Empieza el Agosto 30, 2017 16:00
The path towards secure delegated quantum computing over the internet
When quantum computers finally become available, it is fair to assume (based on the history of ordinary computing) that such devices will be hosted by large institutions and accessed remotely by clients. This has already begun to happen with companies such as DWave and IBM, as well as academic institutions including the University of Bristol, enabling remote access to functioning quantum processors. In instances where the computation involves commercially sensitive or otherwise confidential information, this raises the dual issues of how to insure privacy in such a setting and how to guarantee the correctness of the computation. In particular, the latter captures the idea of verification of quantum computing, i.e., the ability to detect any attempt by a malicious server to deceive the client by tampering with their computation. It remains a major open problem (with important repercussions in complexity theory) whether a classical client can securely delegate a computation to a quantum server in such a way that
the correctness can be verified.
In this talk we will address part of this fundamental issue by taking the first step towards the construction of a blind quantum computing protocol with a completely classical client and single quantum server. We exploit the unique structure of measurement-based quantum computing to show how a classical agent can delegate certain quantum computations to a server while hiding critical details of the structure of the computation. Specifically we show that the transcript of any run of the protocol is consistent with multiple non-equivalent computations. This arises because the movement of quantum information through the resource state is hidden from the server—a feature we call flow ambiguity. This translates into the first protocol for classically driven blind quantum computing: We demonstrate that in a single run of the protocol, the amount of information obtained by the server is bounded below what is necessary to unambiguously distinguish the computation, even up to classical post- processing.