Machine Learning-Based Polarization Drift Compensation for High Speed DV-QKD Homodyne Receiver

Discrete variables quantum key distribution (DV-QKD), with its well-studied and scrutinized BB84 protocol, benefits from being very attractive for highly secure communications. However, current detection schemes rely on the use of InGaAs SPDAs, which limits not only its use in high temperature environments, but also high secure communication rates. A possible approach is the use of coherent homodyne detection schemes for polarization encoding based DV-QKD combined. To deploy polarization encoding DV-QKD over standard optical fiber high speed networks, the polarization drift suffered due to birefringence over the channel must be compensated. In this work, we use a machine learning (ML) polarization tracking and compensation algorithm combined with a coherent homodyne receiver, thus allowing the deployment of high-speed polarization encoding based DV-QKD in standard optical fibers. The ML-algorithm predicts the SOP evolution keeping the error rate below 1 %. In this way, the overhead to polarization monitoring is reduced leading to a secure key exchange rate (SKR) of 79 Mbps for a communication over 40 km optical fiber.