Oscar Dahlsten

In a quantum key distribution protocol, an eavesdropper Eve extracts information by performing the measurement on flying qubits. If the measurement basis does not match with the basis in which the qubit is prepared, it leads to a disturbance in the qubit state with 1/2 probability. There is an optimal condition between Information Gain and tolerable disturbance. Leveraging machine learning techniques, several advances have been made in the rapid development of quantum applications. In order to achieve the (upper) bound for Information Gain vs. disturbance trade-off, we investigate this optimal condition for the BB84 protocol using the gradient descent algorithm. By employing such a numerical method, we report an improvisation in the upper bound of the disturbance compare to the existing best threshold reported in Phys. Rev. A, 56 1163 (1997). Interestingly, the Information Gain by Eve at this bound is relatively less than reported in the above paper. It means with our strategy Alice-Bob are in advantageous position in both ways as they do not have to abort the protocol as early as required in Phys. Rev. A, 56 1163 (1997) at d=0.146447, as well as now the information leaked to Eve is also lowered.