Sifting scheme for continuous-variable quantum key distribution with short samples

Abstract

Preventing the secret key from being stolen is an important issue in practical quantum key distribution systems. In the sifting step, the legitimate parties discard the useless portion of the raw data to form the sifted key. This step is executed at high speed to support the high repetition frequency of the systems without guaranteeing the security of the raw data. In practical systems, useless data contain abnormal data and the key measured by the legitimate party on different bases. Here we propose a sifting scheme based on machine learning that can monitor anomaly quantum signal disturbances in practical continuous-variable quantum key distribution systems. It randomly samples small amounts of data from the data block and uses short samples to preliminarily sift the abnormal one. The results show that the model can quickly distinguish normal communication from most common attacks with the cost of a small part of the raw keys and improve system performance under attacks. In principle, the model can also be generalized and applied to discrete-variable quantum key distribution systems and further enhance the security of quantum key distribution.