The security of device-independent quantum key distribution (QKD) has been deemed ineffective by a team of Canadian researchers, and at least one commercial product already in use for telecommunications is directly affected.
Last month Feihu Xu, Bing Qi and Hoi-Kwong Lo at the University of Toronto reported a successful hack of devices made by ID Quantique, a Geneva-based commercial quantum cryptography company. The products, the company says, are secure, meeting currently International Security standards (Commmon Criteria EAL4+ and FIPS PUB 140-2 Level 3). Key distribution certification, however, does not yet exist.
In the classic Bob and Alice explanation of QKD works something like this: If Alice wants to communicate with Bob, she sends him the key using quantum physics. The idea is to declare photons that are angled in a particular way, for example up is 1 and down is 0. It is also possible that 45 degrees right is 1 and 45 degrees left is 0. Once Bob receives and confirms the key on his end, Alice can then start the transmission.
This form of cryptography maintains that if Eve were to eavesdrop on the quantum communication, she'll have no choice but to introduce errors in the original message when she goes to resend it, tipping off Bob and Alice that someone else is on the line. Thus, it's hard to mount a Man-in-the-Middle attack on a quantum stream.
That is in theory.
In reality, say the researchers, there is a 20 percent error rate within the communication because of naturally occurring background noise. Knowing this, the researchers claim to have found a way for Eve to intercept some of the communications stream, glean its message, and resend the packets without the error rate ever topping that 20 percent threshold. The flaw, it appears, is the base assumption that that 20% is tolerable.
But how likely are these attacks?
ID Quantique says in its FAQ " Some of the attacks that have been proposed by research groups are quite powerful, but it is important to stress that they remain academic and do not apply in practice. They usually require an adjustment phase that would introduce noticeable perturbations in the link. Moreover, they require an in-depth characterization of the actual QKD system to be attacked. This is possible in the lab, but not in the field."
The commercial application of this is that a device-dependent cryptography system, where two devices perform the role of Bob and Alice, could be infected by a Trojan that performs as Eve, intercepting and storing the communications before sending it on. This is known as an "intercept and resend attack." While not fatal, it does poke a hole into the invincibility of quantum cryptography. The full academic paper can be found here. (PDF)