research into quantum-based security mechanisms, particularly Quantum Key Distribution (QKD). However, traditional QKD systems, within dynamic environments incorporating network fluctuation and attacks, have been relatively limited because static protocols cannot support high key generation rates and security. This work addresses these challenges by proposing the integration of AI and machine learning optimization techniques into quantum communication protocols to enhance both security and efficiency. We here propose three advanced models: first, Deep Reinforcement Learning is applied to adaptively optimize QKD protocols by dynamically adjusting the key generation parameters with respect to environmental conditions. In the state-of-the-art method, the DRL-based approach enlarges the secure key generation rate by 15-20 % and suppresses QBER 30-40 % under noisy conditions. A VAE is used for the detection of anomalies in quantum networks that effectively detects eavesdropping. By incorporating quantum-specific feature extraction and latent variable disentanglement, the VAE model detects attack detection accuracy of 85-90 % with a reduction of 25 % in false positives. Finally, it considers the optimization of cryptographic protocols in a distributed quantum network using Multi-Agent Deep Q-Networks. This multi-agent system strengthens both the security and computational efficiency by reducing attack vulnerabilities by 15-18 % and lowering the computational complexity by 20-25 %. In all, the integration of AI with machine learning methods brings far better enhancements in the field of quantum communication system security and efficiency, addressing critical limitations of conventional QKD systems and pointing to the way to more resilient adaptive quantum security solutions.

MethodsX•Jun 21, 2025•11 citations

An Effective Attack on the Quantum Key Distribution Protocol Based on Quantum Encryption

Fei Gao, Su‐Juan Qin, Qiaoyan Wen +1 more

Lecture notes in computer science•Jan 1, 2005•6 citations

Beyond universal attack detection for continuous-variable quantum key distribution via deep learning

Haisen Luo, Ling Zhang, Hao Qin +6 more

Removing quantum hacking attacks is a highly challenging task for quantum key distribution (QKD) since one needs to correctly detect the types of attacks. There is a big gap between the idealized theoretical model and the practical physical system, which results in various hacking strategies exploited from security loopholes. To bridge this gap, previous continuous-variable (CV) QKD systems adopted a universal statistical model to represent different dimensional characters of the physical layer. The challenge is to discriminate different attacks from the exposed key physical details. Here we present a semisupervised deep-learning method to detect the known attacks or potential threats that are hard to be captured in previous CVQKD systems. The proof-of-principle experiment result shows that it could break the present limitations that different attacks are relatively independent, and key features computed from previous threshold-based methods are not correlated with each other. We anticipate that our proposal will give insight into how quantum hacking attacks can be detected under the real system modeling constraints, and will enable the present quantum communication systems to reach a higher level of practical security without adding new assumptions.

Physical Review A•Apr 7, 2022•35 citations
Abstract As quantum computing starts taking over the world, traditional cryptography schemes become less and less secure, in need of quantum-resistant cryptography solutions. In theory, quantum Key Distribution provides a theoretically unconditional level of security; nevertheless, in practice, noise sensitivity, dynamic network behavior, limited transmission range and complexity of the system pose challenges to the practical implementation of quantum key distribution. To overcome these shortcomings, the suggested framework combines deep learning and quantum communication, developing a multi-task problem which simultaneously detects the security of the quantum communication, predicts it, and also classifies attack types in 6G networks.It optimizes the heterogeneous, non-linear and stochastic data to include feature engineering (load ratio, total cryptographic time), normalization, encoding and imputation, to increase model robustness and learning efficiency.. Baseline XGBoost models (27.5%–37% accuracy) and hybrid ensembles (~ 38%) exhibit limited efficacy, motivating the proposed UDMTL architecture leveraging deep fully connected layers (512–256–128–64) with batch normalization and dropout for hierarchical feature abstraction and enhanced generalization. The proposed model achieves ~ 87% overall accuracy (89% QKD detection, 85% attack classification, R² = 0.91), where shared representation learning captures cross-task dependencies, significantly outperforming conventional approaches.
The need for secure communication systems has driven extensive research into quantum-based security mechanisms, particularly Quantum Key Distribution (QKD). However, traditional QKD systems, within dynamic environments incorporating network fluctuation and attacks, have been relatively limited because static protocols cannot support high key generation rates and security. This work addresses these challenges by proposing the integration of AI and machine learning optimization techniques into quantum communication protocols to enhance both security and efficiency. We here propose three advanced models: first, Deep Reinforcement Learning is applied to adaptively optimize QKD protocols by dynamically adjusting the key generation parameters with respect to environmental conditions. In the state-of-the-art method, the DRL-based approach enlarges the secure key generation rate by 15-20 % and suppresses QBER 30-40 % under noisy conditions. A VAE is used for the detection of anomalies in quantum networks that effectively detects eavesdropping. By incorporating quantum-specific feature extraction and latent variable disentanglement, the VAE model detects attack detection accuracy of 85-90 % with a reduction of 25 % in false positives. Finally, it considers the optimization of cryptographic protocols in a distributed quantum network using Multi-Agent Deep Q-Networks. This multi-agent system strengthens both the security and computational efficiency by reducing attack vulnerabilities by 15-18 % and lowering the computational complexity by 20-25 %. In all, the integration of AI with machine learning methods brings far better enhancements in the field of quantum communication system security and efficiency, addressing critical limitations of conventional QKD systems and pointing to the way to more resilient adaptive quantum security solutions.
Removing quantum hacking attacks is a highly challenging task for quantum key distribution (QKD) since one needs to correctly detect the types of attacks. There is a big gap between the idealized theoretical model and the practical physical system, which results in various hacking strategies exploited from security loopholes. To bridge this gap, previous continuous-variable (CV) QKD systems adopted a universal statistical model to represent different dimensional characters of the physical layer. The challenge is to discriminate different attacks from the exposed key physical details. Here we present a semisupervised deep-learning method to detect the known attacks or potential threats that are hard to be captured in previous CVQKD systems. The proof-of-principle experiment result shows that it could break the present limitations that different attacks are relatively independent, and key features computed from previous threshold-based methods are not correlated with each other. We anticipate that our proposal will give insight into how quantum hacking attacks can be detected under the real system modeling constraints, and will enable the present quantum communication systems to reach a higher level of practical security without adding new assumptions.