DRL-assisted quantum attack mitigation in resource allocation of CV-QKD over optical networks

Abstract

Due to possible imperfect implementation of laser sources and coherent detectors, continuous-variable quantum key distribution (CV-QKD) in optical networks is vulnerable to quantum attacks, which would significantly degrade the secret key rate (SKR) and compromise the security of the quantum keys. The attackers may steal the data encrypted with the compromised keys, leading to service failure and resource wastage. To tackle it, we propose a quantum attack mitigation resource allocation scheme (QAM-RA) for CV-QKD over optical networks. It corrects the biased channel parameters of the compromised quantum channels by conducting appropriate countermeasures, re-analyzes the SKR, and allocates additional quantum resources to compensate for the loss of SKR. A deep reinforcement learning (DRL) framework based on the Asynchronous Advantage Actor-Critic (A3C) algorithm is developed to determine QAM-RA solutions intelligently. Extensive simulations have been conducted to evaluate the performance of the DRL-assisted QAM-RA scheme in two test networks. Simulation results have confirmed the effectiveness of the proposed scheme in mitigating quantum attacks and reducing user request blocking probabilities.