Privacy-Preserving and Publicly Verifiable Matrix Multiplication
Jing Liu, Liang Feng Zhang · IEEE Transactions on Services Computing · 2022
Outsourcing computations allows the resource-constrained clients (such as, IoT devices) to offload heavy computations to powerful cloud servers. At the same time, it brings many challenges such as data privacy, result verification and fair payment. In this paper, we propose a privacy-preserving multi-function verifiable computation (MFVC) model and construct an MFVC scheme for outsourcing matrix multiplication computations (MMC), which have many real-life applications, such as machine learning. Our scheme keeps one of the matrices in MMCsemantically secureand allows apublic verificationof the server's work. In particular, the client's work isfasterthan the native MMC for square matrices of order$\geq 7000$, while the best existing scheme requires matrices of order$\geq 270000$. We also propose a smart contract-based framework that equips the proposed scheme withfair paymentproperty. By offloading the verification to the blockchain, the client's work is faster than the native MMC for square matrices of order$\geq 6000$.