# Techniques: Multi-client distributed blind quantum computation with the Qline architecture

## What the paper contributes (one paragraph)
The paper proposes and experimentally demonstrates a lightweight protocol for multi-client blind quantum computation over a linear quantum network (the Qline). Several clients delegate a joint computation to an untrusted remote server while keeping their inputs, algorithm parameters, and outcomes hidden, even against correlated attacks by server nodes colluding with malicious clients. Clients need only trusted single-qubit rotation devices: a maximally entangled state from one untrusted server passes sequentially down the line, each client adding a layer of one-time-pad encryption, before a second untrusted server measures it. Coordination uses secure multi-party computation, which the experiment substitutes with a trusted classical orchestrator to shorten quantum storage time. Security rests on the Remote State Rotation functionality and is proven composably in the Abstract Cryptography framework. A two-client photonic proof of concept shows blindness and correctness for both classical and quantum outputs.

## Techniques offered
- **Multi-client blind (delegated) quantum computation over a Qline** - Lets several weak clients jointly delegate a private quantum computation to untrusted servers using only single-qubit rotations along a linear network, supplying blindness of inputs, algorithm, and outcome to an external joint computation.
  - guarantee: composable blindness (proven in the Abstract Cryptography framework) provided at least one client per Qline is non-corrupt, holding against any strict subset of colluding malicious clients and servers.
  - quote: "In this work, we propose and experimentally demonstrate a lightweight multi-client blind quantum computation protocol based on a novel linear quantum network configuration (Qline)."
- **Secure multi-party computation for client coordination** - Uses classical SMPC (substituted in the experiment by a trusted classical orchestrator) to coordinate the parties and combine their secret keys so the joint computation stays blind even to dishonest colluding clients.
  - guarantee: interactive composable security; the full computation details remain blind to the server and to dishonest clients colluding with it.
  - quote: "enable multi-client BQC by exploiting secure multi-party computation (SMPC), whose aim is to allow several users to collaboratively compute a joint function on their private data."
- **Remote State Rotation delegation primitive** - Reduces each client's role to receiving a single qubit, applying a random z-rotation, and forwarding it, so sequential rotations realise a collaborative Remote State Rotation that drives the blind computation without per-client trusted sources or detectors.
  - guarantee: perfect blindness for delegated universal quantum computation from rotation-only clients with an untrusted server, by reduction to the proven RSR functionality.
  - quote: "From the perspective of an honest client, this behavior is exactly captured by the \emph{Remote State Rotation} (RSR) functionality introduced in~\cite{ma2022qenclave}."

## Where it could apply
- Target primitive(s): measurement-based / cluster-state computation; joint circuits for privacy-preserving quantum machine learning (e.g. federated QML building blocks)
- Target application group(s): quantum_ml, platforms_tooling
- Code family / hardware assumptions: photonic linear quantum networks (polarisation-encoded entangled photons, liquid-crystal rotations, Pockels-cell feed-forward); no error-correcting code; clients limited to trusted single-qubit rotation devices

## Caveats
Full LaTeX source available; nothing paywalled. The protocol supplies blindness only, not verification: the correctness of the computation output is not certified and is left as future work, and whether rotation-only client states suffice for verification is open. The experimental demonstration is a two-client, two-qubit cluster-state proof of concept; scaling needs one Qline per resource qubit (optionally parallelised) and parallel client rotations. The orchestrator must be trusted unless replaced by a composably secure classical SMPC protocol.