CN113052331A - Block chain-based Internet of things personalized federal learning method - Google Patents

Abstract

Translated fromChinese

本发明公开一种基于区块链的物联网个性化联邦学习方法，涉及区块链技术领域，该方法包括：终端设备和边缘计算设备注册和认证，各个终端设备和边缘计算设备向区块链进行注册，区块链对各个设备进行认证，并发放证书；区块链智能合约创建联邦学习任务，初始化训练模型及参数；终端设备加载数据样本，卸载到边缘计算设备上，进行本地模型训练；边缘计算设备将本地训练模型参数加密之后，上传至区块链，区块链节点共识通过之后，产生新的区块；智能合约对模型参数进行聚合，聚合模型参数，并更新整体的模型；智能合约判断是否达到模型预设收敛条件，如果没有则进行下一轮训练，如果到达则终止联邦学习任务；边缘计算设备基于全局模型信息结合自己数据，训练个性化模型。本发明解决传统的联邦学习的全局模型却无法满足物联网设备在存储计算和通信能力方面的异构性，同时提高了隐私数据的安全性以及系统的拜占庭容错性。

The invention discloses a blockchain-based IoT personalized federated learning method, which relates to the technical field of blockchain. The method includes: registration and authentication of terminal devices and edge computing devices; For registration, the blockchain authenticates each device and issues certificates; the blockchain smart contract creates a federated learning task, initializes the training model and parameters; the terminal device loads the data sample, unloads it to the edge computing device, and conducts local model training; The edge computing device encrypts the local training model parameters and uploads them to the blockchain. After the blockchain node consensus is passed, a new block is generated; the smart contract aggregates the model parameters, aggregates the model parameters, and updates the overall model; intelligent The contract judges whether the preset convergence conditions of the model are reached. If not, the next round of training will be performed. If so, the federated learning task will be terminated; the edge computing device will train the personalized model based on the global model information combined with its own data. The invention solves the problem that the traditional global model of federated learning cannot satisfy the heterogeneity of the Internet of Things devices in terms of storage computing and communication capabilities, and at the same time improves the security of private data and the Byzantine fault tolerance of the system.

Description

Block chain-based Internet of things personalized federal learning method

Technical Field

The invention relates to the field of block chains, in particular to a block chain-based Internet of things personalized federal learning method.

Background

The federated learning system is used as a learning method for collaboratively training a unified model by contributing local data by a plurality of clients, and has the limitation that the model only depends on a single central server and is easily influenced by server faults. The purpose of traditional federal learning is to obtain a globally shared model for all participants. However, when the data distribution of each participant is inconsistent, the global model cannot meet the performance requirement of each federal learning participant, and some participants cannot even obtain a model better than a model trained by only using local data.

Based on the method, the personalized federal learning method of the Internet of things based on the block chain is provided, the problem of heterogeneity in the environment of the Internet of things is solved, the federal learning byzantine fault tolerance is improved, and the privacy of data is improved.

Disclosure of Invention

In view of the above defects in the prior art, the technical problem to be solved by the present invention is to provide a personalized federal learning method for internet of things based on a blockchain, which is provided by the present invention, in view of the needs and disadvantages of the current technical development.

Firstly, the invention provides a block chain-based federal learning incentive method, and the technical scheme adopted for solving the technical problems is as follows:

s01, registering and authenticating terminal equipment and edge computing equipment, registering each terminal equipment and edge computing equipment to a block chain, authenticating each equipment by the block chain, and issuing a certificate;

s02, creating a federal learning task by a block chain intelligent contract, and initializing a training model and parameters;

s03, loading the data sample by the terminal equipment, unloading the data sample to the edge computing equipment, and carrying out local model training;

s04, the edge computing equipment encrypts the local training model parameters and uploads the parameters to a block chain, and after the block chain link points are identified together, a new block is generated;

s05, the intelligent contract aggregates the model parameters, aggregates the model parameters and updates the whole model; the intelligent contract judges whether a preset convergence condition of the model is reached, if not, the next round of training is carried out, and if so, the federal learning task is terminated;

and S06, training the personalized model by the edge computing equipment based on the global model information and combining the data of the edge computing equipment.

