Magic Secret Sharing: Threshold Control of Quantum Computational Power via GHZ Entanglement
Abstract
We introduce Magic Secret Sharing (MSS), a quantum cryptographic primitive in which the secret is the computational capability of a quantum state rather than its classical description. In the resource theory of magic, non-stabilizer states fuel universal quantum computation via non-Clifford gates; MSS distributes this resource with an (n-1,n) threshold structure using a pre-shared GHZ state and a single local phase gate P(phi) = diag(1, exp(i*phi)). Any individual party holds the maximally mixed state I/2, with Wigner distance C(I/2) = 0, so no local operation can yield non-Clifford computational advantage regardless of what operations are applied or what noise acts on the device. The authorised coalition reconstructs magic content C(phi) = (|sin(phi)| + |cos(phi)| - 1)/2 exactly, enabling a logical T gate via gate teleportation in multi-server blind quantum computation (BQC). Among diagonal parametric gates, phase gates are the unique class satisfying the security condition, characterised via an exact column-sum condition. The protocol is elevated to a one-sided device-independent (1SDI) setting via a steering inequality: the assemblage produced on the recipient's side certifies magic delivery without trusting the coalition's devices. We demonstrate the (2,3) instance on ibm_marrakesh (156-qubit IBM Heron): security (C(rho_Bob) < 10^-6, the linear-programming solver tolerance) holds in every run and is independently reproduced on a second qubit assignment, and state fidelity reaches 0.959-0.973 for the authorised party, with faithfulness confirmed for all four test values of phi to within 0.025 in magic content, a residual that depolarising noise alone accounts for.
Related
- Sharp universal death of entanglement threshold for Pauli Hamiltonians
- Quantum secret sharing in a tripartite superconducting network
- Combinatorial aspects of holographic quantum secret sharing
- Entanglement Harvesting Landscapes: A Common Computational Framework
- From sharing propensity to diffusion control: temporal catch-up costs of delayed misinformation correction
BibTeX
@misc{dutta2026magic,
title = {{Magic Secret Sharing: Threshold Control of Quantum Computational Power via GHZ Entanglement}},
author = {Soumyojyoti Dutta and Tushar},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2605.16614},
eprint = {2605.16614},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2605.16614},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Magic Secret Sharing: Threshold Control of Quantum Computational Power via GHZ Entanglement
AU - Soumyojyoti Dutta
AU - Tushar
PY - 2026
DA - 2026/10/06
PB - arXiv
DO - 10.48550/arxiv.2605.16614
UR - https://arxiv.org/abs/2605.16614
AB - We introduce Magic Secret Sharing (MSS), a quantum cryptographic primitive in which the secret is the computational capability of a quantum state rather than its classical description. In the resource theory of magic, non-stabilizer states fuel universal quantum computation via non-Clifford gates; MSS distributes this resource with an (n-1,n) threshold structure using a pre-shared GHZ state and a single local phase gate P(phi) = diag(1, exp(i*phi)). Any individual party holds the maximally mixed state I/2, with Wigner distance C(I/2) = 0, so no local operation can yield non-Clifford computational advantage regardless of what operations are applied or what noise acts on the device. The authorised coalition reconstructs magic content C(phi) = (|sin(phi)| + |cos(phi)| - 1)/2 exactly, enabling a logical T gate via gate teleportation in multi-server blind quantum computation (BQC). Among diagonal parametric gates, phase gates are the unique class satisfying the security condition, characterised via an exact column-sum condition. The protocol is elevated to a one-sided device-independent (1SDI) setting via a steering inequality: the assemblage produced on the recipient's side certifies magic delivery without trusting the coalition's devices. We demonstrate the (2,3) instance on ibm_marrakesh (156-qubit IBM Heron): security (C(rho_Bob) < 10^-6, the linear-programming solver tolerance) holds in every run and is independently reproduced on a second qubit assignment, and state fidelity reaches 0.959-0.973 for the authorised party, with faithfulness confirmed for all four test values of phi to within 0.025 in magic content, a residual that depolarising noise alone accounts for.
ER -