CN113347162A - Block link point contribution degree proving consensus method for crowd-sourcing service - Google Patents
Block link point contribution degree proving consensus method for crowd-sourcing service Download PDFInfo
- Publication number
- CN113347162A CN113347162A CN202110547351.4A CN202110547351A CN113347162A CN 113347162 A CN113347162 A CN 113347162A CN 202110547351 A CN202110547351 A CN 202110547351A CN 113347162 A CN113347162 A CN 113347162A
- Authority
- CN
- China
- Prior art keywords
- model
- node
- contribution degree
- training
- data
- 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.)
- Granted
Links
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/04—Network architectures or network communication protocols for network security for providing a confidential data exchange among entities communicating through data packet networks
- H04L63/0428—Network architectures or network communication protocols for network security for providing a confidential data exchange among entities communicating through data packet networks wherein the data content is protected, e.g. by encrypting or encapsulating the payload
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/12—Applying verification of the received information
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L67/00—Network arrangements or protocols for supporting network services or applications
- H04L67/01—Protocols
- H04L67/10—Protocols in which an application is distributed across nodes in the network
- H04L67/1095—Replication or mirroring of data, e.g. scheduling or transport for data synchronisation between network nodes
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L2209/00—Additional information or applications relating to cryptographic mechanisms or cryptographic arrangements for secret or secure communication H04L9/00
- H04L2209/30—Compression, e.g. Merkle-Damgard construction
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L2209/00—Additional information or applications relating to cryptographic mechanisms or cryptographic arrangements for secret or secure communication H04L9/00
- H04L2209/46—Secure multiparty computation, e.g. millionaire problem
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L9/00—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols
- H04L9/50—Cryptographic mechanisms or cryptographic arrangements for secret or secure communications; Network security protocols using hash chains, e.g. blockchains or hash trees
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Abstract
The invention discloses a block chain link point contribution degree proving and identifying method for crowd-sourcing service, which comprises the following steps: competing the accounting right of the block chain distributed account book according to three aspects of the node online time, the local model quality and the data contribution degree; based on a workload proving mechanism, the mine excavation difficulty coefficient is dynamically adjusted, the greater the contribution degree is, the lower the mine excavation difficulty of the node is, and the node participation fairness is improved; and the participating nodes compete for the accounting right through a contribution degree certification consensus algorithm, so that platform rewards are obtained. The reward points can be used for downloading the shared parameters recorded on the block chain and used for improving the quality of the local model; in order to avoid the counterfeiting of the local model quality parameters, the swarm intelligence service participation node local model parameters are automatically verified by triggering an intelligent contract.
Description
Technical Field
The invention relates to the technical field of block chains, in particular to a block chain link point contribution degree proving consensus method for crowd sourcing service.
Background
The crowd-sourcing service is provided by the cooperation of intelligent edge nodes in a group, and the federal learning is one of typical application frameworks. The federated learning is a collaborative machine learning framework, and nodes participating in collaboration train local models by using local data, and model aggregation is performed through a parameter server, so that the prediction effect of multi-source data is realized. In the federal learning model aggregation process, multi-party consensus verification of block chain nodes is required, but the existing consensus algorithm is not completely suitable for the scenario. The Proof of workload (Proof of Work, PoW) not only consumes a lot of node computing power, but also is not beneficial to the participation of lightweight edge nodes. Off-line nodes in Proof of user rights (PoS) may also accumulate coin age, possibly leading to a behavior of taking a note of a car by a participating node.
Therefore, how to provide a block link point contribution degree proving consensus method for crowd sourcing service, which can effectively save the computation overhead and improve the fairness, is a problem to be solved by those skilled in the art.
Disclosure of Invention
In view of the above, the invention provides a block link point contribution degree proving consensus method for a crowd sourcing service, which solves the problems of resource overhead and unfairness caused in the process of crowd sourcing service model consensus verification in the prior art.
