Live preview | 30 minutes started quickly!Look at credible distributed AI chain oar architectural design

At 19:00 on August 3rd, Baidu senior R&D engineer Pan Pan was a guest at the Fly Paddle B station live broadcast room and Baidu Super Chain Hundred HomesNo. Live Room, 30 minutes will take you to quickly get started with the architecture design of the trusted distributed AI chain paddle, let's take a look at the technical principles behind the blockchain + AI industrial application!

Live focus>>

• The essence of blockchain

• Architecture design of trusted distributed AI

• Chain paddle core functions

• Chain Paddle Quick Start

Live & Playback Links>>

Long press to identify the QR code to watch

About Chain Paddles>>

PaddleDTX is the first industrial trusted distributed AI open source product jointly launched by Baidu Super Chain and Paddle, dedicated to promoting >Development of distributed privacy model technology.Based on Baidu's self-developed XuperChain, decentralized storage, trusted computing, distributed machine learning and other core technologies, the chain paddle realizes the security and credibility of the whole process from data collection, storage, calculation to circulation, and ensures the training process of distributed AI.The whole process is traceable and auditable.

At the data level, Chainpaddle uses decentralized storage as a trusted data solution.First of all, a multi-copy and multi-node strategy is customized for each piece of data to combine the encrypted shard copy retention proof. The chain paddle has two original storage retention proof mechanisms, namely the time-first variant Merkle tree algorithm and spaceThe preferred copy retention proof algorithm based on elliptic curve bilinear mapping makes it possible to ensure that the data is not tampered with and the storage party cannot collude with a single copy attack, thus ensuring the authenticity of the data; secondly, the chain paddle maintains the proof of the success ratio according to the copy, and evaluate the health of storage nodes with the effective heartbeat frequency, and automatically migrate data according to the health of the nodes, and dynamically adjust the distribution strategy according to the comprehensive judgment of node stability, historical performance and available resources.

At the algorithmic level, Chain Paddle ensures data privacy by using a combination of KMS+TEE during prediction and trainingand the fairness and auditability of the model.In the non-TEE environment, we deploy the public key part of the KMS, and use the variant ECDH to encrypt the data, the encrypted data is decrypted in another part of the KMS in the TEE environment, and passed to the calculation logic program, and finally the calculation result and the correspondingMeasure the report, and store the report in the blockchain network. The blockchain network will verify the authenticity of the computing environment by calling the decentralized oracle network to ensure the credibility of the algorithm.

Finally, Chain Paddle uses DID Distributed Trusted Digital Identity as a solution for trusted computing power.DID is a distributed digital identity based on blockchain technology, which has the characteristics of verifiability and self-sovereignty.The Chain Paddle framework is integrating Baidu DID, which supports users to independently control private data through DID, and independently verify trusted computing power, so as to achieve the goals of data storage security, verifiability, sharing and traceability, and computing power use control.

About the Paddle Hackathon>>

On July 4th, the third phase of the 2022 PaddlePaddle Hackathon was officially launched. The industry's first industry-level trusted distributed AI open source product - PaddleDTX trackRegistration starts at the same time.

This event is a programming event in the field of deep learning for global developers. It is hosted by the National Engineering Research Center for Deep Learning Technology and Application of Fei Pao and is in the form of online claim tasks.to encourage developers to understand and participate in deep learning open source projects.

Developers can submit entries based on PaddleDTX, in addition to receiving the most cutting-edge hard-core trusted distributed AI technical guidance, there is also a chance to win a prize of 10,000 yuan!Click "read the original text" to register for the competition~