Quantum computers promise surprising powers of computation by exploiting the stunning physical properties of infinitesimally small particles. I focused on designing and proving the security of protocols that allow a purely classical client to use the computational resources of a quantum server, so that the performed computation is never revealed to the server. To this end, I develop a modular tool to generate on a remote server a quantum state that only the client is able to describe, and I show how multi-qubits quantum states can be generated more efficiently. I also prove that there is no such protocol that is secure in a generally composable model of security, including when our module is used in the UBQC protocol. In addition to delegated computation, this tool also proves to be useful for performing a task that might seem impossible to achieve at first sight: proving advanced properties on a quantum state in a non-interactive and non-destructive way, including when this state is generated collaboratively by several participants. This can be seen as a quantum analogue of the classical Non-Interactive Zero-Knowledge proofs. This property is particularly useful to filter the participants of a protocol without revealing their identity, and may have applications in other domains, for example to transmit a quantum state over a network while hiding the source and destination of the message. Finally, I discuss my ongoing independent work on One-Time Programs, mixing quantum cryptography, error correcting codes and information theory.

Publication

Study of Protocols Between Classical Clients and a Quantum Server

Quantum computers promise surprising powers of computation by exploiting the stunning physical properties of infinitesimally small particles. I focused on designing and proving the security of protocols that allow a purely classical client to use the computational resources of a quantum server, so that the performed computation is never revealed to the server. To this end, I develop a modular tool to generate on a remote server a quantum state that only the client is able to describe, and I show how multi-qubits quantum states can be generated more efficiently. I also prove that there is no such protocol that is secure in a generally composable model of security, including when our module is used in the UBQC protocol. In addition to delegated computation, this tool also proves to be useful for performing a task that might seem impossible to achieve at first sight: proving advanced properties on a quantum state in a non-interactive and non-destructive way, including when this state is generated collaboratively by several participants. This can be seen as a quantum analogue of the classical Non-Interactive Zero-Knowledge proofs. This property is particularly useful to filter the participants of a protocol without revealing their identity, and may have applications in other domains, for example to transmit a quantum state over a network while hiding the source and destination of the message. Finally, I discuss my ongoing independent work on One-Time Programs, mixing quantum cryptography, error correcting codes and information theory.