Reinforcement Learning-based Adaptive Error Correction for BB84 Quantum Key Distribution: A Q-Learning Approach to Post-Processing Optimization
Devishree Rohit Naidu, Sweta Jain, Ananya Anand +2 more
In Quantum Key Distribution (QKD) protocol BB84, a crucial step is post-processing. It is where the communicating parties correct errors caused either by channel noise or by an eavesdropper. Standard BB84 implementations usually relies on fixed choices of post-processing parameters. This often leads to extra classical communication and lowers the effective key generation rate. In this work, we propose an adaptive framework based on reinforcement learning that changes the error-correction block size dynamically during the classical reconciliation stage of BB84. We have tested this idea using Qiskit Aer simulations over 200 runs with key lengths ranging from 100 to 500 qubits. Here we observed approximately 30-50% reduction in communication overhead compared to the conventional fixed-block settings. Even though this introduces slightly higher error rates, it still improves the overall efficiency of key generation across all evaluated key sizes. This indicates that reinforcement learning can make QKD post-processing more efficient. This can open opportunities for quantum-secure communication systems.
International Conference on Connected Intelligence for Industrial Applications (CI2A)•Apr 3, 2026•0 citations