CN117094773A - Online migration learning method and system based on blockchain privacy calculation - Google Patents
Online migration learning method and system based on blockchain privacy calculation Download PDFInfo
- Publication number
- CN117094773A CN117094773A CN202311186226.0A CN202311186226A CN117094773A CN 117094773 A CN117094773 A CN 117094773A CN 202311186226 A CN202311186226 A CN 202311186226A CN 117094773 A CN117094773 A CN 117094773A
- Authority
- CN
- China
- Prior art keywords
- data
- model
- local
- blockchain
- user
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 47
- 238000004364 calculation method Methods 0.000 title claims abstract description 29
- 238000013508 migration Methods 0.000 title claims abstract description 25
- 230000005012 migration Effects 0.000 title claims abstract description 25
- 230000002776 aggregation Effects 0.000 claims description 30
- 238000004220 aggregation Methods 0.000 claims description 30
- 238000012545 processing Methods 0.000 claims description 26
- 238000004422 calculation algorithm Methods 0.000 claims description 20
- 230000008569 process Effects 0.000 claims description 17
- 238000012549 training Methods 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 14
- 238000001303 quality assessment method Methods 0.000 claims description 12
- 230000006399 behavior Effects 0.000 claims description 11
- 230000007246 mechanism Effects 0.000 claims description 10
- 238000012795 verification Methods 0.000 claims description 10
- 230000006870 function Effects 0.000 claims description 8
- 238000012550 audit Methods 0.000 claims description 7
- 238000013480 data collection Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 claims description 4
- 238000013441 quality evaluation Methods 0.000 claims description 3
- 238000012544 monitoring process Methods 0.000 claims description 2
- 230000004888 barrier function Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 7
- 238000013526 transfer learning Methods 0.000 description 5
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000001105 regulatory effect Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000002574 poison Substances 0.000 description 1
- 231100000614 poison Toxicity 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0269—Targeted advertisements based on user profile or attribute
- G06Q30/0271—Personalized advertisement
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
- G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
- G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/096—Transfer learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0277—Online advertisement
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- Bioethics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Business, Economics & Management (AREA)
- General Engineering & Computer Science (AREA)
- Medical Informatics (AREA)
- Biophysics (AREA)
- Computer Hardware Design (AREA)
- Databases & Information Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computer Security & Cryptography (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The application discloses an online migration learning method and system based on blockchain privacy calculation, which relate to the field of privacy calculation, in particular to online migration learning. The online migration learning method based on blockchain privacy calculation can break the island barriers of data, effectively combine multiparty data under the condition of protecting local data privacy, realize open sharing of the data and facilitate value circulation of the data.
Description
Technical Field
The application belongs to a blockchain privacy computing technology and transfer learning, and particularly relates to an online transfer learning method and system based on blockchain privacy computing.
Background
In the digital society, there is a strong demand for data production, and a large amount of data is required for user service and business marketing, especially in a distributed collaboration business mode, each party hopes that the data can smoothly circulate and reasonably represent the data value, wherein two problems occur: data islands and data privacy issues. The federal learning is used as a distributed machine learning paradigm, so that the problem of data islanding can be effectively solved, and the participants can jointly model on the basis of not sharing data. In addition, the blockchain is essentially a distributed account book, each full-block node stores a complete same account book, and the stored data is safe, reliable and tamper-proof by using a trust mechanism constructed by encryption algorithm, consensus mechanism and other technologies. The method uses users as centers, exchanges data under the premise of safety privacy, provides high-quality and compliant services, and is a trend of digital social construction.
In practical application scenarios, the sample space and the feature space contained in the data from different sources are different, and conventional federal learning cannot cope with the above problems. The focus of transfer learning is to identify from the source domain that knowledge in the target domain can be utilized to assist in learning and completing tasks on the target domain. In addition, samples of the target area are often not available at one time, so online learning methods are applied to solve this problem.
Disclosure of Invention
Based on the background and the problems existing in the prior art, the application adopts the following technical scheme: in a first aspect, an online migration learning method based on blockchain privacy calculation is provided, which can utilize the online migration learning method to assist learning tasks in different fields, and can also utilize blockchain privacy calculation to protect privacy security of local data in different fields.
