A trap based technique for verification of quantum computations

We present a new verification technique where the correctness of a `target' computation is verified by means of several other classically efficiently predictable `trap' computations run on the same quantum computer. We show that if the target computation is corrupted, then it is likely that the outputs of the trap computations do not correspond to the expected ones. Our results hold when it is possible to prepare single-qubit states in an ideal and noise-free way, while gates and measurements are error-prone and imperfect. Alternatively, they also hold when it is possible to perform measurements in a discrete set of bases in an ideal way, whereas state preparation and gates are imperfect. Our technique is different from existing trap based approaches and offers unique advantages. As an example, it reduces the types of single-qubit states that need to be prepared in an ideal way, or alternatively the number of bases for ideal measurements.