Notes on recent approaches concerning the Kirchhoff-law-Johnson-noise-based secure key exchange

摘要

We critically analyze the results and claims in [P.-L. Liu, Phys. Lett. A 373 (2009) 901]. We show that the strong security leak appeared in the simulations is only an artifact and not caused by "multiple reflections". Since no wave modes exist at cable length of 5% of the shortest wavelength of the signal, no wave is present to reflect it. In the high wave impedance limit, the conditions used in the simulations are heavily unphysical (requiring cable diameters up to 28000 times greater than the measured size of the known universe) and the results are modeling artifacts due to the unphysical values. At the low cable impedance limit. the observed artifacts are due to violating the recommended (and tested) conditions by neglecting the cable capacitance restrictions and using about 100 times longer cable than recommended without cable capacitance compensation arrangement. We implement and analyze the general circuitry of Liu's circulator [P.-L. Liu. Phys. Lett. A 373 (2009) 901] and confirm that they are conceptually secure against passive attacks. We introduce an asymmetric, more robust version without feedback loop. Then we crack all these systems by an active attack: a circulator-based man-in-the middle attack. Finally, we analyze the proposed method to increase security by dropping only high-risk bits. We point out the differences between different types of high-risk bits and show the shortage of this strategy for some simple key exchange protocols.