but,
but,ARPACo-founder and COO Xu Maotong believes that although AI+blockchain sounds beautiful, there are still many problems and challenges in its implementation.
The first step in data on-chain isGuarantee the security of data encryption. For example, data as an important asset is easily copied and leaked during circulation and exchange. There is a gray area in the traditional data trading platform, and data contributors cannot get benefits. At the same time, it is increasingly difficult for large companies with massive amounts of data to monetize their data under increasingly stringent privacy protection regulations.
Odaily has reportedOasis Labs、TRIAS、TaxaThey all chose the technical path of building a trusted execution environment (Trusted Execution Environment, TEE) with trusted hardware. Although TEE balances security and cost, its ability to resist attacks is still lower than SE (smart card), and must be securely activated. this week,Intel Discloses Third SGX TEE Vulnerability of the Year - "L1 Terminal Fault" (L1TF)。
In addition to TEE,sMPC (Secure Multi-party Computation, Secure Multi-party Computation)It is also a decentralized off-chain solution for encrypted computing, and does not depend on hardware.
MPC was first proposed in 1982 by Yao Qizhi, a Chinese computer scientist and Turing Award winner, to solve the "millionaire problem" - how can two millionaires not reveal their true identity without a trusted third party? Property status, compare who is richer. The working principle of MPC can be simply understood as that multiple parties hold calculation data, execute a calculation logic (such as finding the maximum value) together without decryption, and obtain plaintext calculation results. During the whole process, the original data and calculation model are encrypted and hidden. This ensures that the data is not leaked throughout the process, and also separates the right to use and ownership of the data.
At the beginning of last year, as an investor, Xu Maotong, who was looking at AI, blockchain and financial technology projects at Fosun Ruizheng Capital, got acquainted withData annotation company StardustFounder and CEO Zhang Lei. The two are optimistic about the prospects of the blockchain and hope to solve the pain points of "data encryption and privacy protection" in the data field, so they jointly createdARPA, an off-chain privacy computing network based on MPC encryption。
ARPA can cooperate with existing high-performance public chains and has the followingthree characteristics：
Verifiability of calculation process and results;
The confidential operation process is thousands of times faster than homomorphic encryption;
Data security can still be guaranteed even when most nodes are doing evil.
Among them, "the verifiability of calculation process and results" belongs to the characteristics of MPC. Let me briefly talk about the principle behind it. If any party modifies the data during the data transfer process or does not operate according to the MPC protocol, different MAC values will be generated to determine whether it has been tampered with. Since the process of calculation and verification is separated and a consensus is reached under the chain, it isAvoid redundant calculations and low performance in the blockchain。
MPC itself is suitable for general-purpose computing, and network nodes are not necessarily blockchain nodes.The relationship between ARPA and the public chain can be imagined as parallel. For example, the ARPA computing network receives a request from Ethereum, the smart contract is triggered, the computing network starts to execute the encryption algorithm task, and verifies whether the result is correct, and then returns the data to Ethereum through the consensus layer. The data in the smart contract is converted into a bytecode that can be recognized by the MPC network through a compiler, and is transmitted across the chain according to the Gossip protocol.
image description
In recent years, the performance of MPC
Overseas, some companies that are also optimistic about the application prospects of MPC, such asSharemind，Enigmasecret contractsecret contractIn China, there are not many MPC researchers, and Zhang Lei has accumulated rich industrial resources as a service provider, which can help ARPA form a certain first-mover advantage.
ARPA's partners now include Stardust, ANZ, Credit Karma, Quantitative, etc. The main way of cooperation is to jointly explore the application of MPC in high-value data sharing (such as financial scenarios). MPC can collect the data of competitors in the industry (such as banks, retailers, airlines, etc.), and achieve the hiding of back-end data and the privatization of smart contracts. The security is much higher than traditional data desensitization methods.
ARPA can charge computing and storage service fees (legal currency or token) to B. Token is mainly used for security verification, if the node does evil, it will lose token.
Of course, the combination of MPC and blockchain is still in the initial stage, and it will still take some time for commercial implementation. Therefore, in terms of scientific research, ARPA cooperates with NYU, Zhejiang University, Hong Kong Polytechnic and Leuven University,And will continue to supplement system layer R&D personnel。
According to the roadmap, ARPA plans to launch the Proof of concept (POC) in the MPC direction around October, through the logic of offline computing and privatized smart contracts, and strive to test the network early next year.
According to reports, ARPA received a multi-million dollar cornerstone round of financing from Arrington XRP, GBIC, Connect Capital, Ledger Capital, Coefficient Ventures, LYVC and other institutions in June this year, and plans to launch a new round of financing in September.
secondary title
Related academic papers
Semi-Homomorphic Encryption and Multiparty Computation
Practical Covertly Secure MPC for Dishonest Majority – or: Breaking the SPDZ Limits
