CN115238288A - Safety processing method for industrial internet data - Google Patents

Safety processing method for industrial internet data Download PDF

Info

Publication number
CN115238288A
CN115238288A CN202210880056.5A CN202210880056A CN115238288A CN 115238288 A CN115238288 A CN 115238288A CN 202210880056 A CN202210880056 A CN 202210880056A CN 115238288 A CN115238288 A CN 115238288A
Authority
CN
China
Prior art keywords
model parameters
factory
model
ipfs
training
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210880056.5A
Other languages
Chinese (zh)
Inventor
王汝言
景忠源
吴大鹏
杨志刚
张普宁
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Chongqing University of Post and Telecommunications
Original Assignee
Chongqing University of Post and Telecommunications
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Chongqing University of Post and Telecommunications filed Critical Chongqing University of Post and Telecommunications
Priority to CN202210880056.5A priority Critical patent/CN115238288A/en
Publication of CN115238288A publication Critical patent/CN115238288A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/60—Protecting data
    • G06F21/602—Providing cryptographic facilities or services
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/60—Protecting data
    • G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
    • G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/60—Protecting data
    • G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
    • G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
    • G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G06N3/084—Backpropagation, e.g. using gradient descent

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Health & Medical Sciences (AREA)
  • Bioethics (AREA)
  • Computer Security & Cryptography (AREA)
  • Computer Hardware Design (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Evolutionary Computation (AREA)
  • Data Mining & Analysis (AREA)
  • Mathematical Physics (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Molecular Biology (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Storage Device Security (AREA)

Abstract

The invention relates to a safe processing method of industrial internet data, belonging to the field of industrial internet data processing. The method comprises the following steps: the factory and the cooperative party generate a pair of keys by an ElGamal encryption algorithm; initializing model parameters by a contractor, and uploading the model parameters and a public key on a block chain; a factory downloads an initial model and a public key of a cooperator from a block chain, trains the model and extracts model parameters by using a differential privacy algorithm; factory encryption model parameters are stored in an IPFS to obtain a hash value; the encrypted hash value and a factory public key are added to the block chain; the cooperative party searches model parameters in the IPFS through a hash value to train a global model, encrypts the model parameters by using SK and stores the model parameters in the IPFS, encrypts the hash value of the IPFS, encrypts the SK by using a factory public key, and adds a result to a block chain; the factory receives the current global model parameters for updating. The invention solves the privacy and trust problems of machine learning in an industrial Internet system.