An online migration learning method based on blockchain privacy calculation comprises the following steps:
receiving n pieces of local data in a trusted execution environment, the n pieces of local data including n-1 source domain data and one target domain data from a plurality of sources, wherein the source domain data includes: advertisement data, user data and advertisement click data of a user, wherein the target domain data is new user data;
respectively training the n-1 source domain data under the trusted execution environment to obtain n-1 local models, wherein the intermediate parameters of the local models are adjusted and monitored based on a local differential privacy algorithm;
the block chain nodes collect all local models to be aggregated to form a new model, and the new model is respectively sent to all the block chain nodes;
and downloading a new model from the block chain node, recommending advertisements for new users in the target domain by using the new model, and updating the new model based on feedback data recommended by the advertisements of the new users.
As an implementation manner, the local data set is used for an e-commerce advertisement recommendation task, including an advertisement click behavior log of a user, characteristics of the advertisement, characteristics of the user and shopping behaviors of the user;
the user's advertisement click log contains the fields: user ID, timestamp, advertisement ID, resource display position and whether click, wherein, whether click is represented by 0/1, 0 represents no click, 1 represents click;
the characteristics of the advertisement itself include: advertisement ID, commodity category ID, brand ID, commodity price;
the characteristics of the user themselves include: user ID, gender, age level, consumption level, shopping depth, city level;
the shopping behavior of the user includes: user ID, timestamp, behavior type, merchandise category ID, and brand ID, wherein the behavior type includes: browse, join shopping carts, like and purchase.
As an implementation manner, the method for adjusting and monitoring the intermediate parameters of the local model based on the local differential privacy algorithm includes the following steps:
adding noise to the intermediate parameters in the local model updating process by adopting a Laplace mechanism, and if the calculation tasks are the same, the calculation result after noise addition is similar to the calculation result of the original data;
and updating the intermediate parameters after noise addition after each iteration until the sum of privacy budgets accumulated by the iterations exceeds a preset budget constraint, and ending the iterations.
As an implementation manner, the step of collecting all local models by the blockchain node to perform model aggregation, iterating a plurality of times, and sending a new aggregated model to all blockchain nodes includes:
the block chain nodes collect local models participating in training;
performing aggregation treatment on all the participated local models by adopting a self-adaptive model aggregation algorithm to obtain a new model;
and iterating the new model until the global loss function converges or reaches the preset precision, and respectively transmitting the aggregated new model to all the participated blockchain nodes.
As an implementation manner, the aggregation processing of all the participating local models by adopting the adaptive model aggregation algorithm comprises the following steps:
calculating a quality assessment score for each blockchain node based on the cross entropy loss function;
calculating the credit value score of each blockchain node in the cooperation process by adopting a cooperator asynchronous parameter audit mechanism;
and adjusting model aggregation weights through the quality evaluation scores of the blockchain nodes and the reputation scores in the cooperation process, and carrying out aggregation processing on all the participating local models based on the model aggregation weights.
As an implementation manner, the collaborator asynchronous parameter auditing mechanism includes the following steps:
encrypting the model quality audit data of the verified blockchain node and then transmitting the encrypted model quality audit data to other blockchain nodes for verification;
other block chain nodes verify parameters of the verified model based on own local data and broadcast verification results to other block chain nodes;
and after the verification of all the block chain nodes is passed, obtaining a final verification result.
An online migration learning system based on blockchain privacy calculation comprises a data collection module, a first processing module, a second processing module and a third processing module:
the data collection module is configured to receive n pieces of local data in a trusted execution environment, where the n pieces of local data include n-1 pieces of source domain data from a plurality of sources and one piece of target domain data, and the source domain data includes: advertisement data, user data and advertisement click data of a user, wherein the target domain data is new user data;
the first processing module respectively trains the n-1 source domain data under the trusted execution environment to obtain n-1 local models, wherein the intermediate parameters of the local models are adjusted and monitored based on a local differential privacy algorithm;
the second processing module collects all local models by the block chain link points to be aggregated to form a new model, and the new model is respectively sent to all block chain nodes;
and the third processing module is used for downloading a new model from the blockchain node, recommending advertisements for new users in the target domain by using the new model, and updating the new model based on feedback data recommended by the advertisements of the new users.
A computer readable storage medium storing a computer program, the computer program when executed by a processor implementing a method of:
receiving n pieces of local data in a trusted execution environment, the n pieces of local data including n-1 source domain data and one target domain data from a plurality of sources, wherein the source domain data includes: advertisement data, user data and advertisement click data of a user, wherein the target domain data is new user data;
respectively training the n-1 source domain data under the trusted execution environment to obtain n-1 local models, wherein the intermediate parameters of the local models are adjusted and monitored based on a local differential privacy algorithm;
the block chain nodes collect all local models to be aggregated to form a new model, and the new model is respectively sent to all the block chain nodes;
and downloading a new model from the block chain node, recommending advertisements for new users in the target domain by using the new model, and updating the new model based on feedback data recommended by the advertisements of the new users.
