Partial Blind Quantum Computation (2503.10007v1)

Abstract: Quantum computing is rapidly advancing toward cloud-based services, raising significant concerns about the privacy and security of computations outsourced to untrusted quantum servers. Traditional Blind Quantum Computation (BQC) protocols enable clients with limited quantum resources to delegate computations while concealing both inputs and circuit details. However, applying BQC uniformly to an entire quantum circuit incurs additional quantum resources and computational overhead, which can be a significant burden in practical implementations. In many cases, such as Grover's algorithm, only specific subroutines-like oracles-contain sensitive information, while the rest of the circuit does not require the same level of protection. Therefore, selectively applying BQC to critical components can enhance computational efficiency while maintaining security. In this work, we propose a selective application of BQC that targets only the critical components of quantum circuits. By integrating techniques from Quantum Homomorphic Encryption (QHE) and Universal Blind Quantum Computation (UBQC), our approach secures the sensitive subcircuits while allowing the remaining, non-sensitive portions to be executed more efficiently. In our framework, BQC-protected sections output quantum states that are encrypted via bit-flip and phase-flip operations, and we devise a mechanism based on selective X and Z gate corrections to seamlessly interface these with unprotected sections. We provide a security analysis demonstrating that our selective BQC approach preserves universality, correctness, and blindness, and we illustrate its practical advantages through an application to Grover's algorithm. This work paves the way for more efficient and practical secure quantum computing on near-term devices.