Noise-suppressing channel allocation in dynamic DWDM-QKD networks using LightGBM

Area of Science:

• Quantum Information Science
• Optical Network Engineering
• Machine Learning Applications

Background:

• Integrating quantum key distribution (QKD) into existing optical networks is crucial for cost reduction and resource efficiency.
• Real-world scenarios present challenges due to dynamic classical data traffic, causing time-varying noises that degrade quantum channel quality.
• Conventional static channel allocation schemes are inadequate for maintaining quantum channel integrity under fluctuating noise conditions.

Purpose of the Study:

• To propose and evaluate a novel machine learning-based noise-suppressing channel allocation (ML-NSCA) scheme.
• To address the limitations of static allocation in dynamic optical network environments.
• To enhance the secure key rate and operational efficiency of QKD systems coexisting with classical data traffic.

Main Methods:

• Developed a LightGBM-based machine learning framework to predict optimal channel allocations minimizing noise impact.
• Implemented periodic reallocation of quantum channels based on ML predictions to ensure high secure key rates.
• Optimized feature extraction methods within the ML framework to improve accuracy and scalability.

Main Results:

• The ML-NSCA scheme effectively mitigates dynamic noise impacts in realistic optical networks.
• Demonstrated a significant increase in secure key rates compared to previous schemes.
• Achieved lower operational complexity in managing quantum channels.

Conclusions:

• The proposed ML-NSCA scheme offers a robust solution for integrating QKD into dynamic optical networks.
• Machine learning provides an effective tool for real-time noise management and channel allocation in hybrid networks.
• The approach enhances the practical feasibility and performance of secure quantum communication over existing infrastructure.