Specifically, the related internet of things equipment can be various intelligent terminals, such as mobile phones, glasses, watches, cameras, monitoring equipment and the like. The client side is authenticated through the block chain, and the authentication mode can be a digital certificate, token authentication and other authentication modes. The blockchain may be a public chain, a federation chain, or a private chain. The intelligent contracts are run on a block chain, and a user can create a federal learning task through a template. The loading data samples of the terminal equipment are unloaded to the edge computing equipment and are transmitted in an encryption mode, and the local model training is a gradient training method. The encryption mode can be an asymmetric encryption algorithm such as an RSA algorithm, a DSA algorithm, an ECC algorithm, a DH algorithm and the like. The model parameters are aggregated according to the records on the block chain by an intelligent contract, and the integral model is updated. The received model parameter updating method can use weighted average, and can also be median value. And the block chain transmits the updated model to each participating edge computing device, so that a new round of training and learning is started. Each device redefines the global model according to local data of the device to obtain an individualized model, wherein the global model comprises low-level parameters and high-level parameters, and the individualized model learns the specific characteristics of the device by finely adjusting the high-level parameters of the global model through the local data.

Drawings

Some specific embodiments of the invention will be described in detail hereinafter, by way of illustration and not limitation, with reference to the accompanying drawings. The same reference numbers in the drawings identify the same or similar elements or components. Those skilled in the art will appreciate that the drawings are not necessarily drawn to scale. The objects and features of the present invention will become more apparent in view of the following description taken in conjunction with the accompanying drawings, in which:

fig. 1 is a block chain-based personalized federal learning flow chart of the internet of things.

FIG. 2 is a schematic diagram of the system architecture of the present invention.

Detailed Description

In order to make the present invention more comprehensible with respect to its gist, the present invention will be further described with reference to the accompanying drawings and examples. In the following description, numerous specific details and specific examples are set forth in order to provide a more thorough understanding of the present invention and to provide a thorough understanding of the present invention. While this invention is susceptible of embodiment in many different forms than that described herein, there will be many equivalents to those skilled in the art which incorporate such variations and modifications without departing from the spirit and scope of the invention as defined by the appended claims and their equivalents.

The technology of the present invention will be described in detail below with reference to the accompanying drawings in conjunction with specific embodiments.

The embodiment of the invention provides an Internet of things personalized federal learning method based on a block chain, which comprises the following implementation processes:

s1, terminal devices C1, C2, C3 and C4 and edge computing devices E1, E2, E3 and E4 are respectively registered with a blockchain B, the blockchain B verifies registration information of the devices, and certificates are issued to the devices after the certificates are authenticated.

S2, a user U1 creates an intelligent contract S1 according to the intelligent contract template learned by the federation, the intelligent contract S1 runs on a block chain B, a federated learning task T is created, and a training model M and a parameter omega are initialized.

S3, the edge computing devices E1, E2, E3 and E4 download a training model M and a parameter omega from the block chain B respectively, the terminal devices C1, C2, C3 and C4 load data samples respectively and unload the data samples to the edge computing devices connected respectively, the training is carried out through local computing by executing a program P, a gradient descent method is adopted for training, and the gradient computing formula is as follows:

and locally updating the model parameters omega, and uploading the updated parameters omega to the block chain B in an encrypted manner.

And S4, after encrypting the local training model parameters respectively by the edge computing equipment E1, E2, E3 and E4, uploading the parameters to a block chain B, and after passing consensus, generating a new block.

S5, the intelligent contract S1 aggregates the model parameters, aggregates the model parameters omega, and updates the whole model M; the polymerization formula is as follows:

and the intelligent contract judges whether a preset convergence condition of the model is reached, if not, the next round of training is carried out, and if so, the federal learning task is terminated.

And S6, training the personalized model M by the edge computing devices E1, E2, E3 and E4 based on the information of the global model M and the data of the edge computing devices.

Claims (10)

Translated fromChinese

1.一种基于区块链的物联网个性化联邦学习方法，其特征在于，该方法的实现过程包括：1. A blockchain-based personalized federated learning method for the Internet of Things, characterized in that the implementation process of the method comprises:S01、终端设备和边缘计算设备注册和认证，各个终端设备和边缘计算设备向区块链进行注册，区块链对各个设备进行认证，并发放证书；S01. Registration and certification of terminal devices and edge computing devices, each terminal device and edge computing device are registered with the blockchain, and the blockchain certifies each device and issues certificates;S02、区块链智能合约创建联邦学习任务，初始化训练模型及参数；S02. The blockchain smart contract creates a federated learning task, and initializes the training model and parameters;S03、终端设备加载数据样本，卸载到边缘计算设备上，进行本地模型训练；S03, the terminal device loads the data sample, unloads it to the edge computing device, and performs local model training;S04、边缘计算设备将本地训练模型参数加密之后，上传至区块链，区块链节点共识通过之后，产生新的区块；S04. The edge computing device encrypts the parameters of the local training model and uploads it to the blockchain. After the blockchain node consensus is passed, a new block is generated;S05、智能合约对模型参数进行聚合，聚合模型参数，并更新整体的模型；智能合约判断是否达到模型预设收敛条件，如果没有则进行下一轮训练，如果到达则终止联邦学习任务；S05. The smart contract aggregates the model parameters, aggregates the model parameters, and updates the overall model; the smart contract determines whether the model preset convergence condition is reached, if not, the next round of training is performed, and if it is reached, the federated learning task is terminated;S06、边缘计算设备基于全局模型信息结合自己数据，训练个性化模型。S06 , the edge computing device trains a personalized model based on the global model information combined with its own data.2.根据权利要求1所述的一种基于区块链的物联网个性化联邦学习方法，其特征在于，所述步骤S01中物联网设备可以是各种智能终端，如手机、眼镜、手表、摄像头、监控设备等。2. A blockchain-based IoT personalized federated learning method according to claim 1, characterized in that, in the step S01, the IoT devices can be various intelligent terminals, such as mobile phones, glasses, watches, Cameras, surveillance equipment, etc.3.根据权利要求1所述的一种基于区块链的物联网个性化联邦学习方法，其特征在于，所述步骤S01中认证方式可以是数字证书、令牌认证等认证方式。3 . The blockchain-based IoT personalized federated learning method according to claim 1 , wherein the authentication method in the step S01 can be an authentication method such as digital certificate and token authentication. 4 .4.根据权利要求1所述的一种基于区块链的物联网个性化联邦学习方法，其特征在于，所述步骤S01中区块链可以是公有链、联盟链、私有链。4. A blockchain-based IoT personalized federated learning method according to claim 1, wherein in the step S01, the blockchain can be a public chain, a consortium chain, or a private chain.5.根据权利要求1所述的一种基于区块链的物联网个性化联邦学习方法，其特征在于，所述步骤S02中智能合约是运行在区块链上，用户可以通过模板创建联邦学习任务。5. A blockchain-based IoT personalized federated learning method according to claim 1, characterized in that, in the step S02, the smart contract is run on the blockchain, and the user can create federated learning through a template Task.6.根据权利要求1所述的一种基于区块链的物联网个性化联邦学习方法，其特征在于，所述步骤S03终端设备加载数据样本卸载到边缘计算设备上，是通过加密的方式进行传输，本地模型训练是通过梯度训练的方法。6. A blockchain-based personalized federated learning method for the Internet of Things according to claim 1, characterized in that, in step S03, the terminal device loads data samples and unloads them to the edge computing device, which is performed in an encrypted manner. Transfer, local model training is a method of training via gradients.7.根据权利要求1所述的一种基于区块链的物联网个性化联邦学习方法，其特征在于，所述步骤S04加密方式可以是RSA算法、DSA算法、ECC算法、DH算法等非对称加密算法。7. A blockchain-based personalized federated learning method for the Internet of Things according to claim 1, wherein the encryption method of step S04 can be asymmetric such as RSA algorithm, DSA algorithm, ECC algorithm, DH algorithm, etc. Encryption Algorithm.8.根据权利要求1所述的一种基于区块链的物联网个性化联邦学习方法，其特征在于，所述步骤S05将收到的模型参数进行聚合，是通过智能合约依据区块链上的记录，聚合模型参数，并更新整体的模型。对收到的模型参数更新方法可以使用加权平均，也可以是中位数取值。8. A blockchain-based personalized federated learning method for the Internet of Things according to claim 1, wherein the step S05 aggregates the received model parameters, which is based on a smart contract on the blockchain. records, aggregates model parameters, and updates the overall model. The update method for the received model parameters can use a weighted average or a median value.9.根据权利要求1所述的一种基于区块链的物联网个性化联邦学习方法，其特征在于，所述步骤S05下一轮训练，区块链将更新之后的模型下发至各参与边缘计算设备，从而开始新一轮的训练学习。9. A blockchain-based IoT personalized federated learning method according to claim 1, wherein in the next round of training in step S05, the blockchain sends the updated model to each participant Edge computing devices, thus starting a new round of training and learning.10.根据权利要求1所述的一种基于区块链的物联网个性化联邦学习方法，其特征在于，所述步骤S06训练个性化模型，每个设备根据自己的本地数据重新定义全局模型得到个性化模型，其中全局模型分低层参数和高层参数，个性化模型是通过本地数据对全局模型的高层参数进行微调，来学习设备的特定特征。10. A blockchain-based personalized federated learning method for the Internet of Things according to claim 1, wherein the step S06 trains the personalized model, and each device redefines the global model according to its own local data to obtain. Personalized model, in which the global model is divided into low-level parameters and high-level parameters. The personalized model uses local data to fine-tune the high-level parameters of the global model to learn the specific characteristics of the device.

Images