In order to achieve the purpose, the invention adopts the following technical scheme:
a block link point contribution degree certification consensus method for crowd-sourcing service comprises the following steps:
s1, acquiring a timestamp of the last block of the existing block chain, a network timestamp and an off-line timestamp of a node which is added into the block chain for the first time, and calculating the on-line contribution of the node;
s2, training a local model, calculating the cross entropy of the quality evaluation result of the local model, determining the contribution degree of the local model, and broadcasting gradient data for sharing;
s3, calculating a data information entropy, counting sample data size, and determining a data contribution degree through the data information entropy and the data size;
s4, generating an antagonistic network DPGAN through differential privacy to provide an antagonistic sample data set for verifying the model quality for the crowd sourcing service parameter verification node, triggering a model parameter quality verification intelligent contract, verifying shared gradient data and a local model quality evaluation result through the antagonistic sample data set, executing a step S5 if the verification condition is met, and otherwise, discarding the current gradient data;
s5, the participating nodes are trained by using local data, and a joint training model is obtained by combining gradient data obtained by sharing, so that a crowd-sourcing service cooperation process is completed; and adding the node on-line contribution degree, the local model contribution degree and the data contribution degree to obtain an evaluation node contribution degree, wherein the node contribution degree evaluates the contribution degree of the participating nodes to the joint training model, distributes a mining difficulty degree coefficient in inverse proportion to the node contribution degree for the participating nodes, dynamically adjusts the mining difficulty, and achieves the financial book consensus through mining.
It should be noted that:
the node down timestamp refers to a timestamp of a node in the blockchain network exiting the network.
The method judges the contribution degree of the node online resource overhead through the node online time; judging the quality contribution degree of the local model of the node according to the cross entropy data of the model quality evaluation result; finally, calculating the ratio of the entropy of the local data information of the node to the data amount to judge the contribution degree of the node data; judging the contribution degree of the swarm intelligence service participation node to the whole collaborative swarm intelligence service model based on the three dimensions; on the other hand, the local model and the model quality verification result need to be verified by a consensus mechanism and then recorded on the blockchain for data sharing. Firstly, local model parameters and model quality verification results are packaged and linked in a transaction mode, and the mining difficulty coefficient is dynamically adjusted based on the node contribution degree, so that excessive calculation overhead is avoided, and the fairness of the consensus process is ensured. In addition, when the blockchain accounting request Req is received, a master node is selected from the transaction blockchain according to the contribution value of the nodes in the blockchain network to be responsible for receiving the transaction request. And in a period of time, selecting the node with the largest contribution degree as a main node, and responding to the request and promoting the overall consensus process.
The crowd-sourcing-based service participants include: a crowd sourcing service task publisher and an intelligent edge device that executes a local stochastic gradient descent algorithm. The task publisher puts forward a model training requirement, is responsible for constructing a joint training group by the participatory nodes capable of carrying out joint training, and can carry out credit evaluation on the participatory nodes so as to avoid the behavior of carrying out disgust and taking a vehicle by the participatory nodes. After the initial training model is received, the local stochastic gradient descent algorithm continuously optimizes and iterates the model by using local data, the trained intermediate gradient parameters are aggregated with local model parameters shared on the block chain after being broadcasted, and each node continuously optimizes the aggregated model to achieve a model convergence condition.
Preferably, the specific contents of S1 include:
obtaining the offline time interval of the node by subtracting the network time stamp of the node which is added into the block chain for the first time from the offline time stamp, obtaining the online time period of the node by subtracting the network time stamp of the node which is added into the block chain for the first time from the time stamp of the last block, and obtaining the online time contribution by subtracting the offline time interval from the online time period of the node;
and controlling the proportion of the online time contribution degree through an online time contribution degree adjusting coefficient preset in advance.
It should be noted that:
the method is controlled by an online time contribution degree adjusting coefficient (more than 0 and less than 1) preset by a user, and the adjusting coefficient is multiplied by the online time contribution degree to obtain a node online contribution degree value.
Preferably, the content in S2 specifically includes:
the process of training the local model comprises the following steps: a crowd sourcing service task publisher distributes an initial model to each participating node;
each participating node executes a multi-round random gradient descent algorithm by using local data based on the initial model, optimizes the local model, and broadcasts and chains the training gradient and the training loss after iterating and fixing the round number;
testing the training effect of the local model through the label data, wherein the cross entropy is a loss function value for measuring the quality of the model, and the approximation degree of the prediction result and the real result of the model is represented by calculating the deviation between the expected output of the label data and the prediction result;
the smaller the cross entropy, the closer the probability distribution representing the model prediction is to the true result, i.e. the higher the model quality.
It should be noted that:
the cross entropy is a loss value of the trained local model after the label data set is tested, and is used for measuring the quality of model training.
