.png){kind=icon}
.png){kind=icon}
Return
Dynamically quantum attack detection for quantum key distribution based on deep representations
.png){kind=icon}
PRE
AI
DOI：10.1007/s11432-025-4878-3.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Quantum key distribution (QKD) has attracted more attention and has been developed rapidly due to its unconditional security. However, various quantum attacks, exploiting imperfections in practical devices, might be performed by Eve to steal the key. Thus, detecting the existence of Eve and the type of quantum attacks in real time (or online) becomes an important task for practical QKD systems. In this paper, we propose a dynamic evolutionary attack detection scheme for practical QKD, in which, with the online sampled system parameters, Eve’s attack feature could be vectorized by a well-designed embedding model and dynamically self-update to an attack feature vector database. Then, our method not only has high detection accuracy for known quantum attacks but also has the ability to detect and classify unseen quantum attack strategies without retraining the embedding model. Finally, taking the continuous variable (CV) QKD as an example, we demonstrate that the detection accuracy of 100% is achieved for four types of known attacks and over 99% for two types of unseen quantum attack strategies. Thus, our work significantly improves the generalization capability of quantum attack detection and promotes the practical development of QKD.
Keywords:
QKD
quantum attacks
dynamic evolutionary detection scheme
embedding model
attack feature vector database
unseen quantum attack strategies
Journal
S
IF:
7.6
Papers:
87
Citations:
0