Paper 2026/1733

TEE Server-Assisted Aggregated Offline Deployment Scheme for Multiplication Triples

Abstract

With the growing deployment of secure multi-party computation (MPC) in data-intensive applications, the offline generation and distribution of authenticated multiplication triples has become a key scalability bottleneck. Existing software-only preprocessing protocols, such as MASCOT and LowGear, typically incur substantial interaction and computation costs when the number of participants or the triple demand is large. This paper proposes a trusted execution environment (TEE) server–assisted aggregated offline deployment scheme that moves expensive interactive preprocessing into a remotely attested enclave and distributes participant-specific authenticated triple shares over authenticated-encrypted channels. Conceptually, the enclave serves as a TEE-backed pseudorandom correlation generator (PCG) that outputs authenticated correlated randomness for MPC with one-way delivery. We design an end-to-end workflow covering remote attestation, per-participant session establishment, and encrypted distribution. The proposed protocol applies to both semi-honest and malicious adversaries; we provide a security analysis against malicious adversaries with abort. Experimental results show clear efficiency improvements over MP-SPDZ implementations of MASCOT and LowGear. Specifically, the proposed scheme achieves a generation rate three times that of MASCOT, while exhibiting linear scalability up to 100k participants. Furthermore, by employing a streaming processing strategy, it efficiently supports the generation of up to 10M triples with an amortized time significantly lower than that of purely software-based protocols, enabling practical and highly scalable preprocessing for large-scale heterogeneous MPC deployments.