The label data is a data set for testing the training effect of the model and consists of a test set and a verification set, wherein the test set is used for model training, and the verification set is used for testing the accuracy of the model training by marking labels in advance.
Preferably, the data information entropy in S3 represents the data value, and the data size is the proportion of the local training sample data size of the node in the whole data sample.
Preferably, the specific contents of the intelligent contract verification for verifying the quality of the model parameters in S4 include:
after carrying out differential privacy on the basis of local training samples, the participating nodes adopt DPGAN to generate a countermeasure sample data set and broadcast the data set;
other nodes receive the broadcasted gradient data and the training loss, the training loss is a local model quality evaluation result, namely cross entropy, a model parameter quality verification intelligent contract is automatically triggered, and the shared gradient data and the local model quality evaluation result are verified through an anti-sample data set;
if the difference is smaller than the preset range, the verification is passed, otherwise, the verification fails, and the received gradient data is discarded after the verification fails so as to avoid influencing the accuracy of the aggregation model.
It should be noted that:
the difference between the training loss of the verification node and the training loss of the verified node is used for evaluating the accuracy of parameters submitted by the verified node, so that the risk that the node takes a free car to provide a false loss value is reduced.
Preferably, the account book consensus content of S5 includes: after the consensus process starts, selecting the node with the highest contribution degree in the joint training model as a main node to accept a transaction request; each node carries out local training, broadcasts local training loss to other nodes in the joint training model, and packs and sends the local training loss in a transaction form; the gradient data of the local model of the node and the quality evaluation result of the current wheel model are cross entropy packaged into blocks, and the blocks are recorded on a block chain after competition; in order to ensure the fairness of bookkeeping competition, the ore excavation difficulty is dynamically adjusted based on a workload proving mechanism, the node with high contribution degree to the node is reduced, and the high-quality model can be quickly linked.
It should be noted that:
the gradient data of the node local model and the quality evaluation result of the current round model are cross entropy packaged into blocks, and the specific content recorded on the block chain after competition comprises the following contents:
after the node trains the local model, in order to avoid the privacy disclosure of original data, only intermediate gradient parameters of model training and the model training quality evaluation results (cross entropy) of the round are packed into blocks, miners in the block chain network dig mines through a contribution degree certification consensus algorithm and compete for the bookkeeping right, and miners who dig successfully add block information to the tail of the current block chain to finish bookkeeping.
A block chain link point contribution degree certification consensus device for crowd-sourcing service comprises a block chain miner module, a model training module, a model aggregation module and a model verification module;
the block chain miner module is used for realizing data transaction;
the model training module is used for realizing the training process of the model;
the model aggregation module is used for realizing the aggregation of the joint training model;
the model verification module is used for verifying the trained model.
A computer device, the device comprising: a memory and one or more processors;
a memory for storing one or more programs;
the one or more programs, when executed by the processor, cause the processor to perform the above-described method.
A computer-readable storage medium storing computer-executable instructions that, when executed by a processor, implement the above-described method.
According to the technical scheme, compared with the prior art, the block link point contribution degree proving consensus method for the crowd sourcing service can quantitatively evaluate the contribution degree in the collaborative training process based on the participating nodes, dynamically adjust the mining difficulty coefficient in the consensus process based on the contribution degree of the participating nodes, and improve the fairness of the crowd sourcing service process and the consensus mechanism. Because the privacy disclosure problem exists due to the fact that the whole amount of high-dimensional gradient data can be subjected to reasoning attack, the method and the system consider that the gradient is compressed and then broadcast, and therefore privacy disclosure of local data of the participating nodes caused by complete gradient data is avoided. In order to ensure the verifiability of the quality of the local model parameters, an intelligent contract is adopted to carry out the automatic verification operation of the local model parameters. By means of a generated countermeasure network (DPGAN) based on differential privacy protection, original data are trained after differential privacy and noise are added to generate and broadcast countermeasure sample data which can be disclosed, and therefore privacy of the original sample data is prevented from being revealed. The intelligent contract evaluates the quality of the local model by verifying the confrontation sample and the model parameters, and abandons the model parameters with low quality.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.
Fig. 1 is a schematic diagram illustrating a block link point contribution degree certification consensus method for crowd sourcing service according to the present invention;
FIG. 2 is a schematic overall flowchart of a block link point contribution degree identification method for crowd sourcing service according to the present invention;
fig. 3 is a schematic diagram of an intermediate gradient parameter compression and sharing process of the block link point contribution degree certification consensus method for crowd sourcing service according to the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The embodiment of the invention discloses a block link point contribution degree proving consensus method for crowd-sourcing service, which is characterized in that as shown in figure 1, the contribution degree of a crowd-sourcing service participating node is calculated, and the mining difficulty degree in the consensus process is dynamically adjusted based on the node contribution degree to achieve local model parameter distributed accounting consensus. As shown in fig. 2, the process of block chain-based crowd-sourcing service participant contribution degree proof consensus is as follows:
1) acquiring timestamps of a first block and a last block and off-line time of a node, and calculating the on-line contribution degree of the node;
2) the participating nodes provide local data information entropy and data quantity results (taking MB as a unit);
3) a node local training model broadcasts high-dimensional gradient data subjected to gradient compression and a local model quality evaluation result, and step 4) is executed in parallel;
4) triggering a model parameter quality verification intelligent contract, verifying input gradient data and a model quality evaluation result through a countermeasure sample generated by the DPGAN, and executing 5) if the verification condition is met, or executing 6) if the verification condition is not met;
5) summing the contribution degrees of the nodes calculated in the steps 1)2)3), distributing an excavation difficulty coefficient which is inversely proportional to the contribution degrees for the nodes, starting excavation until the excavation is successful, and recording uplink by local model parameters;
6) and punishing the nodes which do not pass the local model parameter verification, discarding the local model parameters provided by the nodes, and not performing uplink operation.
The local model parameter intelligent contract verification process comprises the following steps: the participating nodes compress the intermediate gradient data of the local model training and broadcast the cross entropy loss function of the local training, wherein the intermediate gradient parameter compression and sharing flow diagram is shown in fig. 3. And after carrying out differential privacy on the basis of the local training samples, the participating nodes adopt the DPGAN to generate countermeasure samples and broadcast the countermeasure samples. And the node receiving the broadcast parameters verifies the cross entropy loss function through the local participation and countermeasure sample, if the difference value is smaller than a certain range, the verification is passed, uplink consensus can be carried out, and otherwise, the parameters are discarded so as not to influence the accuracy of the aggregation model.
By using the scheme, the consensus is proved through the contribution degree of the participants, so that the resource overhead of the node consensus process is saved, and the fairness of node participation is improved. The block chain technology is adopted to realize parameter sharing of the crowd sourcing service local model, and the influence of the low-quality model on the accuracy rate of the aggregation model is reduced. Through gradient compression, the risk of privacy disclosure of intermediate gradient data is reduced, the verification of local model parameters is automated through executing an intelligent contract, the cost of parameter management maintenance and management is saved, and the efficiency is improved. The verification data set generates countermeasure samples through a differential privacy generation countermeasure network (DPGAN), and the quality of the local model parameters is verified through the countermeasure samples, so that not only are the local training parameters not exposed, but also the data set of verifiable model parameters can be provided. By the scheme, the fairness of the crowd sourcing service participating in the node cooperative training can be realized, and the calculation and storage burden of the node consensus process is reduced. The scheme accurately solves the problem of node fairness and intermediate parameter privacy disclosure of parameter consensus in the crowd-sourcing service.
The crowd sourcing service is provided by multiple types of intelligent edge devices in cooperation. With the rise of mobile communication technology and intelligent edge equipment, the crowd-sourcing service will have wide application prospects in the fields of smart cities, electronic medical treatment, wireless communication, mobile edge networks and the like. However, the cooperative training of multiple nodes is involved in the crowd sourcing service, and how to ensure the fairness of the participation of the multiple nodes is a key problem for promoting the crowd sourcing service to fall to the ground further. Particularly, in the cooperative training process, the quality of an aggregation model caused by the participation of some malicious nodes and the behavior of taking a free car by the nodes is not high, and the further development of the crowd-sourcing service is influenced. Furthermore, privacy of the local training data may be compromised due to the high-dimensional gradient data, thereby masking the advantages specific to crowd-sourcing services. In order to solve the problems, the block chain is introduced into a multi-node collaborative training scene of the crowd-sourcing service, the contribution degree of the participating nodes is scientifically quantized, the ore excavation difficulty in the block chain consensus process is dynamically adjusted according to the contribution degree, the workload proving consensus mechanism is improved, the resource waste of the nodes is reduced, and the fairness of the participating nodes is improved.
In order to avoid malicious nodes from providing false model parameters, the present disclosure provides countersample verification local model parameters using a Differential Privacy protection-based generation countermeasure Network (DPGAN). The DPGAN is a deep learning network model fusing differential privacy and generation countermeasure network (GAN), and provides a generation countermeasure network model for protecting intermediate gradient privacy. Due to the high complexity of the deep learning network model, the complete intermediate gradient data can easily expose the training samples, so that the gradient noise addition is a common gradient privacy protection method at present. The DPGAN not only protects the privacy of the data samples by using differential privacy, but also obtains countermeasure samples based on the noisy data sample training, and the countermeasure samples do not expose the local data privacy of the nodes and can be used for model quality verification. The quality of the model parameters is automatically verified through the intelligent contract, so that false local model parameters are avoided, and the reliability of the aggregation model is improved. The block chain-based crowd sourcing service participant contribution degree proving method realizes distributed autonomous training of the crowd sourcing service participant nodes, reduces resource waste in the node consensus process through an improved participant contribution degree proving algorithm, and improves node participation fairness. The compressed intermediate gradient parameters and the model quality verification result are stored through the block chain, so that the transparency of shared data is realized, and the characteristics of non-falsification and traceability of the parameters are also realized.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. The device disclosed by the embodiment corresponds to the method disclosed by the embodiment, so that the description is simple, and the relevant points can be referred to the method part for description.
The previous description of the disclosed embodiments is provided to enable any person skilled in the art to make or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the generic principles defined herein may be applied to other embodiments without departing from the spirit or scope of the invention. Thus, the present invention is not intended to be limited to the embodiments shown herein but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.
Claims (9)
1. A block link point contribution degree certification consensus method for crowd-sourcing service is characterized by comprising the following steps:
s1, acquiring a timestamp of the last block of the existing block chain, a network timestamp and an off-line timestamp of a node which is added into the block chain for the first time, and calculating the on-line contribution of the node;
s2, training a local model, calculating the cross entropy of the quality evaluation result of the local model, determining the contribution degree of the local model, and broadcasting gradient data for sharing;
s3, calculating a data information entropy, counting sample data size, and determining a data contribution degree through the data information entropy and the data size;
s4, generating an antagonistic network DPGAN through differential privacy to provide an antagonistic sample data set for verifying the model quality for the crowd sourcing service parameter verification node, triggering a model parameter quality verification intelligent contract, verifying shared gradient data and a local model quality evaluation result through the antagonistic sample data set, executing a step S5 if the verification condition is met, and otherwise, discarding the current gradient data;
s5, the participating nodes are trained by using local data, and a joint training model is obtained by combining gradient data obtained by sharing, so that a crowd-sourcing service cooperation process is completed; and adding the node on-line contribution degree, the local model contribution degree and the data contribution degree to obtain an evaluation node contribution degree, wherein the node contribution degree evaluates the contribution degree of the participating nodes to the joint training model, distributes a mining difficulty degree coefficient in inverse proportion to the node contribution degree for the participating nodes, dynamically adjusts the mining difficulty, and achieves the financial book consensus through mining.
2. The method of claim 1, wherein the details of S1 include:
obtaining the offline time interval of the node by subtracting the network time stamp of the node which is added into the block chain for the first time from the offline time stamp, obtaining the online time period of the node by subtracting the network time stamp of the node which is added into the block chain for the first time from the time stamp of the last block, and obtaining the online time contribution by subtracting the offline time interval from the online time period of the node;
and controlling the proportion of the online time contribution degree through an online time contribution degree adjusting coefficient preset in advance.
3. The method as claimed in claim 1, wherein the step S2 includes:
the process of training the local model comprises the following steps: a crowd sourcing service task publisher distributes an initial model to each participating node;
each participating node executes a multi-round random gradient descent algorithm by using local data based on the initial model, optimizes the local model, and broadcasts and chains the training gradient and the training loss after iterating and fixing the round number;
testing the training effect of the local model through the label data, wherein the cross entropy is a loss function value for measuring the quality of the model, and the approximation degree of the prediction result and the real result of the model is represented by calculating the deviation between the expected output of the label data and the prediction result;
the smaller the cross entropy, the closer the probability distribution representing the model prediction is to the true result, i.e. the higher the model quality.
4. The method of claim 1, wherein the entropy of the data information in S3 represents a data value, and the data size is a ratio of local training sample data size of the node to the whole data sample.
5. The method for consensus certification of block link point contribution degree for crowd-sourcing service according to claim 1, wherein the specific content of the model parameter quality validation intelligent contract validation in S4 comprises:
after carrying out differential privacy on the basis of local training samples, the participating nodes adopt DPGAN to generate a countermeasure sample data set and broadcast the data set;
other nodes receive the broadcasted gradient data and the training loss, the training loss is a local model quality evaluation result, namely cross entropy, a model parameter quality verification intelligent contract is automatically triggered, and the shared gradient data and the local model quality evaluation result are verified through an anti-sample data set;
if the difference between the training loss of the verification node and the training loss of the verified node is smaller than the preset range, the verification is passed, otherwise, the verification fails, and the received gradient data is discarded after the verification fails so as to avoid influencing the accuracy of the aggregation model.
6. The method as claimed in claim 1, wherein the account book consensus of S5 includes: after the consensus process starts, selecting the node with the highest contribution degree in the joint training model as a main node to accept a transaction request; each node carries out local training, broadcasts local training loss to other nodes in the joint training model, and packs and sends the local training loss in a transaction form; the gradient data of the local model of the node and the quality evaluation result of the current wheel model are cross entropy packaged into blocks, and the blocks are recorded on a block chain after competition; in order to ensure the fairness of bookkeeping competition, the ore excavation difficulty is dynamically adjusted based on a workload proving mechanism, the node with high contribution degree to the node is reduced, and the high-quality model can be quickly linked.
7. A block chain link point contribution degree certification consensus device for crowd-sourcing service comprises a block chain miner module, a model training module, a model aggregation module and a model verification module;
the block chain miner module is used for realizing data transaction;
the model training module is used for realizing the training process of the model;
the model aggregation module is used for realizing the aggregation of the joint training model;
the model verification module is used for verifying the trained model.
8. A computer device, the device comprising: a memory and one or more processors;
a memory for storing one or more programs;
the program or programs, when executed by a processor, cause the processor to perform the method of any of claims 1-6.
9. A computer-readable storage medium having stored thereon computer-executable instructions that, when executed by a processor, implement the method of any one of claims 1-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110547351.4A CN113347162B (en) | 2021-05-19 | 2021-05-19 | Block chain link point contribution degree proving consensus method for crowd-sourcing service |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110547351.4A CN113347162B (en) | 2021-05-19 | 2021-05-19 | Block chain link point contribution degree proving consensus method for crowd-sourcing service |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113347162A true CN113347162A (en) | 2021-09-03 |
| CN113347162B CN113347162B (en) | 2023-09-29 |
Family
ID=77469740
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110547351.4A Active CN113347162B (en) | 2021-05-19 | 2021-05-19 | Block chain link point contribution degree proving consensus method for crowd-sourcing service |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113347162B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113762528A (en) * | 2021-09-10 | 2021-12-07 | 北京航空航天大学 | Block chain-based federal credit assessment method |
| CN113961969A (en) * | 2021-12-22 | 2022-01-21 | 北京金睛云华科技有限公司 | Security threat collaborative modeling method and system |
| CN114757674A (en) * | 2022-06-15 | 2022-07-15 | 中科链安(北京)科技有限公司 | Consensus method, system and storage medium for rewarding and punishing ability based on intelligent contract |
| CN115297009A (en) * | 2022-07-08 | 2022-11-04 | 中电信数智科技有限公司 | Block chain-based distributed network digital encryption consistency optimization method |
| WO2024041130A1 (en) * | 2022-08-25 | 2024-02-29 | 华为技术有限公司 | Rights and interests allocation method and apparatus |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110825810A (en) * | 2019-10-28 | 2020-02-21 | 天津理工大学 | Block chain-based crowd sensing double privacy protection method |
| WO2020165256A1 (en) * | 2019-02-13 | 2020-08-20 | Uvue Limited | System and method for evaluating useful work |
| CN111931242A (en) * | 2020-09-30 | 2020-11-13 | 国网浙江省电力有限公司电力科学研究院 | Data sharing method, computer equipment applying same and readable storage medium |
| CN112784994A (en) * | 2020-12-31 | 2021-05-11 | 浙江大学 | Block chain-based federated learning data participant contribution value calculation and excitation method |
-
2021
- 2021-05-19 CN CN202110547351.4A patent/CN113347162B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020165256A1 (en) * | 2019-02-13 | 2020-08-20 | Uvue Limited | System and method for evaluating useful work |
| CN110825810A (en) * | 2019-10-28 | 2020-02-21 | 天津理工大学 | Block chain-based crowd sensing double privacy protection method |
| CN111931242A (en) * | 2020-09-30 | 2020-11-13 | 国网浙江省电力有限公司电力科学研究院 | Data sharing method, computer equipment applying same and readable storage medium |
| CN112784994A (en) * | 2020-12-31 | 2021-05-11 | 浙江大学 | Block chain-based federated learning data participant contribution value calculation and excitation method |
Non-Patent Citations (1)
| Title |
|---|
| 吴梦宇 等: "基于工作量证明和权益证明改进的区块链共识机制", 计算机应用, vol. 40, no. 08 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113762528A (en) * | 2021-09-10 | 2021-12-07 | 北京航空航天大学 | Block chain-based federal credit assessment method |
| CN113961969A (en) * | 2021-12-22 | 2022-01-21 | 北京金睛云华科技有限公司 | Security threat collaborative modeling method and system |
| CN114757674A (en) * | 2022-06-15 | 2022-07-15 | 中科链安(北京)科技有限公司 | Consensus method, system and storage medium for rewarding and punishing ability based on intelligent contract |
| CN114757674B (en) * | 2022-06-15 | 2022-09-23 | 中科链安(北京)科技有限公司 | Consensus method, system and storage medium for rewarding and punishing ability based on intelligent contract |
| CN115297009A (en) * | 2022-07-08 | 2022-11-04 | 中电信数智科技有限公司 | Block chain-based distributed network digital encryption consistency optimization method |
| CN115297009B (en) * | 2022-07-08 | 2023-11-17 | 中电信数智科技有限公司 | Digital encryption consistency optimization method based on blockchain distributed network |
| WO2024041130A1 (en) * | 2022-08-25 | 2024-02-29 | 华为技术有限公司 | Rights and interests allocation method and apparatus |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113347162B (en) | 2023-09-29 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113347162A (en) | Block link point contribution degree proving consensus method for crowd-sourcing service | |
| CN110535836B (en) | Trust block chain consensus method based on role classification | |
| Nguyen et al. | Modeling and analysis of data trading on blockchain-based market in IoT networks | |
| CN109102287A (en) | Block chain bookkeeping methods, device, equipment and storage medium | |
| CN113222752A (en) | Data processing system, method, computing device and storage medium based on block chain | |
| CN106682825A (en) | System and method for evaluating credit of Social Internet of Things based on block chain | |
| US20140169192A1 (en) | System and Method for Estimating an Effective Bandwidth | |
| CN113095510A (en) | Block chain-based federal learning method and device | |
| CN110599179B (en) | Risk control method and related equipment based on blockchain system | |
| Sorkhoh et al. | Optimizing information freshness for MEC-enabled cooperative autonomous driving | |
| CN113553377B (en) | Data sharing method and device based on block chain and federal learning | |
| CN114418109A (en) | Node selection and aggregation optimization system and method for federal learning under micro-service architecture | |
| CN113987080A (en) | Block chain excitation method and device based on reputation consensus and related products | |
| CN114971421B (en) | Hydropower station operation risk management and control system based on block chain technology | |
| CN113052329A (en) | Method and device for jointly updating service model | |
| CN113177670A (en) | Data processing method, device, equipment and readable storage medium | |
| CN114339733A (en) | Model training method and related equipment | |
| CN115841382A (en) | Virtual power plant transaction block chain, credit evaluation method thereof and consensus mechanism thereof | |
| CN109064329B (en) | Calculation power transaction method and device | |
| CN114169387A (en) | Model training method and device, electronic equipment and storage medium | |
| Polschykov et al. | The methodology of modeling available for data traffic bandwidth telecommunications network | |
| Baniata et al. | Prifob: a privacy-aware fog-enhanced blockchain-based system for global accreditation and credential verification | |
| CN111865595A (en) | Block chain consensus method and device | |
| CN109348451A (en) | The storage method and device of vehicle auxiliary transmission record | |
| CN113138847A (en) | Computer resource allocation scheduling method and device based on federal learning |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |