A group of English and Canadian researchers has cast doubt on the nascent push to develop device-independent quantum cryptography standards, asserting that such schemes could be undermined by malicious vendors.
Their paper, Prisoners of their own device: Trojan attacks on device-independent quantum cryptography, is published on Arxiv.org, here.
The paper outlines scenarios which the authors say would be undetectable to the user, but would allow the attacker to obtain sufficient information to snoop on supposedly “uncrackable” quantum cryptography.
The paper, authored by London University mathematician Jonathan Barrett, Roger Colbeck of Canada’s Perimeter Institute of Theoretical Physics, and Adrian Kent of Cambridge’s Centre for Quantum Information and Foundations, states:
“A malicious manufacturer who wishes to mislead users or obtain data from them can equip devices with a memory [El Reg – to clarify, in our reading this refers to a memory included in the devices specifically for attack purposes] and use it in programming them.
“A task is potentially vulnerable to our attacks if it involves secret data generated by devices, and if Eve [El Reg – ie, the attacker] can learn some function of the device outputs.”
Their analysis gives rise, for example, to a scenario in which the attacking equipment might store key exchange communications from “day 1”, use this to analyse the key exchange taking place on “day 2”; and use this to extract the “day 1” key.
This is supposed to be impossible, since any tampering with the quantum communication channel should be revealed – for example, as (entanglement-destroying) noise on the quantum channel.
However, as the authors point out, all real-world channels contain noise; to overcome this, quantum crypto schemes exchange multiple pairs over a noisy channel, and use a statistical analysis to detect interference in the channel.
The malicious manufacturer, however, should be able to conceal its activities below the noise threshold the system uses to decide that the channel remains secure. The attacker could even build systems whose actual noise levels are lower than claimed, and use the gap between specified and real noise to conceal their activity.
If not addressed, the authors say the flaws they have identified effectively turn QKD devices into a “use once” proposition: you can only guarantee security for the first exchange, so the device has to be disposed of. ®
Comment: Before the world proclaims “quantum crypto not secure!” in headlines (too late? Oh well…) El Reg would make a couple of observations.
First, the malicious manufacturer is not a quantum-specific threat: backdoors can be just as easily inserted into classical cryptography kit.
Second, this paper is presenting a discussion not on any mass-deployed system, but on proposed schemes for device-independent QKD. Device independence has come to the fore chiefly because of prior demonstrations suggesting that today’s implementations have exploitable flaws; as a result, there has been ongoing discussion as to how users might verify the security of a quantum communication without knowing anything about the equipment used to create that channel.
For those interested in the kinds of schemes they believe could be compromised, the article cites some key papers on Arxiv, such as:
Third, the authors do not claim to have actually built a working proof-of-concept: their paper is a discussion of how a malicious system may be designed; it’s been published on Arxiv for review, and El Reg would expect a veritable feast of future papers for quantum crypto enthusiasts. ®