Classical Client Controls Malicious Quantum Server via LWE-Hardness Protocol on arXiv

The protocol constructs a maliciously secure two-party computation that lets the client choose and execute arbitrary quantum circuits, prepare states, and maintain hidden quantum memory across rounds. It directly yields negligible-soundness classical verification of quantum sampling and remote preparation of any efficiently preparable state, two open problems the authors resolve. Methods rest on a general-purpose classical-quantum MPC primitive whose security reduces to learning-with-errors; the paper supplies an explicit construction and sketches the reduction.

Prior blind-quantum-computation schemes either required multiple non-communicating servers or trusted setup; this work collapses the requirement to a single server while removing setup. It also strengthens earlier LWE-based verification results by achieving full malicious security rather than honest-but-curious. The advance matters for near-term quantum cloud services, where users cannot verify hardware integrity yet need private, composable access.

Efficiency remains unquantified beyond asymptotic claims; concrete gate overhead and round complexity are left for follow-up. A practical limitation is the assumption that LWE stays hard against quantum attacks—an assumption that could shift with new cryptanalysis. Full end-to-end implementation on existing hardware plus a published security proof in the UC framework would materially strengthen the result.

Next steps include integrating the primitive into quantum-supremacy verification pipelines and testing remote state preparation on ion-trap or superconducting testbeds within three years.