In the paper [Phys. Rev. A 69, 052319 (2004)], a quantum direct communication protocol is proposed which is claimed to be unconditionally secure even for the case of a noisy channel. We show that this is not the case by giving an undetectable attack scheme.