CN114723441A

CN114723441A - Method, device and equipment for constraining behaviors of demander and participator

Info

Publication number: CN114723441A
Application number: CN202110005390.1A
Authority: CN
Inventors: 史家康; 刘诗阳
Original assignee: China Mobile Communications Group Co Ltd; China Mobile Communications Ltd Research Institute
Current assignee: China Mobile Communications Group Co Ltd; China Mobile Communications Ltd Research Institute
Priority date: 2021-01-05
Filing date: 2021-01-05
Publication date: 2022-07-08

Abstract

The invention provides a method, a device and equipment for restraining behaviors of a demand party and a participant, and belongs to the technical field of artificial intelligence. The method comprises the following steps: receiving a data cooperation request sent by a federal node as an initiator; after the initiator and the participant reach a data cooperation agreement and start the federal learning task, the notary monitors the situation that the initiator and the participant execute the data cooperation agreement in the process of the federal learning task; and if the initiator and the participant complete the federal learning task, evaluating the initiator and the participant, updating the corresponding credibility of the federal node, and writing the credibility into a federal account book. A alliance chain with a certain admission mechanism ensures the credibility of the identities and roles of participants from different organizations in a block chain network through a member identity identification mechanism based on MSP.

Description

Method, device and equipment for constraining behaviors of demander and participator

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to a method, a device and equipment for restricting the behaviors of a demander and a participant.

Background

Federal Learning (Federal Learning) is an emerging basic technology in the field of artificial intelligence, and aims to carry out efficient machine Learning among multiple parties or multiple computing nodes on the premise of guaranteeing information security during large data exchange, protecting terminal data and personal data privacy and guaranteeing legal compliance.

In federal learning, how to build an effective incentive mechanism to keep participants participating in the data federation is an important challenge. The goal is to maximize federal sustainability management while minimizing inter-party unfairness, dynamically allocating a given budget to each party in the data federation. The key to achieving this goal is to formulate a reward method to share federally generated profits fairly and equitably with the participants.

The design of a federal learning incentive mechanism is a problem of finding an optimal organization and payment structure for a federation so as to realize a series of optimization targets.

The existing incentive methods include a method based on an income sharing game, a reverse auction method and the like. When designing incentive mechanisms for federal learning, the contribution of a participant to the data federation is an important factor, and the cost of the participant to join the data federation is also considered. In a given market, some companies may have already occupied a large market share, and thus may accumulate a large amount of high quality data, which is very valuable to the data federation if a high quality federal learning model is to be built.

But this type of market leader may not be intended to help its competitors because the federal learning model will be shared among all participants, incurring potentially large opportunity costs to the market leader.

Disclosure of Invention

The invention aims to provide a method, a device and equipment for restricting the behaviors of a demand party and a participant party, and solve the problem that no trusted third party has potential risk in the existing federal learning training mechanism. In order to solve the above technical problem, embodiments of the present invention provide the following technical solutions:

the embodiment of the invention provides a method for restricting the behaviors of a demand party and a participant, which comprises the following steps:

receiving a data cooperation request sent by a federal node as an initiator;

after the initiator and the participant reach a data cooperation agreement and start the federal learning task, the notary monitors the situation that the initiator and the participant execute the data cooperation agreement in the process of the federal learning task;

and if the initiator and the participant complete the federal learning task, evaluating the initiator and the participant, updating the corresponding credibility of the federal node, and writing the credibility into a federal account book.

In a preferred embodiment, the initiator, the participant and the notary are all federal nodes; the structure of the federal node includes:

an ID of the federated node;

the credibility of the federal node;

the type labels of the federal nodes comprise an initiator, a participant and a notarization party;

an account book entry array;

the latest ledger entry, the largest known submitted ledger entry index value.

In a preferred embodiment, the initiator and the participant entering a data collaboration agreement includes:

step A, as the federal node of the initiator, receiving the information of the participant: if the message is the data collaboration protocol ID sent by the participant, executing the step B; if the risk is over-limit or the credibility of the federal node is not enough, ignoring;

step B, receiving the data cooperation protocol ID and related information;

step one, calculating the protocol feasibility of the initiator, and if the protocol feasibility is greater than a threshold value, sending a message of agreeing to the protocol ID to the participants; otherwise, sending a message of not agreeing with the protocol ID to the participant;

and step D, sending the data cooperation protocol ID to the notarization party.

In a preferred embodiment, the method for the initiator and the participant to reach the data collaboration agreement further comprises the following steps:

a first step, as a federation node of a participant, receiving a message of an initiator, and if the message is a data collaboration request, executing a second step; if the message is an agreement protocol ID from the initiator, adding the agreement ID into a waiting queue according to the weighting of the protocol urgency degree to wait for service; if the message is not the agreement protocol ID, cancelling the data cooperation protocol ID;

a second step of generating a data cooperation protocol ID;

the third step, verify whether the federal node credit degree of the sponsor meets the basic condition of signing the agreement, if meet go to the fourth step, if not meet send "credit degree not enough" message refuse to participate to the sponsor;

step four, calculating the own protocol feasibility of the participant, and if the protocol feasibility is greater than a threshold value, sending the data cooperation protocol ID to the initiator; otherwise, sending a risk overrun message to refuse participation.

In a preferred embodiment, monitoring for situations where the initiator and the participant execute a data collaboration protocol during the federal learning task includes:

coordinating and monitoring a federal learning task and notarizing and arbitrating a data cooperative protocol as a federal node of a notarization party, and executing a second step if a new protocol ID message sent by an initiator is received; if receiving the 'complete protocol ID' message sent by the participant, executing the step four;

step two, storing the protocol of the corresponding protocol ID into a federal account book;

step three, coordinating and monitoring the federal learning task and timing;

and step four, monitoring whether the initiator and the participant successfully complete the federal learning task, and recording the completion time of the initiator and the participant.

In a preferred embodiment, the method further comprises:

designing a data collaboration protocol for describing a collaboration relationship between an initiator and a participant in a federated learning task, the data collaboration protocol comprising:

protocol ID: numbering the data collaboration protocol;

initiator ID: the number of the initiator;

participant ID: the number of the participant;

initiator reputation value: the current federal node credit degree of the initiator;

participant reputation value: the current federal node credit degree of the participant;

notary party ID list: an ID list of a data collaboration protocol notarization party;

data file abstract: ID and hash value of the file provided by the participant;

agreement price and payment mode: the data price and payment mode achieved by the initiator and the participant;

penalty cost: the initiator and the participator do not obey the cost required to be paid by the data cooperation protocol;

protocol period: the validity time interval of the data collaboration protocol.

In a preferred embodiment, the method further comprises: designing the protocol feasibility degree according to the credibility, expected income and default cost of the federal nodes of the initiator and the participants, the length of the current task queue and the urgency of the task, wherein the protocol feasibility degree is designed according to the credibility, the expected income and the default cost of the initiator and the participants, and the current task queue length and the urgency of the task

Protocol feasibility of initiator B

Setting a threshold value mu larger than 0 to limit the feasibility of the protocol, and enabling the participants with the feasibility of the protocol exceeding the threshold value to enter an alternative participant list;

data provider A as participant, agreement feasibility

In the formula:

the Tbase is a guaranteed value of the credibility of the federal node, and a new participant is guaranteed to participate in the credibility of the federal node;

trust (ID, A) is the credibility of the federal node of a participant A when a data cooperation protocol is established, and the Trust (ID, A) of a newly added participant is 0;

waiting (ID, B) is the length of the current Waiting queue of party a when the data coordination protocol is established;

the Urgency degree of the initiator B to subscribe the data cooperation protocol is represented by Unrgency (ID, B), the Unrgency (ID, B) belongs to [0, 1], and the Urgency degree is the highest when the value is 1;

gain (ID) is the maximum gain increment obtained after the participant A completes the federal learning task, and the relationship with the value of the required document is gain (ID) k sigma_f∈F Value(f)；

Penalty (ID, A) is the default cost of Party A;

w₁+w₂+w₃1, and w₁，w₂，w₃∈(0，1)。

In a preferred embodiment, the initiator and the participant are evaluated, and updating the corresponding federated node reputation comprises:

after a data cooperation protocol is executed, calculating the credibility increment of the federal nodes of an initiator and a participant by calling corresponding protocol evaluation functions; the agreement evaluation function is related to the credibility of the federal node of the initiator and the participant before the data cooperation agreement is signed, expected income, default cost and mutual evaluation results of the initiator and the participant.

In a preferred embodiment, the agreement valuation function for party a to calculate the federated node reputation degree increment is:

Trust(ID，A)＝MAX(0，Trust(ID′，A)+p₁×Eval(ID，B)×Gain(ID)-p₂x Penalty (ID, A)), Trust (ID', A) is the reputation of the federal node before transaction A, and Eval (ID, B) is the evaluation of the initiator B on the participation A to fulfill the data collaboration protocol;

the protocol evaluation function of the initiator B for calculating the credibility increment of the federal node is as follows:

Trust(ID，B)＝MAX(0，Trust(ID′，B)+q₁×Eval(ID，A)-q₂x Penalty (ID, B)), Trust (ID', B) is the credibility of the federal node after the previous transaction of the initiator B, and Eval (ID, A) is the evaluation of the participant A on the fulfillment of the data collaboration protocol of the initiator B.

In a preferred embodiment, evaluating the initiator and the participant, and updating the corresponding federated node reputation further comprises:

for the federal node which completes the federal learning task, the credit degree of the federal node is accumulated, and the calculation formula of the credit degree of the federal node is

Wherein tj is the federate node reputation degree increment, Trust, for completing the federate learning task for the ith time_iAnd the credibility of the federal node is after the federal node completes the federal learning task for the ith time.

In a preferred embodiment, after updating the corresponding reputation of the federal node and writing the reputation into the federal ledger, the method further includes:

calculating default cost according to a Penalty function Penalty (ID) of the data cooperation protocol which is not fulfilled by an initiator or a participant in the data cooperation protocol, wherein the ID is the ID of the data cooperation protocol, and the participant _ ID is the ID of the participant;

and executing punishment according to the calculated default cost.

The embodiment of the invention also provides a device for restricting the behaviors of the demand party and the participation party, which comprises the following steps:

the interaction module is used for receiving a data cooperation request sent by a federal node as an initiator;

the monitoring module is used for monitoring the situation that the initiator and the participant execute the data cooperation protocol in the process of the federal learning task by the notarization party after the initiator and the participant reach the data cooperation protocol and the federal learning task is started;

and the evaluation module is used for evaluating the initiator and the participant if the initiator and the participant complete the federal learning task, updating the corresponding credibility of the federal node and writing the credibility into the federal account book.

In a preferred embodiment, further comprising:

the federal node management module is used for managing the structure of the federal node, and the structure of the federal node comprises:

an ID of the federated node;

the credibility of the federal node;

the type labels of the federal nodes comprise an initiator, a participant and a notary party;

an account book entry array;

the latest ledger entry, the known largest submitted ledger entry index value;

the initiator, the participant and the notarization party are all federal nodes.

The initiator and the participant execute a data collaboration protocol during the federal learning task, and in the process, the initiator, the participant and the notary need to perform their functions. In a preferred embodiment, the federal node as initiator performs:

step A, receiving the information of a participant: if the received data collaboration protocol ID is sent by the participant, continuing to execute the step B; if the received message is that the risk is over-limit or the credibility of the federal node is not enough, ignoring the message;

step B, receiving the protocol ID and related information;

step one, calculating the protocol feasibility of the initiator, and if the protocol feasibility of the initiator is greater than a threshold value, sending a message of agreeing with the protocol ID to the participants; otherwise, sending the non-agreement protocol ID to the participant;

and step D, sending the protocol ID to the notarization party.

In a preferred embodiment, the federate node, as a participant, performs:

a first step of receiving a message of an initiator, and if the message is a data cooperation request message, executing a second step; if the message is an agreement protocol ID from the initiator, adding the agreement ID into a waiting queue according to the weighting of the protocol urgency degree to wait for service; if the message is not agreeing to the protocol ID, canceling the protocol ID;

a second step of generating a data cooperation protocol ID;

the third step, verify whether the credibility of the federal node of the initiating side meets the basic condition of signing the agreement, if meet go to the fourth step, if not meet and send "the credibility is not enough" message refuse to participate in;

step four, calculating the own protocol feasibility of the participant, and if the protocol feasibility of the participant is greater than a threshold value, sending the data cooperation protocol ID to the initiator; otherwise, sending a risk overrun message to refuse participation.

In a preferred embodiment, the monitoring module comprises:

a monitoring unit for performing:

step one, for the coordination of the federated learning process and the notarization and arbitration of the data cooperative protocol, if a new protocol ID message sent by an initiator is received, step two is executed; if receiving the message of 'finishing protocol ID' sent by the participant, executing the step four;

step two, the protocol of the specified protocol ID is stored into a federal account book;

step three, coordinating and monitoring the federal learning task and timing;

In a preferred embodiment, further comprising:

a protocol design module, configured to design a data collaboration protocol used for describing a collaboration relationship between an initiator and a participant in a primary data collaboration process, where the data collaboration protocol includes:

protocol ID: numbering the data collaboration protocol;

initiator ID: the number of the initiator;

participant ID: the number of the participant;

participant reputation value: the current federal node credibility of the participant;

data file abstract: ID and hash value of the file that the participant provides;

penalty cost: the cost for the initiator and the participant not to abide by the protocol is paid;

protocol period: the validity time interval of the data collaboration protocol.

In a preferred embodiment, further comprising:

and the feasibility degree design module is used for designing the protocol feasibility degree according to the credibility of the federal nodes of the initiator and the participant, the expected income and the default cost, the current task queue length and the urgency of the task, wherein the feasibility degree is designed according to the credibility of the federal nodes of the initiator and the participant, the expected income and the default cost, and the current task queue length and the urgency of the task

Protocol feasibility of initiator B

Setting a threshold value mu larger than 0 to limit the feasibility of the protocol, and entering the candidate participant list when the feasibility of the protocol exceeds the threshold value;

data provider A as participant, agreement feasibility

In the formula:

the Tbase is a guaranteed value of the credibility of the federal node, and a new participant is guaranteed to have an opportunity to participate in the transaction;

trust (ID, A) is the credibility of the federal node of a participant A when an agreement is established, and t (ID, A) of a newly added participant is 0;

waiting (ID, B) is the length of the credit wait queue of party a when the data cooperation protocol is subscribed;

gain (ID) is the maximum gain increment which can be obtained after the participant A completes the federal learning task, and the relationship with the value of the required file is gain (ID) ═ k Σ_f∈F Value(f)；

Penalty (ID, A) is the default cost of Party A;

w₁+w₂+w₃1, and w₁，w₂，w₃E (0, 1). A requirement file: in the federal learning process, the participants participate in the data files provided during learning. Called "requirements file" because it is needed by the initiator.

In a preferred embodiment, further comprising:

the federated node credit degree module is used for calculating the federated node credit degree increments of the initiator and the participants by calling corresponding protocol evaluation functions after the data cooperation protocol is executed; the agreement evaluation function is related to the credibility of the federal node, expected income, default cost and mutual evaluation results of the participating parties before the data cooperation agreement is signed.

In a preferred embodiment, further comprising:

the protocol evaluation function module is used for realizing that the protocol evaluation function of the participant A for calculating the credibility increment of the federal node is as follows:

Trust(ID，A)＝MAX(0，Trust(ID′，A)+p₁×Eval(ID，B)×Gain(ID)-p₂×Penalty(ID，A))：

in the formula, Trust (ID', A) is the credibility of the federal node before transaction A, and Eval (ID, B) is the evaluation of an initiator B on the data collaboration protocol executed by a participant A;

Trust(ID，B)＝MAX(0，Trust(ID′，B)+q₁×Eval(ID，A)-q₂×Penalty(ID，B))

in the formula, Trust (ID', B) is the credibility of the federal node after the previous transaction of the initiator B, and Eval (ID, A) is the evaluation of the initiator B by the participant A in fulfilling the data collaboration protocol.

In a preferred embodiment, the evaluation module comprises:

a federate node credit degree accumulating unit for accumulating the federate node credit degree of the federate node for the federate node which completes the federate learning task, wherein the calculation formula of the federate node credit degree is

Wherein, t_jIs the federate node credit degree increment, Trust, for completing the federate learning task for the ith time_iIs the credibility of the federal node after the i-th completion of the federal learning task

In a preferred embodiment, further comprising:

the Penalty module is used for calculating default cost according to a Penalty function Penalty (ID, participant _ ID) of the data cooperation protocol which is not fulfilled by an initiator or a participant in the data cooperation protocol, wherein the ID is the ID of the data cooperation protocol, and the participant _ ID is the ID of the participant;

and executing punishment according to the calculated default cost.

The embodiment of the invention also provides a device for restricting the behaviors of the demand party and the participation party, which is applied to the federal node and comprises the following steps:

an initiator module to perform:

step B, receiving the protocol ID and related information;

and step D, sending the protocol ID to the notarization party.

The embodiment of the invention also provides a device for restricting the behaviors of the demand party and the participant, which is applied to the federal node and comprises the following steps:

a participant module to perform:

a first step of receiving a message of an initiator, and if the message is a data cooperation request message, executing a second step; if the message is agreement ID from the initiator, adding the agreement ID into a waiting queue according to the weighting of the protocol urgency degree to wait for service; if the message is a non-agreement protocol ID, canceling the protocol ID;

a second step of generating a data cooperation protocol ID;

the third step, verify whether the federal node credit degree of the initiator meets the basic condition of signing agreement, if meet, transfer to the fourth step, if not meet and send "credit degree is not sufficient" message refuse to participate in;

The embodiment of the invention also provides equipment for restricting the behaviors of the demand party and the participation party, which comprises the following steps: a processor, a memory, and a program stored on the memory and executable on the processor, the program, when executed by the processor, implementing steps in a method of selecting a trusted third party as a coordinator for federal learning.

Embodiments of the present invention also provide a computer-readable storage medium having stored thereon a computer program that, when executed by a processor, performs the steps in the method for selecting a trusted third party as a coordinator for federated learning as described above.

The embodiment of the invention has the following beneficial effects: a coalition chain with a certain admission mechanism ensures the credibility of the identities and roles of participants from different organizations in a block chain network through a member identity identification mechanism based on MSP. And isolates participants of different organizations by defining channels. On the basis, the block chain has the advantages of high safety by combining the inherent decentralization characteristic, common recognition mechanism, non-tampering and the like of the block chain; social optimization: the social index is maximized; robustness: independent of probability distribution knowledge; and (3) collusion resistance: no participant can gain greater utility through collusion fraud.

Drawings

FIG. 1 is a flow chart illustrating a method for sharing incentives among data collaboration participants in accordance with an embodiment of the present invention;

FIG. 2 is a system architecture diagram illustrating the use of a hierarchy to define federated learning techniques in accordance with an embodiment of the present invention;

FIG. 3 is a schematic diagram of a deployment architecture of federation chain-based federated learning technique according to an embodiment of the present invention;

FIG. 4 is a flowchart illustrating an originator signing a data collaboration protocol according to an embodiment of the present invention

FIG. 5 is a flowchart illustrating a data collaboration protocol signed by a participant according to an embodiment of the present invention;

FIG. 6 is a diagram illustrating a protocol evaluation process of a notarization party in accordance with an embodiment of the present invention;

FIG. 7 is a diagram illustrating interaction processes for parties involved in establishing and executing a collaboration protocol according to an embodiment of the invention;

fig. 8 is a schematic structural diagram of an apparatus for implementing a shared incentive between data collaboration participants according to an embodiment of the present invention.

Detailed Description

In order to make the technical problems, technical solutions and advantages to be solved by the embodiments of the present invention clearer, the following detailed description will be given with reference to the accompanying drawings and specific embodiments.

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 some, not all, embodiments of the present invention. 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 terms "comprises," "comprising," or any other variation thereof, in the description and claims of this application, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus. Furthermore, the use of "and/or" in the specification and claims means that at least one of the connected objects, such as a and/or B, means that three cases, a alone, B alone, and both a and B, exist.

In the embodiments of the present invention, words such as "exemplary" or "for example" are used to mean serving as examples, illustrations or descriptions. Any embodiment or design described as "exemplary" or "e.g.," an embodiment of the present invention is not necessarily to be construed as preferred or advantageous over other embodiments or designs. Rather, use of the word "exemplary" or "such as" is intended to present concepts related in a concrete fashion.

In federal learning, how to build an effective incentive mechanism to keep participants participating in the data federation is an important challenge. The goal is to maximize federal sustainability management while minimizing unfairness among the participants, dynamically allocating a given budget to each participant in the data federation. There is a need to formulate a reward method to share federally generated profits fairly and equitably between participants.

The negotiation mechanism in the social economic activity is to restrict the behaviors of the demand party and the participation party through the interest relations and the agreements of the supply and demand parties.

Therefore, the design principle of the present application is: firstly, a demand side, namely an initiator, exchanges the contribution amount agreed by a protocol for consumption amount; a protocol notarization party is arranged to be responsible for evaluating the protocol execution condition and storing a third party of the protocol, so that deception is reduced; the participant calculates the agreement feasibility degree according to the file value, the participant credit value, the income increment available to the participant and the like, and the agreement feasibility degree of the two parties exceeds a threshold value to reach an agreement; after reaching the cooperative agreement, the two parties can do transaction, but must fulfill the commitment of the agreement before the end time point of the agreement, otherwise accept the punishment.

In the present application, a blockchain is used as an intermediary between a data provider and a data demander to add a notarization party, an embodiment of the present invention provides a method for implementing a sharing incentive between data collaboration participants, as shown in fig. 1, where the method includes:

step 11: receiving a data cooperation request sent by a federal node as an initiator;

step 12: after the initiator and the participant reach a data cooperation agreement and start the federal learning task, the notary and notary monitor the situation that the initiator and the participant execute the data cooperation agreement in the process of the federal learning task;

step 13: and if the initiator and the participant complete the federal learning task, evaluating the initiator and the participant, updating the corresponding credibility of the federal node, and writing the credibility into a federal account book.

According to the technical scheme provided by the application embodiment, on the basis of an original federal learning mechanism, a blockchain is used as an intermediary between a data provider and a data demander and is used for storing certificates of protocols, transactions and the like.

To help those skilled in the art understand the technical solution provided by the present application, as shown in fig. 2, a system architecture of the federal learning technology is defined by using a hierarchical structure based on a federation chain as a carrier of data federation:

and the resource layer is used for bearing data and equipment resources.

And the entity layer is used for bearing each organization (also called organization member) in the data alliance, each organization member in the data alliance can initiate a federal learning task, the organization members participating in the federal learning task are participating parties, and the organization members not participating in the federal learning task are third parties. The data federation includes a plurality of organizations, each organization forming a federation chain.

And the platform layer defines core platform components of the system architecture, and comprises a alliance chain platform and a federal learning platform. By adding the trusted third party selection component in the alliance chain platform, the trusted third party selection function is provided for federal learning.

And the application layer defines the application provided for the user and provides views corresponding to various functions according to different roles of the user.

It can be seen that, as shown in fig. 3, a deployment architecture of an actual federated learning system based on a federation chain is described, in the deployment architecture of the federated learning technology based on the federation chain, there are N organizations in a data federation, where the N organizations form a federation chain, and according to a specific data collaboration situation, a series of channels are defined, each channel corresponds to one data federation, and each organization may add multiple channels (data federations).

As illustrated, organization 1, organization 2, organization N, and the like belong to the data federation 1, and organization 3, organization N, and the data federation N. Nodes in the same data federation may participate in a common federated learning task.

In the technical scheme provided by the embodiment of the application, the block chain technology is fused, and on the basis of the original federal learning technology, the block chain is used as an intermediary between a data provider and a data demander and is used for storing certificates of protocols, transactions and the like.

And recording the transactions among all users in the federal book, including the change of the corresponding user reputation value and the like.

The related business logic is realized, compiled and deployed in the form of intelligent contract codes. Such as management of data supply and demand among users, e.g., establishment, execution, and post-event evaluation of collaborative protocols, scheduling of federal learning tasks, federal ledger operations, and the like.

Data consistency is maintained for each node in the data federation through a consensus mechanism.

Therefore, in order to realize the corresponding technical scheme, a network structure is designed, and the network structure is divided into four layers: a data layer (shared file), a user layer (participant), a node layer (Blockchain node), and a ledger layer (Blockchain ridge). Each participant in the user layer is a data holder and is responsible for updating the local model in the process of federal learning. Each block chain link point in the node layer provides safety aggregation of partial local models for the participants of the user layer, and the whole local model is received towards the account layer. The account book layer is responsible for storing data transactions among all users, receiving data uploaded by block chain nodes and finishing the aggregation of the Federal learning global model.

It is also necessary to design a structure describing the current state of the federated node, with various fields in the structure. In a preferred embodiment, the initiator, participant and notary are all federal nodes; the structure of the federal node comprises:

an ID of the federated node;

the credibility of the federal node;

an account book entry array;

the latest ledger entry, the largest known committed ledger entry index value.

And, designing a data file:

data structures describing shared data in the data federation, basic information from multiple dimension description files, such as:

file ID: numbering of data files

Document summarization: the summary of the data file includes, but is not limited to, file description, file type, file structure, field name, data type, data size, application, and the like.

File fingerprint: hash value of data file

File access path: data file access path in data federation

Data value: the calibration or evaluation value of the data.

step A, as the federal node of the initiator, receiving the information of the participant: if the received data collaboration protocol ID is sent by the participant, continuing to execute the step B; if the received message is that the risk is over-limit or the credibility of the federal node is not enough, ignoring the message;

step B, receiving the protocol ID and related information;

and step D, sending the protocol ID to the notarization party.

And the algorithm that the initiator signs the data cooperation protocol is realized in the corresponding application scene. In the data collaboration process of the initiator, as shown in fig. 4, the algorithm for signing up the data collaboration protocol is as follows:

step 41, the initiator B receives the message of the participant a: if the protocol ID sent by the participant is received, the step 42 is executed; otherwise, turning to step 45;

step 42, the initiator receives the protocol ID and the related information sent by the participant;

43, the initiator calculates the feasibility of the protocol;

step 44, if the initiator judges that the protocol feasibility is greater than the threshold, turning to step 45; otherwise go to step 47;

step 45, sending a message of 'agreeing to protocol ID' to the participants;

and step 46, the initiating party sends the protocol ID to the protocol notarization party, and the process is finished.

And step 47, sending a message of 'not agreeing with the protocol ID' to the participant, and ending.

And step 48, if the message of 'risk overrun' or 'reputation insufficient' sent by the participant is received, ignoring the message.

a first step, as a federation node of a participant, receiving a message of an initiator, and if the message is a data collaboration request message, executing a second step; if the message is an agreement protocol ID from the initiator, adding the agreement ID into a waiting queue according to the weighting of the protocol urgency degree to wait for service; if the message is a non-agreement protocol ID, canceling the protocol ID;

a second step of generating a data cooperation protocol ID;

And the algorithm that the participants sign the data collaboration protocol is realized in the corresponding application scene. In the data collaboration process of the participants, as shown in fig. 5, the algorithm for signing up the data collaboration protocol is as follows:

step 51, the participant receives the message of the initiator: if the received data collaboration request message is the data collaboration request message, continuing to execute step 52; otherwise go to step 59;

step 52, the participant receives the request of the initiator and generates a data cooperation protocol ID;

step 53, judging whether the participant verifies that the credit of the initiator meets the basic condition of signing the agreement, wherein the basic condition is whether the initiator has the payment capability of the claimed agreement price, and if the initiator meets the basic condition, turning to step 54; if not, go to step 58;

if the initiator meets the basic conditions, step 54, the participants calculate their own protocol feasibility,

step 55, judging whether the protocol feasibility of the initiator is greater than a threshold, and if so, turning to step 56; otherwise go to step 57.

Step 56, the protocol ID is sent to the initiator, and the process is ended.

And 57, sending a risk overrun message to the initiator to refuse participation, and ending.

Step 58, sending a "reputation not sufficient" message to refuse participation.

Step 59, if the message of 'agreement to protocol ID' of the initiator is received, adding the message into a waiting queue according to the weighting of the protocol urgency degree to wait for service; if the message that the initiator does not agree with the protocol ID is received, the data cooperation protocol ID is cancelled.

In a preferred embodiment, monitoring for situations where the initiator and the participant perform a data collaboration protocol during the federal learning task comprises:

step one, a federal node as a notarization party performs notarization and arbitration on a coordination and data cooperation protocol in a federation learning process, and if a new protocol ID message sent by an initiator is received, step two is executed; if receiving the message of 'finishing protocol ID' sent by the participant, executing the step four;

step two, the protocol of the specified protocol ID and the protocol are stored into a federal account book;

step three, coordinating and monitoring the federal learning task and timing;

In a corresponding application scenario, as shown in fig. 6, the protocol evaluation process of the notary party is as follows:

step 61, selecting a notarization party from the data federation, and receiving messages from all the participating parties;

step 62, judging the specific source and content of the message, and executing step 63 if the received message is the 'new protocol ID' message sent by the initiator B; if the message of 'completing protocol ID' sent by the participant A/the initiator B is received, executing the

steps

65 and 66;

step 63, the notarization party stores the protocol of the specified ID into the federal account book;

and step 64, the notarization party starts the coordination and monitoring of the federal learning task, times and ends.

Step 65, receiving a message of 'completing protocol ID' sent by the participant A, monitoring whether all the participants successfully complete the federal learning task, and recording the completion time of the participant A; turning to step 67;

step 66, receiving a message of 'completing protocol ID' sent by the initiator B, monitoring whether all the participants successfully complete the federal learning task, and recording the completion time of the participants B; turning to step 67;

and step 68, the notarization party evaluates the protocol.

And in order to describe the primary data agreement, if all the participants successfully complete the federal learning task, performing agreement evaluation, updating the node credit value and writing the node credit value into a federal account book. In the same process, the cooperation relationship between the initiator and the participant is needed, and therefore, a data cooperation protocol needs to be designed. In a preferred embodiment, before receiving a data cooperation request sent by a federal node as an initiator, the method further includes:

designing a data collaboration protocol for describing a collaboration relationship between an initiator and a participant in a data collaboration process, wherein the data collaboration protocol comprises the following steps:

protocol ID: numbering the data collaboration protocol;

initiator ID: the number of the initiator;

participant ID: the number of the participant;

penalty cost: the cost for the initiator and the participant not to abide by the protocol is paid; including but not limited to monetary amount, credit value, etc.;

protocol period: the validity time interval of the data collaboration protocol.

The estimation of the possibility of protocol implementation by both participating parties from their own interests is related to the reputation value, expected profit, default cost of the data provider, the current length of the task queue, the urgency of the task, and the like. In a preferred embodiment, before receiving a data cooperation request sent by a federal node as an initiator, the method further includes: designing the protocol feasibility degree according to the credibility, expected income and default cost of the federal nodes of the initiator and the participants, the length of the current task queue and the urgency of the task, wherein the protocol feasibility degree is designed according to the credibility, the expected income and the default cost of the initiator and the participants, and the current task queue length and the urgency of the task

Protocol feasibility of initiator B

data provider A as participant, agreement feasibility

In the formula:

trust (ID, A) is the credibility of the federal node of a participant A when a protocol is established, and t (ID, A) of a newly added participant is equal to 0;

Penalty (ID, A) is the default cost of Party A;

w₁+w₂+w₃1, and w₁，w₂，w₃E (0, 1). A requirement file: in the federal learning process, the participants participate in the data files provided during learning. It is called a "requirement document" because it is needed by the initiator.

It can be seen that the federal learning initiator B has the right to decide which data providers' data to use, and the above agreement feasibility provides a decision basis for the federal learning initiator (strong data demander). A threshold value μ > 0 may be set to limit the agreement feasibility, which is exceeded before entering the list of alternate participants.

The system may then set up the agreement for automatic or manual bargaining. And when the automatic transaction is carried out, judging by a system intelligent contract, automatically selecting the participants according to the protocol feasibility of different participants, and signing a data cooperative protocol. When the manual transaction is made, the initiator synthesizes a plurality of factors according to the candidate parties (according to the feasibility degree of the agreement and the like) screened out by the intelligent contract, and manually selects the participants to sign the data cooperation agreement.

Agreement evaluation is the determination of the effect of the agreement being fulfilled by the two parties to the transaction. And after the protocol is executed, the reputation value of the participant is adjusted through a protocol evaluation function. The evaluation function is related to factors such as reputation value before the agreement is signed, expected income, default cost, mutual evaluation results of the participating parties and the like. In a preferred embodiment, the initiator and the participant are evaluated, and updating the corresponding federated node reputation comprises:

after a data cooperation protocol is executed, calculating the credibility increment of the federal nodes of an initiator and a participant by calling corresponding protocol evaluation functions; the agreement evaluation function is related to the credibility of the federal node, expected income, default cost and mutual evaluation results of the participating parties before the data cooperation agreement is signed.

For example, a data provider (participant) a in the agreement may evaluate by calculating a reputation value increment through the following function, and in a preferred embodiment, the agreement evaluation function for participant a to calculate the reputation degree increment of the federated node is:

the protocol evaluation function of the initiator B for calculating the credibility increment of the federated node is as follows:

In a preferred embodiment, evaluating the initiator and the participant, and updating the corresponding federal node reputation degree further comprises:

Wherein, t_jIs the federate node credit degree increment, Trust, for completing the federate learning task for the ith time_iAnd the credibility of the federal node after the i-th completion of the federal learning task.

After the cooperative agreement is reached, the two parties can carry out transaction, but must fulfill the promise of the agreement on demand before the agreement ending time point, otherwise, the penalty is accepted. In a preferred embodiment, after updating the corresponding reputation of the federal node and writing the reputation into the federal ledger, the method further includes:

and executing punishment according to the calculated default cost.

And, design penalty cost:

the penalty of failure of a party to perform in the data collaboration protocol is defined, and may be described as:

Penalty(ID，participant_ID)

ID: data collaboration protocol ID

participant _ ID: participant ID

And designing a protocol notarization party: namely, the trusted third party of the federal study, is responsible for the coordination of the federal study process and the notarization and arbitration of the data collaboration protocol.

In an application scenario, using the technical means provided by the above embodiment, for a typical federal learning task in a typical collaborative protocol interaction process, as shown in fig. 7, an interaction process for relevant parties to establish and execute a collaborative protocol includes:

step 71, the federal study initiates a data cooperation request to the data federal;

step 72, calling an intelligent contract, and screening out an alternative participant list;

step 73, generating a data cooperation protocol for each alternative party in the alternative party list;

step 74, calculating the feasibility of the protocols in the alternative list, then judging by using a threshold value, updating the list and sequencing;

step 75, selecting the final participant according to the ranking of the feasibility degree, and the initiator can select automatic transaction or manual transaction:

judging and selecting a final participant by a system intelligent contract when the automatic transaction is carried out;

when the hand-held transaction is made, the initiator sorts according to the feasibility degree, integrates various factors and manually selects the final participant.

Step 76, calling an intelligent contract to sign a data cooperation protocol, and writing the data cooperation protocol into a federal account book;

77, each party reaching the agreement starts a federal learning task and executes the agreement, wherein, when 3 or more parties exist, the initiator and each party sign a cooperative agreement respectively;

step 78, the protocol notarization party evaluates the protocol execution condition, updates the node credit value and writes the node credit value into a federal account book; and selecting the notarization party, namely the trusted third party in the federal study, through a certain mechanism.

A federal account book, namely a block chain account book in the data federation adopts a distributed storage structure to store transactions and protocols among all users, including changes of corresponding user reputation values and the like.

An embodiment of the present invention further provides a device for restricting behaviors of a demanding party and a participating party, as shown in fig. 8, the device includes:

the interaction module 81 is configured to receive a data cooperation request sent by a federal node as an initiator;

the monitoring module 82 is used for monitoring the situation that the initiator and the participant execute the data cooperation agreement in the process of the federal learning task by the notary after the initiator and the participant reach the data cooperation agreement and the federal learning task is started;

and the evaluation module 83 is configured to evaluate the initiator and the participant if both the initiator and the participant complete the federal learning task, update the corresponding federal node credit degree, and write the updated federal node credit degree into a federal book.

In a preferred embodiment, further comprising:

an ID of the federated node;

the credibility of the federal node;

an account book entry array;

the latest ledger entry, the known largest submitted ledger entry index value;

step B, receiving the protocol ID and related information;

and step D, sending the protocol ID to the notarization party.

In a preferred embodiment, the federate node, as a participant, performs:

a first step of receiving a message of an initiator, and if the message is a data cooperation request message, executing a second step; if the message is an agreement protocol ID from the initiator, adding the agreement ID into a waiting queue according to the weighting of the protocol urgency degree to wait for service; if the message is a non-agreement protocol ID, canceling the protocol ID;

a second step of generating a data cooperation protocol ID;

step four, calculating the own protocol feasibility of the participant, and if the protocol feasibility of the participant is greater than a threshold value, sending the data cooperation protocol ID to the initiator; otherwise, sending a 'risk overrun' message to refuse participation.

In a preferred embodiment, the monitoring module 82 includes:

a monitoring unit for performing:

step three, coordinating and monitoring the federal learning task and timing;

In a preferred embodiment, further comprising:

a protocol design module, configured to design a data collaboration protocol used for describing a collaboration relationship between an initiator and a participant in a data collaboration process, where the data collaboration protocol includes:

protocol ID: numbering the data collaboration protocol;

initiator ID: the number of the initiator;

participant ID: the number of the participant;

protocol period: the validity time interval of the data collaboration protocol.

In a preferred embodiment, further comprising:

Protocol feasibility of initiator B

data provider A as participant, and the feasibility degree of the protocol

In the formula:

gain (ID) as a participantThe maximum gain increment obtained after the A completes the federal learning task and has the relationship with the value of a demand file of gain (ID) k sigma_f∈F Value(f)；

Penalty (ID, A) is the default cost of Party A;

In a preferred embodiment, further comprising:

In a preferred embodiment, the evaluation module 83 comprises:

the federal node credit degree accumulation unit is used for accumulating the federal node credit degree of the federal node for the federal node which completes the federal learning task, and the calculation formula of the federal node credit degree is

In a preferred embodiment, further comprising:

the Penalty module is used for calculating default cost according to a Penalty function Penalty (ID, participant _ ID) of the data cooperation protocol which cannot be fulfilled by an initiator or a participant in the data cooperation protocol, wherein the ID is the ID of the data cooperation protocol, and the participant _ ID is the ID of the participant;

and executing punishment according to the calculated default cost.

an initiator module to perform:

step B, receiving the protocol ID and the related information;

step C, calculating the protocol feasibility of the initiator, and if the protocol feasibility of the initiator is greater than a threshold value, sending a message of agreeing to the protocol ID to the participants; otherwise, sending the non-agreement protocol ID to the participant;

and step D, sending the protocol ID to the notarization party.

a participant module to perform:

a second step of generating a data cooperation protocol ID;

The technical scheme provided by the application embodiment has a alliance chain with a certain admission mechanism, and the identity and the role credibility of the participants from different organizations in the block chain network are ensured through a member identity identification mechanism based on the certificate Management (MSP) of alliance chain members. And isolates participants of different organizations by defining channels. On the basis, the block chain has the advantages of high safety by combining the inherent decentralization characteristic, common recognition mechanism, non-tampering and the like of the block chain; social optimization: the social index is maximized; robustness: independent of probability distribution knowledge; and (3) collusion resistance: no participant can gain greater utility through collusion fraud. MSP is a fundamental technology mechanism for alliance blockchains. The membership service provider MSP is used to determine which certificate authorities CA are trusted, list organization membership, identify the specific roles participants play within the organization, define network and channel access rights, etc.

Computer-readable media, including both permanent and non-permanent, removable and non-removable media, may implement the information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), Static Random Access Memory (SRAM), Dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), Read Only Memory (ROM), Electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technologies, compact disc read only memory (CD-ROM), Digital Versatile Disks (DVD) or other optical storage, magnetic cassettes, magnetic tape storage or other magnetic storage terminal devices to be detected, or any other non-transmission medium that can be used to store information that can be accessed by a computer to be detected terminal device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

While the foregoing is directed to the preferred embodiment of the present invention, it will be understood by those skilled in the art that various changes and modifications may be made without departing from the spirit and scope of the invention as defined in the appended claims.

Claims

1. A method of constraining the behavior of a requiring party and a participating party, the method comprising:

receiving a data cooperation request sent by a federal node as an initiator;

after the initiator and the participant reach a data cooperation agreement and start the federal learning task, the notary and notary monitor the situation that the initiator and the participant execute the data cooperation agreement in the process of the federal learning task;

2. The method of claim 1, wherein the initiator, the participant, and the notary are all federal nodes; the structure of the federal node includes:

an ID of the federal node;

the credibility of the federal node;

an account book entry array;

the latest ledger entry, the largest known committed ledger entry index value.

3. The method of claim 1, wherein the initiator and the participant entering a data collaboration agreement comprises:

step B, receiving the data cooperation protocol ID and related information;

and step D, sending the data cooperation protocol ID to the notarization party.

4. The method of claim 1, wherein the initiator and the participant entering a data collaboration agreement further comprises:

a second step of generating a data cooperation protocol ID;

5. The method of claim 1, wherein monitoring for instances in which the initiator and the participant execute a data collaboration protocol during a federal learning task comprises:

step three, coordinating and monitoring the federal learning task and timing;

6. The method according to any one of claims 3-5, characterized in that the method further comprises:

protocol ID: numbering the data collaboration protocol;

initiator ID: the number of the initiator;

participant ID: the number of the participant;

the notarization party ID list is an ID list of a notarization party of the data cooperation protocol;

the agreement price and the payment mode are the data price and the payment mode achieved by the initiator and the participant;

protocol period: the validity time interval of the data collaboration protocol.

7. The method according to any one of claims 3-5, characterized in that the method further comprises: designing the feasibility degree of the protocol according to the credibility, expected income and default cost of the federal nodes of the initiator and the participant, the length of the current task queue and the urgency of the task, wherein the feasibility degree of the protocol is designed according to the credibility, the expected income and the default cost of the federal nodes of the initiator and the participant, and the feasibility degree of the protocol is designed according to the length of the current task queue and the urgency of the task

Protocol feasibility of initiator B

Setting a threshold value mu more than 0 to limit the feasibility degree of the protocol, and enabling the participants with the feasibility degree exceeding the threshold value to enter an alternative participant list;

data provider A as participant, agreement feasibility

In the formula:

gain (ID) is the maximum gain increment obtained after the participant A completes the federal learning task, and the relationship with the value of the demand document is gain (ID) ═ k Σ_f∈FValue(f)；

Penalty (ID, A) is the default cost of Party A;

w₁+w₂+w₃1, and w₁,w₂,w₃∈(0，1)。

8. The method of claim 7, wherein evaluating the initiator and the participant and updating the corresponding federated node reputation comprises:

after a data cooperation protocol is executed, the corresponding protocol evaluation functions are called to calculate the credibility increment of the federal nodes of an initiator and a participant; the agreement evaluation function is related to the credibility of the federal node of the initiator and the participant before the data cooperation agreement is signed, expected income, default cost and mutual evaluation results of the initiator and the participant.

9. The method of claim 8,

the protocol evaluation function of the participant A for calculating the credibility increment of the federal node is as follows:

Trust(ID,A)＝MAX(0,Trust(ID′,A)+p₁×Eval(ID,B)×Gain(ID)－p₂x Penalty (ID, A)), Trust (ID', A) is the reputation of the federal node before transaction A, and Eval (ID, B) is the evaluation of the initiator B on the participation A to fulfill the data collaboration protocol;

Trust(ID,B)＝MAX(0,Trust(ID′,B)+q₁×Eval(ID,A)－q₂x Penalty (ID, B)), Trust (ID ', B) is the federal node reputation after the initiator B's previous transaction, Eval (ID, a) is the participant a's evaluation of the initiator B's performance of the data collaboration protocol.

10. The method of claim 9, wherein evaluating the initiator and the participant and updating the corresponding federated node reputation further comprises:

Wherein, t_jIs the federate node credit degree increment, Trust, for completing the federate learning task for the ith time_iAnd the credibility of the federal node is after the federal node completes the federal learning task for the ith time.

11. The method of claim 8, wherein after updating the corresponding federated node reputation and writing into a federated ledger, further comprising:

and executing punishment according to the calculated default cost.

12. An apparatus for constraining the behavior of a requiring party and a participating party, the apparatus comprising:

the monitoring module is used for monitoring the situation that the initiator and the participant execute the data cooperation protocol in the process of the federal learning task by a notary party after the initiator and the participant reach the data cooperation protocol and the federal learning task is started;

13. An apparatus for constraining the behavior of a requiring party and a participating party, applied to a federated node, the apparatus comprising:

an initiator module to perform:

step B, receiving the protocol ID and related information;

and step D, sending the protocol ID to the notarization party.

14. An apparatus for constraining the behavior of a requiring party and a participating party, applied to a federated node, the apparatus comprising:

a participant module to perform:

a second step of generating a data cooperation protocol ID;

15. A behavior device that constrains a requiring party and a participating party, comprising: processor, memory and program stored on the memory and executable on the processor, which when executed by the processor implements the steps of the method according to any of claims 1-11.

16. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method of carrying out a shared incentive between data cooperating parties as claimed in any one of claims 1 to 11.