An online migration learning apparatus based on blockchain privacy computation, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor, when executing the computer program, implements the method of:
receiving n pieces of local data in a trusted execution environment, the n pieces of local data including n-1 source domain data and one target domain data from a plurality of sources, wherein the source domain data includes: advertisement data, user data and advertisement click data of a user, wherein the target domain data is new user data;
respectively training the n-1 source domain data under the trusted execution environment to obtain n-1 local models, wherein the intermediate parameters of the local models are adjusted and monitored based on a local differential privacy algorithm;
the block chain nodes collect all local models to be aggregated to form a new model, and the new model is respectively sent to all the block chain nodes;
and downloading a new model from the block chain node, recommending advertisements for new users in the target domain by using the new model, and updating the new model based on feedback data recommended by the advertisements of the new users.
The application at least comprises the following beneficial effects:
(1) The online migration learning method and system based on the blockchain privacy calculation can solve the problem of data island under the condition of not sharing local data, fully and effectively mine data and merge across fields under the condition of protecting privacy, and promote the value circulation of the data.
(2) The online migration learning method and system based on the blockchain privacy calculation are provided, and the blockchain decentralization and strong trust attributes are utilized to conduct partner parameter audit on model parameters, so that the poison attack and the taking behavior of malicious nodes are effectively identified.
Additional advantages, objects, and features of the application will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the application.
Drawings
FIG. 1 is a schematic diagram of the online migration learning method and system based on blockchain privacy calculation.
FIG. 2 is a flowchart of an online migration learning method and system based on blockchain privacy computation.
Detailed Description
In order to clearly illustrate the present application and make the objects, technical solutions and advantages of the embodiments of the present application more clear, the technical solutions of the embodiments of the present application are clearly and completely described below in conjunction with the drawings in the embodiments of the present application, so that those skilled in the art can implement the embodiments according to the description and the text of the present application.
In the context of the present application, a trusted execution environment provides an isolated execution environment from the perspective of underlying hardware and operating systems, capable of protecting code and data running therein from external attacks, including attacks from the operating systems, hardware, and other applications. Some fields have used this technology to achieve the above-described objects, some basic principles of this technology are also known to those skilled in the art, but those skilled in the art will know how to apply this technology in this scenario after reading this application, and will clearly know that this technology combines with other features in a specific scenario to be novel.
The technology of the present application will be described in detail with reference to the following drawings.
The application relates to an online migration method based on blockchain privacy calculation, wherein the schematic steps of the application are shown in fig. 1, and the specific steps are as follows:
(1) A local data set is received in a trusted execution environment from n-1 source domain data from multiple sources and one target domain data, and fig. 2 is a flow chart of the present application. Define the local dataset as d= { D 1 ,n 2 ,…,D}。
(2) Under the trusted execution environment, training the n-1 source domain data and one target domain data respectively to obtain n local models, and protecting intermediate parameters in the local model training process by using a local differential privacy algorithm and not being capable of reversely pushing original data.
Definition f= { F 1 ,f 2 ,…,f n And the local model obtained by training on the local data set is M= { M 1 ,m 2 ,…,m n And is an intermediate parameter set in the local model training process.
In order to realize differential privacy in the local model training process, the local differential privacy algorithm adopts a Laplacian mechanism to add noise to intermediate parameters in the model updating training process, and is formed as follows:
wherein the method comprises the steps of<Lap(s/∈)>Is Laplacian noise, s is local sensitivity, and ε is the privacy budget. Is a dynamic privacy noise regulating coefficient, the noise size is controlled by cross entropy, and after normalization, the noise regulating coefficient gamma epsilon [0,1 ] is controlled]。H k (f(x i ),y i ) Is the cross entropy of model quality assessment, the smaller the value, the higher the model quality.
(3) The blockchain nodes collect all local models and aggregate the models.
Defining a blockchain node as p= { P 1 ,P 2 ,…,P n For a single blockchain node P } k Which stores data from D k A trained model.
And according to the model quality evaluation result and the credit value score of the node in the cooperation process, cooperatively adjusting the model aggregation weight, and increasing the contribution degree duty ratio of the high-quality model parameters in the aggregation model, thereby improving the accuracy of the aggregation model.
The model quality assessment is mainly realized through cross entropy in the local training process, namely, the error between the model output result and the real result is measured. Assume a tag dataset D for verifying model quality k ＝{(x 1 ,y 1 ),(x 2 ,y 2 ),…,(x n ,y n ) Cross entropy H of the kth model }, then k (f(x i ),y i ) Can be calculated by the following formula:
wherein y is i Is the desired output, f (x i ) Representing the predicted outcome of the model.
Model quality assessment mainly depends on blockchain nodes P k Model loss determination for local training at the ith round, using cross entropy H (f (x i ),y i ) To evaluate local model loss. Quality assessment weight Q of adaptive model aggregation algorithm i (P k ) The definition is as follows:
in order to avoid the false model quality assessment result provided by the participating nodes, the application adopts collaborator asynchronous parameter examinationThe calculation mechanism is used for parameter verification, and in each iteration process, training collaborators calculate model errors on model parameters to be verified by using own local data, and the block chain node P k Broadcast model parameter m i And cross entropy H (f (x) i ),y i ) The collaborator node verifies the model parameters by using the local tag data set, so that the model quality assessment not only depends on the training error of the model itself, but also needs to consider the verification results of other nodes on the model parameters, thereby obtaining a more real and fair local model quality assessment result.
Quality assessment weights for each node are recorded on the blockchain for each model aggregate, blockchain node P k Reputation expressions in an aggregation process are represented by a cumulative reputation score S (P k ) Expressed by S (P k ) Formalizing as follows:
where τ is the blockchain node P k Number of times of participation in model aggregation, Q i (P k ) Is a blockchain node P k Quality assessment weights recorded on blockchains in participating in ith model aggregation, will be S (P k ) As blockchain node P k And the historical accumulated reputation value in the federal transfer learning is used as the basis for the federal transfer learning model aggregation and the participation node selection.
The global model update calculation process in the adaptive model aggregation algorithm is expressed as follows:
wherein n is the number of blockchain nodes involved in model aggregation, Q i (P k ) Is a blockchain node P k Quality assessment weights of local models in the ith model aggregation, S (P k ) Is a blockchain node P k Is included.
(4) And performing a downstream task on a new sample in the target field by using the aggregated model, and updating the aggregated model based on feedback data of the downstream task of the new sample.
In summary, according to the online migration learning method and system based on blockchain privacy calculation provided by the application, local data is trained to obtain a local model under a trusted execution environment, meanwhile, intermediate parameters in the model training process are protected by utilizing a local differential privacy algorithm, the local model is collected by utilizing blockchain nodes, the model is aggregated by utilizing a self-adaptive model aggregation algorithm, a new sample is learned by utilizing the aggregation model, and the aggregation model is updated based on feedback data learned by the new sample. The application can use the existing knowledge in the source field to realize task solution in the target field, protect the privacy of local data, effectively combine multiparty data, break the island barriers of the data and help realize the value circulation of the data.
Example 2:
an online migration learning system based on blockchain privacy calculation comprises a data collection module, a first processing module, a second processing module and a third processing module:
the data collection module is configured to receive n pieces of local data in a trusted execution environment, where the n pieces of local data include n-1 pieces of source domain data from a plurality of sources and one piece of target domain data, and the source domain data includes: advertisement data, user data and advertisement click data of a user, wherein the target domain data is new user data;
the first processing module respectively trains the n-1 source domain data under the trusted execution environment to obtain n-1 local models, wherein the intermediate parameters of the local models are adjusted and monitored based on a local differential privacy algorithm;
the second processing module collects all local models by the block chain link points to be aggregated to form a new model, and the new model is respectively sent to all block chain nodes;
and the third processing module is used for downloading a new model from the blockchain node, recommending advertisements for new users in the target domain by using the new model, and updating the new model based on feedback data recommended by the advertisements of the new users.
All changes and modifications that come within the spirit and scope of the application are desired to be protected and all equivalent thereto are deemed to be within the scope of the application.
In this specification, each embodiment is described in a progressive manner, and each embodiment is mainly described in a different manner from other embodiments, so that identical and similar parts of each embodiment are mutually referred to.