Description

Safety processing method for industrial internet data
Technical Field
The invention belongs to the technical field of industrial internet data security processing, and relates to a security processing method of industrial internet data.
Background
In recent years, with the large-scale deployment and application of intelligent sensing equipment, massive data is generated, and the data contains information of various fields of production and life and becomes an important production element. The industrial internet is used as a product of deep integration of a new generation of information technology and manufacturing industry, a novel industrial production manufacturing and service system with complete elements, a complete industrial chain and a complete value chain which are completely connected is constructed through the complete interconnection of people, machines and objects, the method is a realization way of digital transformation, and is key force for realizing the transformation of new and old kinetic energy. The industrial internet is in a new stage of data scale explosion, calculation capacity sharp increase and algorithm performance continuous improvement, and intelligent technologies such as deep learning, reinforcement learning and federal learning become important supports for improving the comprehensive service performance of the industrial internet.
The industrial production data security is high, the problem of data privacy disclosure exists in the current common centralized model training method, the dual requirements of data sharing and privacy protection are difficult to be considered, the novel privacy protection calculation paradigm is needed to complete multiparty combined modeling, and the flexible and intelligent adaptation method is needed to guarantee the efficient operation of the novel calculation paradigm. And the safety is the guarantee of the industrial Internet. In the future, the construction of industrial internet as a new infrastructure will be pushed to a wider range, a deeper degree and a higher level. Facing new technologies and new potentials of security protection, the security ecology of the industrial internet continuously faces new challenges, and therefore, a security processing method facing the industrial internet data is urgently needed to improve the capability of coping with security risks and promote the prosperity and development of the industrial internet.
Disclosure of Invention
In view of this, the present invention aims to provide a method for securely processing industrial internet data, which combines federal learning, differential privacy, elGamal encryption, etc. to create a secure industrial internet in cooperation, so as to better enable the industrial internet and promote industrial upgrade, aiming at the problem that it is difficult to consider dual requirements of industrial data sharing and privacy protection.
In order to achieve the purpose, the invention provides the following technical scheme:
a safe processing method of industrial internet data specifically comprises the following steps:
s1: the factory and the cooperative party generate a pair of secret keys; the collaborator then initializes the model parameters w 0 And the model parameter w is calculated 0 Uploading the public key of the block chain to the block chain;
s2: each factory downloads initialization model parameters w from the blockchain 0 The public key of the cooperative party is used for training a deep neural network model by using a certain number of local data samples; processing the model by using a differential privacy algorithm after training is finished to generate a local differential privacy machine learning model, and then extracting model parameters in a factory;
s3: the factory encrypts the model parameters by using a public key of a cooperator, then stores the encrypted model parameters in the IPFS, and obtains a unique IPFS hash value; then, the IPFS hash value is encrypted by using the public key of the collaborator, and is packed with the factory public key by means of an intelligent contract and added to the block chain; the factory informs the cooperative party that the local model training of the current round is finished;
the encrypted model parameters are stored in the IPFS, which is a peer-to-peer network protocol for file storage, using content-based addressing, rather than location-based. Files in the IPFS network are assigned a hash value, which is similar to a "fingerprint" and is computed from the contents of the file.
S4: the cooperative party decrypts by using a private key, encrypts an IPFS hash value through a public key of the cooperative party, retrieves the IPFS hash value in the IPFS to obtain a corresponding encrypted model parameter, trains a global model after decryption and extracts the model parameter;
s5: encrypting the model parameters by using SK (symmetric key produced by a cooperative party) and storing the model parameters in IPFS, then giving an IPFS hash value, encrypting the IPFS hash value by using SK, simultaneously encrypting SK by using public keys of various factories, adding the encryption result to a block chain (the multi-layer encryption method ensures the safety of data), and informing the factories that the aggregation is finished in the current round;
s6: after a factory receives a notice that a current federal period is finished from a cooperative party, retrieving the encrypted global model parameters, decrypting by using a private key to obtain SK, decrypting the encrypted IPFS hash value by using the SK, and retrieving the model parameters by using the IPFS hash value; the plant uses the updated model parameters for local training, begins the next joint training round, and repeats all steps in a predefined number of joint rounds.
Further, in step S1, the factory and the cooperator generate a pair of keys by using ElGamal encryption algorithm; the ElGamal encryption algorithm is an asymmetric encryption algorithm based on Diffie-Hellman key exchange, and the EIGamal encryption algorithm is based on the following principle: solving discrete logarithms is difficult, and the inverse of it can be efficiently calculated using the square multiplication method. In the corresponding group G, the exponential function is a one-way function.
Further, step S2 specifically includes: the factory trains a deep neural network by using a specific number of local data samples; obtaining the latest model parameters from the server, from 1 to the batch number
Figure BDA0003763832890000021
Batch number b, calculating a batch gradient g k (b) And locally updating model parameters:
Figure BDA0003763832890000022
wherein eta represents the learning rate, w t Representing the current model parameters, D k Representing the dataset owned by the kth collaborator, M representing the size of the mini-batch that the specified client update is using;
in the training process, the neural network training based on the difference privacy is realized in the SGD calculation, and the experience loss function is minimized
Figure BDA0003763832890000023
To train model parameters; at SAt each step of GD, gradients are calculated, and for a subset of samples, the norm l of each gradient is clipped 2 And calculating an average value, adding noise for protecting privacy, and stepping a step in the opposite direction of the average noise gradient to perform gradient descent backward propagation to finish training and finally outputting a model.
Further, step S4 specifically includes: assuming that there are K participants (i.e., plants) in a Federal learning System, D k Represents a data set owned by the kth participant, P k An index set representing data points located at participant k; let n k Is represented by P k Assuming that there is a kth participant with n k And (3) when the data points total K participants, the collaborator aggregates the received model parameters, namely performs weighted average on the received model parameters:
Figure BDA0003763832890000031
updating model parameters:
Figure BDA0003763832890000032
wherein, therein
Figure BDA0003763832890000033
Represents the pair of samples (x) over a given model parameter w i ,y i ) Making a prediction of the resulting loss, x i And y i Respectively, the ith training data point and its associated label. η represents the learning rate, and n represents the number of training data; the cooperator checks whether the loss function converges or reaches the maximum training round; if yes, the cooperative party sends signals to all the participants, and all the participants stop model training.
Further, in step S6, the updated model parameters of the factory are encrypted, stored and uploaded.
The invention has the beneficial effects that: the invention enhances the privacy and credibility of industrial internet data by combining differential privacy, federal learning, block chains and intelligent contracts. The capability of coping with security risks is improved, and the prosperity and development of the industrial Internet are promoted.
Additional advantages, objects, and features of the invention will be set forth in part in the description which follows and in part will become apparent to those having ordinary skill in the art upon examination of the following or may be learned from practice of the invention. The objectives and other advantages of the invention will be realized and attained by the structure particularly pointed out in the written description and claims thereof.
Drawings
For a better understanding of the objects, aspects and advantages of the present invention, reference will now be made to the following detailed description taken in conjunction with the accompanying drawings in which:
FIG. 1 is an architecture diagram of a secure processing method for industrial Internet data according to the present invention;
FIG. 2 is a flow chart of a factory data processing process of the industrial Internet data security processing method of the present invention;
fig. 3 is a flow chart of a data processing process of a cooperator of the industrial internet data security processing method of the present invention.
Detailed Description
The following embodiments of the present invention are provided by way of specific examples, and other advantages and effects of the present invention will be readily apparent to those skilled in the art from the disclosure herein. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention in a schematic way, and the features in the following embodiments and examples may be combined with each other without conflict.
Referring to fig. 1 to 3, fig. 1 is an architecture diagram of an industrial internet data security processing method, fig. 2 is a factory data processing process of the industrial internet data security processing method, and fig. 3 is a cooperator data processing process of the industrial internet data security processing method.
The method solves the privacy and trust problems of machine learning in an industrial internet system, and comprises entities such as factories, collaborators, IPFS (Internet protocol file system), block chains and the like and data transmission among the entities. The method specifically comprises the following steps:
step 1: the factory and the cooperator firstly generate a pair of secret keys by an ElGamal encryption algorithm, wherein the ElGamal encryption algorithm is an asymmetric encryption algorithm based on the key exchange of Diffie-Hellman, and the EIGamal encryption algorithm is based on the following principle: solving discrete logarithms is difficult, and the inverse of it can be efficiently calculated using the square multiplication method. In the corresponding group G, the exponential function is a one-way function.
Step 2: downloading initialization model w from blockchain per factory 0 And a public key of the collaborator, training a deep neural network using a certain number of local data samples. And after the training is finished, processing the model by using a differential privacy algorithm to generate a local differential privacy machine learning model.
The factory trains a deep neural network using a specified number of local data samples. Obtaining the latest model parameters from the server, from 1 to the batch number
Figure BDA0003763832890000041
Batch number b, calculate batch gradient g k (b) And locally updating model parameters:
Figure BDA0003763832890000042
η represents the learning rate.
In the training process, the neural network training based on the difference privacy is realized in the SGD calculation, and the experience loss function is minimized
Figure BDA0003763832890000043
To train the model parameters. At each step of the SGD, gradients are computed, and for a subset of samples, the norm l of each gradient is clipped 2 Calculating an average value, adding noise for privacy protection, and stepping down the average noise gradient in the opposite directionAnd (5) finishing training by back propagation and finally outputting the model.
Federated learning enables deep learning models to be trained with the help of a central server while keeping training data distributed at the client. The client only needs to submit the local gradient of the client to the cloud server during model training, and therefore the risk that the privacy of user data is revealed is avoided to a certain extent. Differential privacy is an efficient privacy protection technology, can quantify the data privacy protection degree, sets a proper privacy budget, and can achieve good balance between data availability and privacy protection. The differential privacy is commonly used for privacy protection of training data of an artificial intelligence model, and can provide privacy guarantee for federal study.
On the premise that user data is not local, a common model is established through parameter exchange and optimization under an encryption mechanism or a disturbance mechanism. The performance of this common model is close to aggregating each party's data into a trained model. The data joint modeling scheme does not reveal user privacy and accords with the principle of data safety protection.
And step 3: the factory encrypts the model parameters by using the public key of the collaborator, then stores the encrypted model parameters in the IPFS, and obtains a unique IPFS hash value. The IPFS hash value is then encrypted, again using the public key of the contractor, and packed with the factory public key by means of the smart contract to be added to the blockchain. And the factory informs the cooperative party that the local model training of the current round is finished.
The encrypted model parameters are stored in the IPFS, which is a peer-to-peer network protocol for file storage, using content-based addressing, rather than location-based. Files in the IPFS network are assigned a hash value, which is similar to a "fingerprint" and is computed from the contents of the file.
And 4, step 4: and the cooperative party decrypts the IPFS hash value encrypted by the public key by using the private key, retrieves the IPFS hash value in the IPFS to obtain corresponding encrypted model parameters, trains the global model after decryption and extracts the model parameters.
And the cooperative party firstly decrypts to obtain the IPFS hash value, retrieves to obtain the corresponding encryption model parameter by using the IPFS hash value, and continuously decrypts by using the private key to obtain the updated parameter of the local model.
Assuming that there are K participants in a Federal learning System, D k Representing a data set, P, owned by the Kth participant k An index set representing data points located at customer k. Let n k Represents P k Assuming that there is a kth participant with n k Data points, total K participants
Figure BDA0003763832890000051
The collaborator aggregates the received model parameters, namely performs weighted average on the received model parameters:
Figure BDA0003763832890000052
where η represents the learning rate, the cooperator checks whether the loss function converges or reaches the maximum training round. If yes, the cooperative party sends signals to all the participants, and all the participants stop model training.
And 5: the method comprises the steps of encrypting model parameters by using SK (symmetric key produced by a cooperative party) and storing the model parameters in IPFS, then giving an IPFS hash value, encrypting the IPFS hash value by using the SK, encrypting the SK by using a factory public key, adding an encryption result to a block chain, and informing a factory that the current aggregation is finished.
Step 6: and (3) local training is carried out by using the updated model parameters in a factory, firstly, the encrypted global model parameters are retrieved, the SK is obtained by decryption through a private key, the IPFS hash value is encrypted by decryption through the SK, the model parameters are retrieved through the IPFS hash value, and model training is carried out in the same mode as the step 2 after the model parameters are obtained. And then, encrypting, storing and uploading the model parameters.
Finally, although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that various changes and modifications may be made therein without departing from the spirit and scope of the invention as defined by the appended claims.

Claims (5)

1. A safe processing method of industrial internet data is characterized by comprising the following steps:
s1: the factory and the cooperative party generate a pair of secret keys; the collaborator then initializes the model parameters w 0 And the model parameter w is calculated 0 Uploading the public key of the block chain to the block chain;
s2: each factory downloads initialization model parameters w from the blockchain 0 The public key of the cooperative party is used for training a deep neural network model by using a specific number of local data samples; processing the model by using a differential privacy algorithm after training is finished to generate a local differential privacy machine learning model, and then extracting model parameters in a factory;
s3: the factory encrypts the model parameters by using a public key of a cooperator, then stores the encrypted model parameters in the IPFS, and obtains a unique IPFS hash value; then, the IPFS hash value is encrypted by using the public key of the collaborator, and is packed with the factory public key by means of an intelligent contract and added to the block chain; the factory informs the cooperative party that the local model training of the current round is finished;
s4: the cooperative party decrypts by using a private key, encrypts an IPFS hash value through a public key of the cooperative party, retrieves the IPFS hash value in the IPFS to obtain a corresponding encrypted model parameter, trains a global model after decryption and extracts the model parameter;
s5: encrypting the model parameters by using a symmetric key SK produced by a cooperative party and storing the model parameters in an IPFS, then giving an IPFS hash value, encrypting the IPFS hash value by using the SK, simultaneously encrypting the SK by using a public key of each factory, adding an encryption result to a block chain, and informing the factory that the current round of aggregation is finished;
s6: after a factory receives a notice that a current federal period is finished from a cooperative party, retrieving encrypted global model parameters, decrypting by using a private key to obtain SK, decrypting an encrypted IPFS hash value by using the SK, and retrieving the model parameters by using the IPFS hash value; the plant uses the updated model parameters for local training, starts the next joint training round, and repeats all steps in a predefined number of joint rounds.
2. The method for securely processing industrial internet data according to claim 1, wherein in step S1, the factory and the cooperator generate a pair of keys by using ElGamal encryption algorithm; the ElGamal encryption algorithm is an asymmetric encryption algorithm based on the diffie-hellman key exchange.
3. The method for safely processing industrial internet data according to claim 1, wherein the step S2 specifically comprises: the factory trains a deep neural network by using a specific number of local data samples; obtaining the latest model parameters from the server, from 1 to the batch number
Figure FDA0003763832880000011
Batch number b, calculate batch gradient g k (b) And locally updating model parameters:
Figure FDA0003763832880000012
wherein eta represents the learning rate, w t Representing the current model parameters, D k Representing the dataset owned by the kth collaborator, M representing the size of the mini-batch that the specified client update is using;
in the training process, the neural network training based on the difference privacy is realized in the SGD calculation, and the experience loss function is minimized
Figure FDA0003763832880000013
To train model parameters; at each step of the SGD, gradients are computed, and for a subset of samples, the norm of each gradient is clipped
Figure FDA0003763832880000014
Calculating the averageAnd (4) stepping to the opposite direction of the average noise gradient to perform gradient descent back propagation to finish training and finally outputting the model.
4. The method for safely processing industrial internet data according to claim 3, wherein the step S4 specifically comprises: assuming that there are K participants in a Federal learning System, D k Representing a data set, P, owned by the kth participant k An index set representing data points located at participant k; let n k Is represented by P k Assuming that there is a kth participant with n k At data points, when there are K participants in total, the collaborator aggregates the received model parameters, that is, performs weighted average on the received model parameters:
Figure FDA0003763832880000021
updating model parameters:
Figure FDA0003763832880000022
wherein, therein
Figure FDA0003763832880000023
Representing the pair of samples (x) over a given model parameter w i ,y i ) Making a prediction of the resulting loss, x i And y i Respectively representing the ith training data point and the related label; η represents the learning rate, and n represents the number of training data; the cooperator checks whether the loss function converges or reaches the maximum training round; if yes, the cooperative party sends signals to all the participators, and all the participators stop model training.
5. The method for securely processing industrial internet data according to claim 1, wherein in step S6, the updated model parameters of the factory are encrypted, stored and uploaded.
CN202210880056.5A 2022-07-25 2022-07-25 Safety processing method for industrial internet data Pending CN115238288A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210880056.5A CN115238288A (en) 2022-07-25 2022-07-25 Safety processing method for industrial internet data

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210880056.5A CN115238288A (en) 2022-07-25 2022-07-25 Safety processing method for industrial internet data

Publications (1)

Publication Number Publication Date
CN115238288A true CN115238288A (en) 2022-10-25

Family

ID=83674636

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210880056.5A Pending CN115238288A (en) 2022-07-25 2022-07-25 Safety processing method for industrial internet data

Country Status (1)

Country Link
CN (1) CN115238288A (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115619947A (en) * 2022-12-19 2023-01-17 江西农业大学 Three-dimensional modeling cooperation method and system based on block chain
CN115865487A (en) * 2022-11-30 2023-03-28 四川启睿克科技有限公司 Abnormal behavior analysis method and device with privacy protection function

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115865487A (en) * 2022-11-30 2023-03-28 四川启睿克科技有限公司 Abnormal behavior analysis method and device with privacy protection function
CN115865487B (en) * 2022-11-30 2024-06-04 四川启睿克科技有限公司 Abnormal behavior analysis method and device with privacy protection function
CN115619947A (en) * 2022-12-19 2023-01-17 江西农业大学 Three-dimensional modeling cooperation method and system based on block chain
CN115619947B (en) * 2022-12-19 2023-12-26 江西农业大学 Three-dimensional modeling cooperation method and system based on blockchain

Similar Documents

Publication Publication Date Title
Yang et al. A quasi-newton method based vertical federated learning framework for logistic regression
CN111600707B (en) Decentralized federal machine learning method under privacy protection
Wang et al. A privacy-enhanced retrieval technology for the cloud-assisted internet of things
CN115238288A (en) Safety processing method for industrial internet data
CN113434873A (en) Federal learning privacy protection method based on homomorphic encryption
Jiang et al. Flashe: Additively symmetric homomorphic encryption for cross-silo federated learning
WO2021259357A1 (en) Privacy-preserving asynchronous federated learning for vertical partitioned data
CN110084377A (en) Method and apparatus for constructing decision tree
CN113688999B (en) Training method of transverse federated xgboost decision tree
Zhang et al. Achieving efficient and privacy-preserving neural network training and prediction in cloud environments
CN114547643B (en) Linear regression longitudinal federal learning method based on homomorphic encryption
CN113420232B (en) Privacy protection-oriented federated recommendation method for neural network of graph
CN113515760B (en) Horizontal federal learning method, apparatus, computer device, and storage medium
Ou et al. A homomorphic-encryption-based vertical federated learning scheme for rick management
CN112347500B (en) Machine learning method, device, system, equipment and storage medium of distributed system
Ruttor et al. Dynamics of neural cryptography
CN111460478B (en) Privacy protection method for collaborative deep learning model training
CN113947211A (en) Federal learning model training method and device, electronic equipment and storage medium
CN111143471A (en) Ciphertext retrieval method based on block chain
CN113535808B (en) Key value pair model safety training and reasoning method based on safety multi-party calculation
CN111431898A (en) Multi-attribute mechanism attribute-based encryption method with search function for cloud-assisted Internet of things
CN113065143A (en) Block chain based secure sharing of industrial data
CN108259517B (en) Encryption method for realizing key isolation attribute of ciphertext strategy
Kurupathi et al. Survey on federated learning towards privacy preserving AI
CN116523074A (en) Dynamic fairness privacy protection federal deep learning method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination