Process Tomography can identify the quantum channel, while Blind Quantum Computation strives to hide the inputs using quantum gates.

Given the user executes the same blind circuit multiple times (of the order of times required to do process tomography), why can't the server do process tomography to break the blindness?

  • $\begingroup$ The whole point of a blind computation is that the client chooses different random phase encodings each time. Sure, if you use the same encodings every time, you're asking for it! $\endgroup$
    – DaftWullie
    Commented May 7, 2021 at 19:51
  • $\begingroup$ with Universal Blind QC, there is no requirement for the client to have any QC capability. In that case, 'maybe' the pseudo-random generator for the phase encoding can also be inverted. $\endgroup$
    – Aritra
    Commented May 8, 2021 at 13:30
  • 1
    $\begingroup$ First, there's always a quantum part for the client in Universal Blind QC; either the client can prepare & send single qubits, or they can receive & measure. Secondly, in proofs of security of blind QC, a perfect uniform random source/generator is assumed for the client, as the inputs (& conversely the outputs) are essentially OTP'ed. If the client uses pseudo-random generators, the security is flawed. $\endgroup$
    – JSdJ
    Commented May 8, 2021 at 23:25
  • $\begingroup$ Third, note that for process tomography you need to be able to pick your own well-defined input states for the channel, which the server cannot do. I also don't quite understand what channel the server should be characterizing; the client just makes a state and sends this (in prepare-and-send protocols), or the server just sends something to the client (and, depending on the protocol, there might not even be any input from the client). $\endgroup$
    – JSdJ
    Commented May 8, 2021 at 23:34


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.