It will be apparent to those skilled in the art that embodiments of the present application may be provided as a method, apparatus, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, terminal devices (systems), and computer program products according to the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It should be noted that:
reference in the specification to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment is included in at least one embodiment of the application. Thus, the appearances of the phrase "one embodiment" or "an embodiment" in various places throughout this specification are not necessarily all referring to the same embodiment.
The previous description of the embodiments is provided to facilitate a person of ordinary skill in the art in order to make and use the present application. It will be apparent to those having ordinary skill in the art that various modifications to the above-described embodiments may be readily made and the generic principles described herein may be applied to other embodiments without the use of inventive faculty. Therefore, the present application is not limited to the above-described embodiments, and those skilled in the art, based on the present disclosure, should make improvements and modifications within the scope of the present application.
Claims (9)
1. The online migration learning method based on the blockchain privacy calculation is characterized by comprising the following steps of:
receiving n pieces of local data in a trusted execution environment, the n pieces of local data including n-1 source domain data and one target domain data from a plurality of sources, wherein the source domain data includes: advertisement data, user data and advertisement click data of a user, wherein the target domain data is new user data;
respectively training the n-1 source domain data under the trusted execution environment to obtain n-1 local models, wherein the intermediate parameters of the local models are adjusted and monitored based on a local differential privacy algorithm;
the block chain nodes collect all local models to be aggregated to form a new model, and the new model is respectively sent to all the block chain nodes;
and downloading a new model from the block chain node, recommending advertisements for new users in the target domain by using the new model, and updating the new model based on feedback data recommended by the advertisements of the new users.
2. The blockchain privacy calculation-based online migration learning method of claim 1, wherein the local data set is used for an e-commerce advertisement recommendation task, and comprises an advertisement click behavior log of a user, characteristics of the advertisement, characteristics of the user and shopping behaviors of the user;
the user's advertisement click log contains the fields: user ID, timestamp, advertisement ID, resource display position and whether click, wherein, whether click is represented by 0/1, 0 represents no click, 1 represents click;
the characteristics of the advertisement itself include: advertisement ID, commodity category ID, brand ID, commodity price;
the characteristics of the user themselves include: user ID, gender, age level, consumption level, shopping depth, city level;
the shopping behavior of the user includes: user ID, timestamp, behavior type, merchandise category ID, and brand ID, wherein the behavior type includes: browse, join shopping carts, like and purchase.
3. The online migration learning method based on blockchain privacy calculation according to claim 1, wherein the adjusting and monitoring of the intermediate parameters of the local model based on the local differential privacy algorithm comprises the following steps:
adding noise to the intermediate parameters in the local model updating process by adopting a Laplace mechanism, and if the calculation tasks are the same, the calculation result after noise addition is similar to the calculation result of the original data;
and updating the intermediate parameters after noise addition after each iteration until the sum of privacy budgets accumulated by the iterations exceeds a preset budget constraint, and ending the iterations.
4. The online migration learning method based on blockchain privacy calculation of claim 1, wherein the step of collecting all local models by the blockchain node to perform model aggregation, iterating a plurality of times, and sending a new aggregated model to all blockchain link points comprises:
the block chain nodes collect local models participating in training;
performing aggregation treatment on all the participated local models by adopting a self-adaptive model aggregation algorithm to obtain a new model;
and iterating the new model until the global loss function converges or reaches the preset precision, and respectively transmitting the aggregated new model to all the participated blockchain nodes.
5. The online migration learning method based on blockchain privacy calculation of claim 4, wherein the aggregation processing of all participating local models by adopting the adaptive model aggregation algorithm comprises the following steps:
calculating a quality assessment score for each blockchain node based on the cross entropy loss function;
calculating the credit value score of each blockchain node in the cooperation process by adopting a cooperator asynchronous parameter audit mechanism;
and adjusting model aggregation weights through the quality evaluation scores of the blockchain nodes and the reputation scores in the cooperation process, and carrying out aggregation processing on all the participating local models based on the model aggregation weights.
6. The online migration learning method based on blockchain privacy computation of claim 5, wherein the collaborator asynchronous parameter auditing mechanism comprises the following steps:
encrypting the model quality audit data of the verified blockchain node and then transmitting the encrypted model quality audit data to other blockchain nodes for verification;
other block chain nodes verify parameters of the verified model based on own local data and broadcast verification results to other block chain nodes;
and after the verification of all the block chain nodes is passed, obtaining a final verification result.
7. The online migration learning system based on the blockchain privacy calculation is characterized by comprising a data collection module, a first processing module, a second processing module and a third processing module:
the data collection module is configured to receive n pieces of local data in a trusted execution environment, where the n pieces of local data include n-1 pieces of source domain data from a plurality of sources and one piece of target domain data, and the source domain data includes: advertisement data, user data and advertisement click data of a user, wherein the target domain data is new user data;
the first processing module respectively trains the n-1 source domain data under the trusted execution environment to obtain n-1 local models, wherein the intermediate parameters of the local models are adjusted and monitored based on a local differential privacy algorithm;
the second processing module collects all local models by the block chain link points to be aggregated to form a new model, and the new model is respectively sent to all block chain nodes;
and the third processing module is used for downloading a new model from the blockchain node, recommending advertisements for new users in the target domain by using the new model, and updating the new model based on feedback data recommended by the advertisements of the new users.
8. A computer readable storage medium storing a computer program, which when executed by a processor implements the method of any one of claims 1 to 6.
9. An online migration learning apparatus based on blockchain privacy computation, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the method of any one of claims 1 to 6 when executing the computer program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311186226.0A CN117094773A (en) | 2023-09-13 | 2023-09-13 | Online migration learning method and system based on blockchain privacy calculation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311186226.0A CN117094773A (en) | 2023-09-13 | 2023-09-13 | Online migration learning method and system based on blockchain privacy calculation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117094773A true CN117094773A (en) | 2023-11-21 |
Family
ID=88776936
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311186226.0A Pending CN117094773A (en) | 2023-09-13 | 2023-09-13 | Online migration learning method and system based on blockchain privacy calculation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117094773A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117709444A (en) * | 2024-02-06 | 2024-03-15 | 盛业信息科技服务(深圳)有限公司 | Differential privacy model updating method and system based on decentralised federal learning |
-
2023
- 2023-09-13 CN CN202311186226.0A patent/CN117094773A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117709444A (en) * | 2024-02-06 | 2024-03-15 | 盛业信息科技服务(深圳)有限公司 | Differential privacy model updating method and system based on decentralised federal learning |
| CN117709444B (en) * | 2024-02-06 | 2024-05-28 | 盛业信息科技服务(深圳)有限公司 | Differential privacy model updating method and system based on decentralised federal learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Wang et al. | Blockchain‐enabled data sharing in supply chains: Model, operationalization, and tutorial | |
| Himeur et al. | Blockchain-based recommender systems: Applications, challenges and future opportunities | |
| CN110189192B (en) | Information recommendation model generation method and device | |
| CN110990871B (en) | Machine learning model training method, prediction method and device based on artificial intelligence | |
| Yli-Huumo et al. | Where is current research on blockchain technology?—a systematic review | |
| WO2022206510A1 (en) | Model training method and apparatus for federated learning, and device and storage medium | |
| Wang et al. | Evidence-based trust: A mathematical model geared for multiagent systems | |
| CN112465627B (en) | Financial loan auditing method and system based on block chain and machine learning | |
| US11177937B1 (en) | Apparatus and method for establishing trust of anonymous identities | |
| Azad et al. | PrivBox: Verifiable decentralized reputation system for online marketplaces | |
| CN113748440A (en) | Secure multi-party computing attribution | |
| CN112039702B (en) | Model parameter training method and device based on federal learning and mutual learning | |
| Tavakolifard et al. | A taxonomy to express open challenges in trust and reputation systems. | |
| Majd et al. | A trust model for recommender agent systems | |
| Bellagarda et al. | An updated survey on the convergence of distributed ledger technology and artificial intelligence: Current state, major challenges and future direction | |
| CN117094773A (en) | Online migration learning method and system based on blockchain privacy calculation | |
| Zheng et al. | A matrix factorization recommendation system-based local differential privacy for protecting users’ sensitive data | |
| Putra et al. | Blockchain for trust and reputation management in cyber-physical systems | |
| Rafi et al. | Fairness and privacy preserving in federated learning: A survey | |
| Yan et al. | User‐centric trust and reputation model for personal and trusted service selection | |
| He et al. | Federated learning intellectual capital platform | |
| Chatterjee et al. | Next generation financial services: Role of blockchain enabled federated learning and metaverse | |
| Vahabli et al. | A novel trust-based access control for social networks using fuzzy systems | |
| CN117235692A (en) | Web 3-based software supply chain risk management system, method and platform | |
| US11652803B2 (en) | System, method and computer-accessible medium for supporting at least one cyber-physical signaling game |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |