Privacy protection method and device for resisting internal dishonest participants

Abstract

The invention provides a privacy protection method and device for resisting internal dishonest participants. The method comprises the steps that local model parameters to be uploaded by the participants are acquired; performing commitment calculation on the local model parameters based on a commitment function to obtain a commitment function value; the local model parameters and the corresponding commitment function values are sent to the bookkeeping right miner, so that the bookkeeping right miner writes the local model parameters and the corresponding commitment function values into blocks in a BFL framework, and a global model is synthesized; and performing data privacy enhancement on the global model parameter based on each participant, and writing the global model parameter after data privacy enhancement into a block in the BFL framework. According to the method, the intelligent contract based on the commitment function and the blind signature is written into the block chain, and when the dishonest participant denies or leaks the model, the dishonest participant is determined according to the contract opening commitment or signature, so that the resistance and defense capability of the internal dishonest participant is improved, and the privacy security is guaranteed.

Description

Technical Field

[0001] The present invention relates to the field of blockchain technology, and in particular to a privacy protection method and device for resisting internal dishonest participants. Background Art

[0002] The Blockchain-based Federated Learning (BFL) framework uses blockchain instead of a central server. Participants submit parameters such as the gradient of a local model trained locally to miners. Miners perform gradient verification and aggregation according to the consensus mechanism or smart contract, generate a global model for training and write its parameters into the block. The global model parameters are then stored and broadcasted using blocks. Participants can download the global model parameters and conduct the next round of local model training.

[0003] The existing BFL framework has weak resistance and defense capabilities against internal dishonest participants. Internal dishonest participants may cheat the global model parameters obtained by the overall training aggregation of the system by providing invalid local model parameters, or obtain the training data, identity information or local model information of other participants through interaction and monitoring within the BFL system, or even collude with external adversaries to actively leak the obtained global model or local model information, posing a huge privacy and security risk and possibly causing overall privacy leakage. Summary of the invention

[0004] The present invention provides a privacy protection method and device for resisting internal dishonest participants, which are used to enhance the resistance and defense capabilities of a blockchain-based federated learning framework against internal dishonest participants, thereby ensuring privacy security.

[0005] The present invention provides a privacy protection method for resisting internal dishonest participants, comprising the following steps: Obtain the local model parameters to be uploaded by participants in the blockchain-based federated learning BFL framework; Based on a predetermined commitment function, performing commitment calculation on the local model parameters to obtain a commitment function value; Send the local model parameters and their corresponding commitment function values ​​to the bookkeeping rights miner, so that the bookkeeping rights miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework and synthesizes the global model; Data privacy is enhanced for global model parameters in the global model based on each participant respectively, and the global model parameters after data privacy enhancement are written into the blocks in the BFL framework.

[0006] According to a privacy protection method for resisting internal dishonest participants provided by the present invention, after determining that the accounting right miner writes the local model parameters and the corresponding commitment function values ​​into the block in the BFL framework and synthesizes the global model, it also includes: Receive abnormal local model parameters, where the abnormal local model parameters are local model parameters that the participants disagree with during the verification process or deny uploading. The verification process is the verification before the accounting right miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework; Based on the commitment function, the abnormal local model parameters are parsed for public verification.

[0007] According to a privacy protection method for resisting internal dishonest participants provided by the present invention, the

[0008] The data privacy is enhanced for the global model parameters in the global model based on each participant, respectively, including: Adding different parameter noises to the global model parameters in the global model based on each participant respectively, and performing a data blinding operation on the global model parameters with added parameter noise based on the public key of each participant to generate blinded global model parameters; The blinded global model parameters are signed based on the private key of each participant.

[0009] According to a privacy protection method for resisting internal dishonest participants provided by the present invention, the

[0010] Perform data blinding on global model parameters with added parameter noise, including: Based on the blinding factor, data blinding is performed on the global model parameters with added parameter noise.

[0011] A privacy protection method for resisting internal dishonest participants provided by the present invention also includes: In the case where there is a global model parameter leakage in the BFL framework, obtaining the leaked global model parameter; Based on the blinding factor and the public key of the participant, the leaked global model parameter is parsed to determine the target participant corresponding to the leaked global model parameter.

[0012] According to a privacy protection method for resisting internal dishonest participants provided by the present invention, after determining the target participant corresponding to the leaked global model parameter, the method further includes: Deduct a preset amount of the target participant's token deposit, and after determining that the deduction of the target participant's token deposit is completed, delete the target participant from the BFL framework, wherein the token deposit is the deposit that the target participant needs to pay before joining the BFL framework.

[0013] The present invention also provides a privacy protection device for resisting internal dishonest participants, comprising the following modules: The parameter acquisition module is used to obtain the local model parameters to be uploaded by participants in the blockchain-based federated learning BFL framework; A commitment calculation module, used to perform commitment calculation on the local model parameters based on a predetermined commitment function to obtain a commitment function value; An aggregation module, used to send the local model parameters and their corresponding commitment function values ​​to the bookkeeping rights miners, so that the bookkeeping rights miners write the local model parameters and their corresponding commitment function values ​​into the blocks in the BFL framework and synthesize the global model; The privacy enhancement module is used to perform data privacy enhancement on the global model parameters in the global model based on each participant, and write the global model parameters after data privacy enhancement into the blocks in the BFL framework.

[0014] The present invention also provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the privacy protection method for resisting internal dishonest participants as described in any one of the above methods is implemented.

[0015] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the privacy protection methods for resisting internal dishonest participants as described above.

[0016] The present invention also provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-described privacy protection methods for resisting internal dishonest participants.

[0017] The privacy protection method and device for resisting internal dishonest participants provided by the present invention perform commitment calculation on the uploaded local model parameters through a predetermined commitment function, and obtain the commitment function value for resisting the situation where the internal dishonest participants use invalid local model parameters to maliciously defraud the global model parameters. It realizes writing a smart contract based on the commitment function in the blockchain. When a dishonest participant denies, the dishonest participant is determined by opening the commitment according to the contract, which improves the resistance and defense capabilities of the blockchain-based federated learning framework against internal dishonest participants and ensures privacy security. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0019] Figure 1 It is a flow chart of the privacy protection method for resisting internal dishonest participants provided by the present invention.

[0020] Figure 2 It is a schematic diagram of the privacy protection architecture provided by the present invention.

[0021] Figure 3 It is a structural schematic diagram of the privacy protection device for resisting internal dishonest participants provided by the present invention.

[0022] Figure 4 It is a structural schematic diagram of the electronic device provided by the present invention. DETAILED DESCRIPTION

[0023] In order to make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be clearly and completely described below in conjunction with the drawings of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0024] Federated learning can break data silos and complete machine learning training without leaving the local data. Therefore, it has received widespread attention and has gradually been applied to various industries. However, the centralized nature of federated learning poses a great privacy and security risk. The decentralized nature of blockchain provides opportunities for the development of federated learning. In recent years, the Blockchain-based Federated Learning (BFL) framework has been proposed. Blockchain is used instead of the central server. Participants submit parameters such as the gradient of a local model trained locally to miners. Miners perform gradient verification, aggregation, and generate a global model for training according to the consensus mechanism or smart contract, and write its parameters into the block. Then, the global model parameters are stored and broadcasted using the block. Participants can download the global model parameters and conduct the next round of local model training.

[0025] The BFL framework in the related methods has weak resistance and defense capabilities against internal dishonest participants: internal dishonest participants may cheat the global model parameters obtained by the overall training aggregation of the system by providing invalid local model parameters, or obtain the training data, identity information or local model information of other participants through the interaction and monitoring within the BFL system, or even collude with external adversaries to actively leak the obtained global model or local model information, which will form a huge privacy and security risk and may cause overall privacy leakage. Therefore, it is necessary to screen out the dishonest participants in the system and remove them from the system.

[0026] In the research on blockchain-based federated learning among related methods, there are four main types of privacy protection schemes, but all of them have problems: 1) Some schemes use differential privacy technology to add noise to local gradients to ensure privacy, which can easily cause deviations in the joint model and cannot screen out dishonest participants; 2) Some schemes use homomorphic encryption or secret sharing to achieve external privacy, which cannot prevent internal dishonest participants from colluding to leak privacy; 3) Some schemes use channel isolation technology to ensure the privacy of information transmission, which is only applicable to consortium chain systems that support this technology and is not universal; 4) Some schemes use the method of regularly electing committees or leaders to aggregate models, which provides dishonest participants with the opportunity to obtain all data and increases the risk of privacy leakage due to single point failures.

[0027] In view of the defects of the related methods, the present invention provides a privacy protection method for resisting internal dishonest participants. Figure 1 is a flow chart of the privacy protection method for resisting internal dishonest participants provided by the present invention, such as Figure 1 As shown, the method includes the following: Step 110, obtaining local model parameters to be uploaded by participants in the blockchain-based federated learning BFL framework; Step 120, based on a predetermined commitment function, performing commitment calculation on the local model parameters to obtain a commitment function value; Step 130, sending the local model parameters and their corresponding commitment function values ​​to the bookkeeping rights miner, so that the bookkeeping rights miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework and synthesizes the global model; Step 140 , data privacy enhancement is performed on the global model parameters in the global model based on each participant, and the global model parameters after data privacy enhancement are written into the block in the BFL framework.

[0028] The execution subject of the privacy protection method for resisting internal dishonest participants provided by the present invention can be an electronic device, a component in an electronic device, an integrated circuit, or a chip. The electronic device can be a mobile electronic device or a non-mobile electronic device. Exemplarily, the mobile electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, an ultra-mobile personal computer (UMPC), a netbook, or a personal digital assistant (PDA), etc., and the non-mobile electronic device can be a server, a network attached storage (NAS), or a personal computer (PC), etc., which is not specifically limited by the present invention.

[0029] The following takes the privacy protection method for resisting internal dishonest participants provided by the present invention executed by a computer as an example to explain the technical solution of the present invention in detail.

[0030] In step 110, local model parameters to be uploaded by participants in the blockchain-based federated learning (BFL) framework are obtained.

[0031] In federated learning, each participant (such as different data holders or computing nodes) will first train the model locally. During this process, the participant will use their own data set to train the model and calculate the local model parameters. These local model parameters reflect the optimization direction and degree of the model on the participant's local data set.

[0032] The local model parameters to be uploaded by the participant are obtained. The local model parameters may be calculated local gradient parameters.

[0033] In step 120, based on a predetermined commitment function, a commitment calculation is performed on the local model parameters to obtain a commitment function value.

[0034] In order to ensure the authenticity and integrity of local model parameters while protecting their privacy, the BFL framework introduces a commitment function. A commitment function is a cryptographic tool that allows participants to make some form of "commitment" or "encapsulation" of data without exposing the actual data. Specifically, participants will perform commitment calculations on local model parameters and obtain a commitment function value. This commitment function value will be used to verify the authenticity and integrity of local model parameters in subsequent processes.

[0035] The main purpose of the commitment function is to prevent internal dishonest participants from using invalid local model parameters to maliciously defraud global model parameters. A smart contract based on the commitment function is written into the blockchain. When a dishonest participant denies the responsibility, the commitment is opened according to the contract to announce and punish the dishonest participant.

[0036] In step 130, the local model parameters and their corresponding commitment function values ​​are sent to the accounting rights miner, so that the accounting rights miner verifies the local model parameters and their corresponding commitment function values ​​and writes them into the block in the BFL framework to synthesize the global model.

[0037] In the BFL framework, accounting miners play an important role. They are responsible for verifying the local model parameters of the participants and their corresponding commitment function values ​​and writing them into the block. To achieve this process, the local model parameters and commitment function values ​​are sent to the accounting miners together. After receiving this information, the accounting miners will verify and record it. The verification process includes checking whether the commitment function value matches the local model parameters to ensure the authenticity and integrity of the data. Once the verification is passed, the accounting miners will write this information into the block and broadcast it to other nodes in the network.

[0038] In the BFL framework, the synthesis of the global model is achieved by aggregating the local model parameters of all participants. Specifically, after the accounting rights miners write the local model parameters into the block, they will use these parameters to update the global model. This process usually involves weighted averaging or other forms of aggregation operations on the local model parameters. Through this process, the BFL framework is able to use distributed data and resources to jointly train a global model, thereby improving the generalization ability and performance of the model.

[0039] In step 140, data privacy enhancement is performed on the global model parameters in the global model based on each participant, and the global model parameters after data privacy enhancement are written into the blocks in the BFL framework.

[0040] In the BFL framework, data privacy protection is crucial. In order to enhance the data privacy of global model parameters, the BFL framework can use a variety of technical means to enhance privacy. These technical means include but are not limited to differential privacy, homomorphic encryption, and the privacy protection mechanism of federated learning itself. Through the application of these technical means, the BFL framework can realize the training and updating of the global model while protecting the data privacy of participants.

[0041] Specifically, before writing the global model parameters with enhanced data privacy into the block, the BFL framework performs necessary processing and transformation on these parameters. These processing and transformations are designed to ensure the security and privacy of the global model parameters during transmission and storage.

[0042] Optionally, for local model parameters, a reusable commitment function can be constructed to perform commitment calculation. A reusable commitment function can be constructed based on Pedersen commitment as follows: Assumptions and are two safe prime numbers, let is the modulus of an RSA scheme. yes The only cyclic subgroup of , whose order is , in the group The discrete logarithm problem is difficult: g and h are two elements of the group, and It is difficult to calculate.

[0043] Assumptions is a global hash function: .

[0044] Key generation algorithm : Each participant Generate the public and private key pair of the commitment function: Let and are two safe prime numbers that satisfy ,but .in, is a safety parameter.

[0045] Commitment Generation Algorithm : Each participant Generate commitments for local model parameters for this round according to system requirements: ; i yes The participants will use the local model parameters of this round and the corresponding commitment function Send to the system.

[0046] Commitment Verification Algorithm : Miners with accounting rights (for example, in a blockchain system that adopts the PoS consensus mechanism, miners with the highest equity have accounting rights) have the ability to commit to verification.

[0047] The accounting miners verify the commitment function of each participant: ; ; verify s=tIs it true?

[0048] If the commitment verification is passed, the commitment value will be written to the block storage, and subsequent global model aggregation and other operations will be performed. If the commitment verification fails, it means that the participant has acted dishonestly. In order to prevent information leakage or data being maliciously obtained, the system will disclose the identity of the participant. i and data , anyone can use the public key to verify the commitment, making it impossible to deny the dishonest identity.

[0049] Among them, the hash function H Used to resist counterfeit attacks with multiple coefficients. pk Will not leak sk This ensures that commitment verification cannot be forged and is therefore reusable.

[0050] The commitment function satisfies data hiding: is indistinguishable from any random element value in the group. Data binding property is satisfied: For a polynomial-time adversary, since computing the discrete logarithm The value of is difficult, so it is difficult to find a pair of values satisfy Even if an all-powerful adversary can compute The value of , then we can get ,because is a collision-resistant hash function, so it is difficult to find out another identity The same hash value is calculated, so the identity and commitment cannot be forged.

[0051] The commitment function is mainly used to prevent internal dishonest participants from using invalid local model parameters to maliciously defraud global model parameters. A smart contract based on the commitment function is written into the blockchain. When a dishonest participant uses invalid data and denies it, the commitment is opened according to the contract to announce and punish the dishonest participant.

[0052] It can be understood that based on the constructed commitment function, it is possible to resist dishonest participants in the BFL system, that is, internal dishonest participants cannot cheat the global model parameters of the system by providing invalid local model parameters, and cannot actively leak local model information. Once there is data fraud or active leakage, the system can accurately locate and screen out the dishonest participants, and remove them from the system to ensure the privacy security of the BFL system.

[0053] Optionally, a reusable commitment function can be used to ensure the security of local model parameters generated by each participant, differential privacy and blind signatures can be used to ensure the security of global model parameters generated by aggregation, and the blockchain token reward and punishment mechanism can be combined to reward honest participants and punish and remove dishonest participants. The overall architecture of the solution is as follows: Figure 2 The privacy protection architecture provided by the present invention is shown in the schematic diagram.

[0054] The specific steps of the scheme include: 1) constructing a reusable commitment function, and participants perform commitment calculations before uploading local gradient parameters, and then send the local gradient parameters together with the corresponding commitment function values ​​to miners; 2) Miners with accounting rights verify local gradient parameters and commitment function values, and discard mismatching data; evaluate the validity and contribution of local gradient parameters according to the contract provisions, and discard invalid data; 3) Miners aggregate correct and valid local gradients, and write the commitment function values ​​corresponding to the local model parameters of this round into block storage; 4) When participants have objections to the verification or deny the uploaded local gradient parameters, they can open their commitment functions for public verification. 5) After the accounting right miner synthesizes the global model, it adds different parameter noises to the global model parameters according to the contract requirements and uses the public keys of the participants to perform data blinding operations to generate different global model parameters consistent with the number of participants, and then sends the blinded global model parameters to the participants with the corresponding public keys; 6) The participants sign the blinded information, and it is impossible to interpret the blinded data or forge the signature. The global model parameters after each blind signature are different but are within the allowable range of privacy error; 7) The accounting right miner writes the blinded and signed global model parameters into the block, and sends the corresponding de-blinded global model parameters to the participants for the next round of training; 8) If the global model information is leaked, the system uses the blinding factor to de-blind the signature and make it public, compare it with the leaked information, and publicly verify the identity of the dishonest participant.

[0055] The privacy protection method for resisting internal dishonest participants provided by the present invention performs commitment calculation on the uploaded local model parameters through a predetermined commitment function, and obtains a commitment function value for resisting the situation where internal dishonest participants use invalid local model parameters to maliciously defraud global model parameters. It realizes writing a smart contract based on the commitment function in the blockchain. When a dishonest participant denies, the dishonest participant is determined by opening the commitment according to the contract, thereby improving the resistance and defense capabilities of the blockchain-based federated learning framework against internal dishonest participants and ensuring privacy security.

[0056] In one embodiment, after determining that the accounting rights miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework after verification and synthesizing the global model, it also includes: receiving abnormal local model parameters, the abnormal local model parameters are local model parameters that the participants object to the verification process or deny uploading local model parameters, and the verification process is the verification before the accounting rights miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework; based on the commitment function, the abnormal local model parameters are parsed for public verification.

[0057] When a participant objects to the verification or denies the uploaded local gradient parameters, the committed function value can be opened for public verification.

[0058] It is understandable that the commitment function is mainly used to defend against the situation where dishonest internal participants use invalid local model parameters to maliciously defraud global model parameters. A smart contract based on the commitment function is written into the blockchain. When a dishonest participant denies his or her responsibility, the commitment is opened according to the contract to announce and punish the dishonest participant.

[0059] The commitment function is predetermined, and after the abnormal local model parameters are determined, the abnormal local model parameters can be parsed based on the commitment function for performing commitment calculations to implement a public verification process for parsing the abnormal local model parameters.

[0060] In one embodiment, data privacy is enhanced for global model parameters in the global model based on each participant, including: adding different parameter noises to the global model parameters in the global model based on each participant, and performing data blinding operations on the global model parameters with added parameter noise based on the public key of each participant to generate blinded global model parameters; and signing the blinded global model parameters based on the private key of each participant.

[0061] The performing a data blinding operation on the global model parameters with added parameter noise includes: performing a data blinding operation on the global model parameters with added parameter noise based on a blinding factor.

[0062] Optionally, for global model parameters, blind signatures and differential privacy schemes are constructed to achieve data privacy enhancement process.

[0063] The construction scheme using RSA blind signature algorithm and centralized differential privacy technology is as follows: Assuming that the blockchain adopts the PoS consensus mechanism, the miner with the highest equity has the right to keep accounts and execute privacy-preserving smart contracts. According to the contract, after completing this round of local gradient aggregation, Laplace noise is first added to the global model parameters, and then blinding operations are performed, and finally sent to each participant for blind signature calculation.

[0064] make and are two safe prime numbers, let is the modulus of an RSA scheme.

[0065] Adding noise algorithm : For each participant , to the global model parameters M Add different parameters, but privacy budget Noise within the threshold range , generating each participant Unique global parameters : ; in is the learning rate, is the gradient function, is the dynamic noise adjustment coefficient, is the sensitivity of the function.

[0066] Key generation algorithm : Each participant Generate a public and private key pair for signature: and are two safe prime numbers that satisfy ,but .

[0067] The signature public and private key pairs can be used throughout the training process and do not need to be changed frequently.

[0068] Data Blinding Algorithm : Selecting the blinding factor , using participants The public key , generate the blind data that needs to be signed: ; and will Send to participants .

[0069] Signature Algorithm : Participants Received Then, use your personal private key Sign it: ; and will Sent to the accounting rights miners.

[0070] Due to the message A blinded operation was performed and participants The specific content of the message is not known at the time of signing.

[0071] Signature deblinding algorithm : Miners with bookkeeping rights according to participants The unblinding factor Calculate the original correct signature: ; Signature Verification Algorithm : When the global model parameters in this training When a leak occurs, the system will use the smart contract to sign the leaker’s signature. With public key Public, anyone can verify it: ; Note: Due to the addition of noise with different parameters, each participant obtains slightly different global model parameters, making it easier to determine the source of the leaked information.

[0072] This solution is mainly used to prevent internal dishonest participants from leaking global model parameters. A smart contract based on a blind signature mechanism is written into the blockchain. When a global model leak occurs, the signature of the leaked data is published according to the contract, and the dishonest participants are publicly verified, punished, and removed to ensure the privacy and security of the system. This solution is suitable for a small number of participants. An increase in the number of participants will lead to a decrease in computing efficiency and a linear increase in storage costs.

[0073] In one embodiment, the method further includes: in the case where there is a global model parameter leakage in the BFL framework, obtaining the leaked global model parameter, parsing the leaked global model parameter based on the blinding factor and the public key of the participant, and determining the target participant corresponding to the leaked global model parameter.

[0074] It is understandable that since each global model parameter has a signature executed by the corresponding participant using the public key, after parsing the obtained leaked global model parameter, the identity information of the specific participant can be located based on the signature information.

[0075] After locating the identity information of the participants corresponding to the leaked global model parameters, relevant punishment measures can be implemented to restrain dishonest participants from defrauding data or maliciously leaking information to ensure privacy security.

[0076] In one embodiment, after determining the target participant corresponding to the leaked global model parameter, it also includes: deducting a preset number of token deposits of the target participant, and after determining that the deduction of the token deposit of the target participant is completed, deleting the target participant from the BFL framework, the token deposit is the deposit that the target participant needs to pay before joining the BFL framework.

[0077] A token reward and punishment mechanism is adopted. Participants are required to pay a certain token deposit before joining federated learning. If there is a commitment mismatch and invalid data, or if a dishonest participant is identified by a public signature, a certain number of tokens will be deducted as a punishment. When the participant's deposit is zero, he or she will be removed from the system. At the same time, honest participants will receive token rewards based on their contribution to the global model.

[0078] The privacy protection device for resisting internal dishonest participants provided by the present invention is described below. The privacy protection device for resisting internal dishonest participants described below and the privacy protection method for resisting internal dishonest participants described above can be referenced to each other.

[0079] like Figure 3 As shown, the device comprises: The parameter acquisition module 310 is used to obtain the local model parameters to be uploaded by the participants in the blockchain-based federated learning BFL framework; A commitment calculation module 320, configured to perform commitment calculation on the local model parameters based on a predetermined commitment function to obtain a commitment function value; Aggregation module 330, used to send the local model parameters and their corresponding commitment function values ​​to the bookkeeping right miner, so that the bookkeeping right miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework and synthesizes the global model; The privacy enhancement module 340 is used to perform data privacy enhancement on the global model parameters in the global model based on each participant, and write the global model parameters after data privacy enhancement into the blocks in the BFL framework.

[0080] The privacy protection device for resisting internal dishonest participants provided by the present invention performs commitment calculation on the uploaded local model parameters through a predetermined commitment function, and obtains a commitment function value for resisting the situation where internal dishonest participants use invalid local model parameters to maliciously defraud global model parameters. It realizes writing a smart contract based on the commitment function in the blockchain. When a dishonest participant denies, the dishonest participant is determined by opening the commitment according to the contract, which improves the resistance and defense capabilities of the blockchain-based federated learning framework against internal dishonest participants and ensures privacy security.

[0081] In one embodiment, the aggregation module 330 is specifically configured to: After determining that the accounting right miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework and synthesizes the global model, it also includes: Receive abnormal local model parameters, where the abnormal local model parameters are local model parameters that the participants disagree with during the verification process or deny uploading. The verification process is the verification before the accounting right miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework; Based on the commitment function, the abnormal local model parameters are parsed for public verification.

[0082] In one embodiment, the privacy enhancement module 340 is specifically used to: The data privacy is enhanced for the global model parameters in the global model based on each participant, respectively, including: Adding different parameter noises to the global model parameters in the global model based on each participant respectively, and performing a data blinding operation on the global model parameters with added parameter noise based on the public key of each participant to generate blinded global model parameters; The blinded global model parameters are signed based on the private key of each participant.

[0083] In one embodiment, the privacy enhancement module 340 is further specifically configured to: The performing of a data blinding operation on the global model parameters to which parameter noise is added comprises: Based on the blinding factor, data blinding is performed on the global model parameters with added parameter noise.

[0084] In one embodiment, the privacy enhancement module 340 is further specifically configured to: In the case where there is a global model parameter leakage in the BFL framework, obtaining the leaked global model parameter; Based on the blinding factor and the public key of the participant, the leaked global model parameter is parsed to determine the target participant corresponding to the leaked global model parameter.

[0085] In one embodiment, the privacy enhancement module 340 is further specifically configured to: After determining the target participant corresponding to the leaked global model parameter, the method further includes: Deduct a preset amount of the target participant's token deposit, and after determining that the deduction of the target participant's token deposit is completed, delete the target participant from the BFL framework, wherein the token deposit is the deposit that the target participant needs to pay before joining the BFL framework.

[0086] Figure 4 An example of a physical structure diagram of an electronic device is shown in FIG. Figure 4 As shown, the electronic device may include: a processor 410, a communication interface 420, a memory 430 and a communication bus 440, wherein the processor 410, the communication interface 420 and the memory 430 communicate with each other through the communication bus 440. The processor 410 may call the logic instructions in the memory 430 to execute a privacy protection method for resisting internal dishonest participants, the method comprising: obtaining local model parameters to be uploaded by participants in a blockchain-based federated learning BFL framework; Based on a predetermined commitment function, performing commitment calculation on the local model parameters to obtain a commitment function value; Send the local model parameters and their corresponding commitment function values ​​to the bookkeeping rights miner, so that the bookkeeping rights miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework and synthesizes the global model; Data privacy is enhanced for global model parameters in the global model based on each participant respectively, and the global model parameters after data privacy enhancement are written into the blocks in the BFL framework.

[0087] In addition, the logic instructions in the above-mentioned memory 430 can be implemented in the form of a software functional unit and can be stored in a computer-readable storage medium when it is sold or used as an independent product. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk, etc. Various media that can store program codes.

[0088] On the other hand, the present invention further provides a computer program product, the computer program product includes a computer program, the computer program can be stored on a non-transitory computer-readable storage medium, when the computer program is executed by a processor, the computer can execute the privacy protection method for resisting internal dishonest participants provided by the above methods, the method includes: obtaining local model parameters to be uploaded by participants in a blockchain-based federated learning BFL framework; Based on a predetermined commitment function, performing commitment calculation on the local model parameters to obtain a commitment function value; Send the local model parameters and their corresponding commitment function values ​​to the bookkeeping rights miner, so that the bookkeeping rights miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework and synthesizes the global model; Data privacy is enhanced for global model parameters in the global model based on each participant respectively, and the global model parameters after data privacy enhancement are written into the blocks in the BFL framework.

[0089] In another aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which is implemented when the computer program is executed by a processor to execute the privacy protection method for resisting internal dishonest participants provided by the above methods, the method comprising: obtaining local model parameters to be uploaded by participants in a blockchain-based federated learning BFL framework; Based on a predetermined commitment function, performing commitment calculation on the local model parameters to obtain a commitment function value; Send the local model parameters and their corresponding commitment function values ​​to the bookkeeping rights miner, so that the bookkeeping rights miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework and synthesizes the global model; Data privacy is enhanced for global model parameters in the global model based on each participant respectively, and the global model parameters after data privacy enhancement are written into the blocks in the BFL framework.

[0090] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0091] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a necessary general hardware platform, and of course, can also be implemented by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.

[0092] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A privacy protection method against internal dishonest participants, characterized in that: The method comprises: Obtain the local model parameters to be uploaded by participants in the blockchain-based federated learning BFL framework; Based on a predetermined commitment function, performing commitment calculation on the local model parameters to obtain a commitment function value; Send the local model parameters and their corresponding commitment function values ​​to the bookkeeping rights miner, so that the bookkeeping rights miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework and synthesizes the global model; Data privacy is enhanced for global model parameters in the global model based on each participant respectively, and the global model parameters after data privacy enhancement are written into the blocks in the BFL framework.

2. The privacy protection method against internal dishonest participants according to claim 1, characterized in that: After determining that the accounting right miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework and synthesizes the global model, it also includes: Receive abnormal local model parameters, where the abnormal local model parameters are local model parameters that the participants disagree with during the verification process or deny uploading. The verification process is the verification before the accounting right miner writes the local model parameters and their corresponding commitment function values ​​into the block in the BFL framework; Based on the commitment function, the abnormal local model parameters are parsed for public verification.

3. The privacy protection method against internal dishonest participants according to claim 1, characterized in that: The data privacy enhancement of the global model parameters in the global model based on each participant respectively includes: Adding different parameter noises to the global model parameters in the global model based on each participant respectively, and performing a data blinding operation on the global model parameters with added parameter noise based on the public key of each participant to generate blinded global model parameters; The blinded global model parameters are signed based on the private key of each participant.

4. The privacy protection method against internal dishonest participants according to claim 3, characterized in that: The performing of a data blinding operation on the global model parameters to which parameter noise is added comprises: Based on the blinding factor, data blinding is performed on the global model parameters with added parameter noise.

5. The privacy protection method against internal dishonest participants according to claim 4, characterized in that: Also includes: In the case where there is a global model parameter leakage in the BFL framework, obtaining the leaked global model parameter; Based on the blinding factor and the public key of the participant, the leaked global model parameter is parsed to determine the target participant corresponding to the leaked global model parameter.

6. The privacy protection method against internal dishonest participants according to claim 5, characterized in that: After determining the target participant corresponding to the leaked global model parameter, the method further includes: Deduct a preset amount of the target participant's token deposit, and after determining that the deduction of the target participant's token deposit is completed, delete the target participant from the BFL framework, wherein the token deposit is the deposit that the target participant needs to pay before joining the BFL framework.

7. A privacy protection device for resisting internal dishonest participants, characterized in that: include: The parameter acquisition module is used to obtain the local model parameters to be uploaded by participants in the blockchain-based federated learning BFL framework; A commitment calculation module, used to perform commitment calculation on the local model parameters based on a predetermined commitment function to obtain a commitment function value; An aggregation module, used to send the local model parameters and their corresponding commitment function values ​​to the bookkeeping right miners, so that the bookkeeping right miners write the local model parameters and their corresponding commitment function values ​​into the blocks in the BFL framework and synthesize the global model; The privacy enhancement module is used to perform data privacy enhancement on the global model parameters in the global model based on each participant, and write the global model parameters after data privacy enhancement into the blocks in the BFL framework.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the privacy protection method for resisting internal dishonest participants as described in any one of claims 1 to 6 is implemented.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the privacy protection method for resisting internal dishonest participants as described in any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the privacy protection method for resisting internal dishonest participants as described in any one of claims 1 to 6 is implemented.

Images