Relativistic Bennett-Brassard cryptographic scheme, relativistic errors, and how to correct them

The Bennett-Brassard cryptographic scheme needs two bases, at least one of them linearly polarized. The problem is that linear polarization, formulated in terms of helicities, is not a relativistically covariant notion: State which is linearly polarized in one reference frame becomes depolarized in another one. We show that a relativistically moving receiver of information should define linear polarization with respect to projection of Pauli-Lubanski's vector in a principal null direction of the Lorentz transformation which defines the motion, and not with respect to the helicity basis. Such qubits do not depolarize.