CN115392487A - Privacy protection nonlinear federal support vector machine training method and system based on homomorphic encryption - Google Patents
Privacy protection nonlinear federal support vector machine training method and system based on homomorphic encryption Download PDFInfo
- Publication number
- CN115392487A CN115392487A CN202210758385.2A CN202210758385A CN115392487A CN 115392487 A CN115392487 A CN 115392487A CN 202210758385 A CN202210758385 A CN 202210758385A CN 115392487 A CN115392487 A CN 115392487A
- Authority
- CN
- China
- Prior art keywords
- participant
- training
- privacy protection
- local
- homomorphic encryption
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/10—Machine learning using kernel methods, e.g. support vector machines [SVM]
-
- 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
- G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
Abstract
The invention belongs to the technical field of privacy protection machine learning, and particularly relates to a privacy protection nonlinear federal support vector machine training method and a system based on homomorphic encryption, wherein each participant acquires a group key and a random seed for generating an original data mapping function by using a key agreement protocol; generating a mapping function by using the random seeds, and mapping the local data set of each participant as original data to the same high-dimensional space by using the mapping function so as to obtain high-dimensional data corresponding to each participant; and high-dimensional data is used as a training sample, local model parameters are trained by using a privacy protection Federal SVM algorithm, all the local model parameters are aggregated by using homomorphic encryption operation through a server, and a final model is obtained through joint training. The invention can realize the federated SVM joint training with privacy protection, can give consideration to the privacy protection and the training speed of each participant, can relieve the high time overhead brought by introducing a password system, and is convenient for practical scene application.
Description
Technical Field
The invention belongs to the technical field of privacy protection machine learning, and particularly relates to a privacy protection nonlinear federal support vector machine training method and system based on homomorphic encryption.
Background
With the rapid development of big data and artificial intelligence, privacy protection, data security and other data security problems are more and more emphasized, and how to perform multi-source data combined mining becomes a problem to be solved urgently on the premise of effectively protecting user privacy data. Homomorphic encryption is a privacy protection technology which is established on the basis of the public key cryptography theory and ensures the security of a ciphertext through the mathematical problem, and an effective solution is provided for solving the problem of jointly mining the privacy of multi-source data.
The current SVM research on privacy protection is mainly developed from four directions, which are: safe multi-party calculation, data disturbance, safe outsourcing calculation and federal learning. In the research of the SVM based on the safe multi-party computing, the joint modeling of multiple participants is realized through a safe multi-party computing protocol, the privacy of the training data of each participant can be ensured, but the defects of time and large communication overhead exist; the SVM based on data disturbance ensures data privacy by injecting noise into a Gram matrix, but reduces the accuracy of a model; the SVM based on the secure outsourcing computation is that original data are encrypted and uploaded to a cloud by means of the strong computing capacity of a cloud server, model training and prediction are carried out in a ciphertext domain by the cloud, the overall computing complexity is high, and local data need to be encrypted and uploaded to the cloud server. And the federal learning can realize the establishment of a global model only by exchanging model parameters among all participants on the premise of ensuring that data cannot be output locally. Federal learning has superior training efficiency and broader application background than other approaches. Thus, the interest in privacy protection federal learning based on homomorphic encryption has increased in recent years.
However, the currently existing federal learning SVM research also presents many challenges. First, some studies directly apply the federal SVM algorithm to the corresponding scene without using any privacy protection means. Hidden danger of privacy disclosure exists; secondly, the main stream of transverse federated learning is solved through gradient descent, and relevant research through solution of the SMO algorithm is not available for a while because the global optimal coefficient calculated by the dual problem is different from the local optimal coefficient calculated by the local data. Since each party has data of one data subset, solving the dual problem on the data subsets makes it difficult to generate global optimal coefficients. This means that it would be difficult to solve the non-linear federal SVM directly with nuclear skills in a lateral federal learning scenario.
Disclosure of Invention
Therefore, the invention provides a privacy protection nonlinear federal support vector machine training method and system based on homomorphic encryption, which realize federal SVM joint training for privacy protection, can give consideration to privacy protection and training speed of each participant and is convenient for practical scene application.
According to the design scheme provided by the invention, a privacy protection nonlinear federal support vector machine training method based on homomorphic encryption is provided, which comprises the following contents:
each participant acquires a group key and a random seed for generating an original data mapping function by using a key negotiation protocol;
each participant generates a mapping function by using the random seeds, and uses the mapping function to map the local data set of each participant as original data to the same high-dimensional space so as to obtain high-dimensional data corresponding to each participant;
each participant utilizes high-dimensional data as a training sample, local model parameters are trained by using a privacy protection Federal SVM algorithm, each local model parameter is aggregated by utilizing homomorphic encryption operation through a server, and a final model is obtained through joint training.
As the privacy protection nonlinear federal support vector machine training method based on homomorphic encryption, furthermore, each participant adopts a Burmester-Desmedt protocol as a key negotiation protocol to obtain a public and private key pair; broadcasting the public key of the user to other participants; the group key is obtained by each participant participating in intermediate parameters of the key agreement.
As the privacy protection non-linear Federal support vector machine training method based on homomorphic encryption, further, a group key is used as the input of a pseudo-random number generator, and a random seed used for generating an original data mapping function is obtained by the pseudo-random number generator.
As the privacy protection nonlinear federal support vector machine training method based on homomorphic encryption, further, public and private key pairs are generated by using a Burmester-Desmedt protocol, and the selected group is a 2048-bit multiplication cycle group in RFC 3526.
As the privacy protection nonlinear federal support vector machine training method based on homomorphic encryption, a Chacha pseudo-random number generator is further adopted as the pseudo-random number generator.
As the privacy protection nonlinear federal support vector machine training method based on homomorphic encryption, the original data of each participant is mapped to the same high-dimensional space by combining a random Fourier feature algorithm aiming at the acquired random seeds.
As the privacy protection nonlinear federal support vector machine training method based on homomorphic encryption, further, in a random Fourier characteristic algorithm, a random number generator is arranged according to random seeds; constructing a mapping function through a random number generator; and mapping the original data of the local data set of each participant by using a mapping function to obtain correspondingly mapped high-dimensional data.
The privacy protection nonlinear federal support vector machine training method based on homomorphic encryption is characterized in that in the process of carrying out joint training on a model by using a federal SVM algorithm of privacy protection, a server generates initial model parameters and broadcasts the initial model parameters to all participants; each participant carries out iterative training on a local model, and in the iterative training, each participant firstly carries out model training by using a local training sample to obtain local model parameters, homomorphic encryption is carried out on the local model parameters and the local model parameters are uploaded to a server, the server uses homomorphic encryption operation to aggregate the local model parameters of each participant to obtain a global model parameter ciphertext and sends the global model parameter ciphertext to each participant, each participant obtains the local model parameters by decryption according to the received global model parameter ciphertext and enters the next round of model training until the preset maximum iteration round is met.
The privacy protection nonlinear federal support vector machine training method based on homomorphic encryption further comprises the steps of setting participant contribution degrees according to the size of a local data set sample of each participant in joint training, training current round model parameters by using local training samples of each participant in iterative training, updating the global model parameters received in the current round by using the received global model parameters, the local model parameters of the current round training and the participant contribution degrees, carrying out homomorphic encryption on the local model parameters by using the updated global model parameters as the local model parameters of the participants, and sending the local model parameters to a server.
Further, the invention also provides a privacy protection nonlinear federal support vector machine training system based on homomorphic encryption, which comprises: a key negotiation module, a sample construction module and a joint training module, wherein,
the key negotiation module is used for acquiring a group key and a random seed for generating an original data mapping function by each participant through a key negotiation protocol;
the sample construction module is used for generating a mapping function by each participant by using the random seeds and mapping a local data set of each participant as original data to the same high-dimensional space by using the mapping function so as to obtain high-dimensional data corresponding to each participant;
and the joint training module is used for training local model parameters by using the high-dimensional data as training samples and using a privacy protection Federal SVM algorithm, aggregating the local model parameters by using homomorphic encryption operation through the server, and acquiring a final model through joint training.
The invention has the beneficial effects that:
the method includes the steps that a group key and random seeds are generated through negotiation in a plurality of participants, original training data are mapped to the same high-dimensional space through a random Fourier characteristic algorithm, and joint training of a model is conducted through a Federal SVM (support vector machine) training algorithm with privacy protection, so that privacy and training speed are considered at the same time; furthermore, CKKS homomorphic encryption is used in the training stage, privacy of model parameters of each participant and contribution (the size of a local data set) of the participant in the training process is guaranteed, homomorphic multiplication with large time overhead can be avoided, and high time overhead caused by introducing a password system is relieved. Through experimental verification, compared with other Federal SVM training methods, the scheme can obtain lossless model precision while ensuring privacy, and has a deeper application value.
Description of the drawings:
FIG. 1 is a schematic flow of privacy protection non-linear Federal support vector machine training based on homomorphic encryption in the embodiment;
FIG. 2 is a schematic representation of the CKKS homomorphic encryption used in the examples;
FIG. 3 is a schematic illustration of 1 generated data set Circle employed in the embodiment;
FIG. 4 is a schematic of 1 generated data set Moon employed in the example;
fig. 5 is a graph of performance over 4 data sets employed in the example.
The specific implementation mode is as follows:
in order to make the objects, technical solutions and advantages of the present invention clearer and more obvious, the present invention is further described in detail below with reference to the accompanying drawings and technical solutions.
The embodiment of the invention, as shown in fig. 1, provides a privacy protection nonlinear federal support vector machine training method based on homomorphic encryption, which comprises the following contents:
s101, each participant acquires a group key and a random seed for generating an original data mapping function by using a key agreement protocol;
s102, each participant generates a mapping function by using a random seed, and the local data set of each participant is mapped to the same high-dimensional space as original data by using the mapping function so as to obtain high-dimensional data corresponding to each participant;
s103, each participant uses high-dimensional data as a training sample, local model parameters are trained by using a privacy protection Federal SVM algorithm, each local model parameter is aggregated by using homomorphic encryption operation through a server, and a final model is obtained through joint training.
According to the scheme, a group key and a random seed are generated through co-negotiation of a plurality of participants, original training data are mapped to the same high-dimensional space through a random Fourier characteristic algorithm, and joint training of a model is carried out through a privacy protection Federal SVM training algorithm, so that privacy and training speed are considered at the same time, and Federal SVM joint training with privacy protection is achieved.
Further, each participant adopts a Burmester-Desmedt protocol as a key negotiation protocol to carry out random seed security negotiation, and the same random seed is obtained at each participant. The participator broadcasts the self public key to other participators; the group key is obtained by each participant participating in intermediate parameters of the key agreement.
The algorithm process for random seed security negotiation using the Burmester-Desmedt protocol may be designed as follows:
step 1.1: generating public and private keys of Burmester-Desmedt protocol locally at each participant
Step 1.2: will public key
Broadcast to other participants
Step 1.3: the participating party calculates X from the received public key i Wherein X is i ＝Z i+1 /Z i ，
Step 1.4: intermediate result X i Broadcast to other participants
Step 1.5: calculating a group key K i Wherein
Step 1.6: inputting the group key obtained by negotiation into PRNG to obtain random seed
Wherein, the group selected by the Burmester-Desmedt protocol can be a multiplication cycle group of 2048-bit in RFC 3526; in step 1.5, the length of the group key can adopt 2048; in step 1.6, the selected PRNG may be a cryptographically secure ChaCha pseudorandom number generator.
Further, in the embodiment of the present disclosure, for the obtained random seeds, a random fourier feature algorithm is combined to map the original data of each participant into the same high-dimensional space. In the random Fourier characteristic algorithm, a random number generator is arranged according to a random seed; constructing a mapping function through a random number generator; and mapping the original data of the local data set of each participant by using a mapping function to obtain correspondingly mapped high-dimensional data.
The random seeds obtained by negotiation are utilized, and data mapping is carried out by combining a random Fourier characteristic algorithm, and the specific algorithm can be designed to comprise the following steps:
step 2.1: using seed setting obtained by negotiation in step 1 to obtain random number generator random, and setting Gamma value of Gaussian kernel
Step 2.2: generating a standard normal distribution of fixed sequence random numbers from random seeds by a random number generator random
And is uniformly distributed
Step 2.3: sampling ω from the distribution p (ω), uniform (0,2 π) i ,b i ，
Step 2.5: obtaining a mapping function
Wherein
Step 2.6: for data set D i Mapping each data in the data set to obtain a mapped data set X i
In the embodiment of the scheme, further, in the process of carrying out combined training on the model by using a privacy protection Federal SVM algorithm, the server generates initial model parameters and broadcasts the initial model parameters to all participants; each participant carries out iterative training on a local model, and in the iterative training, each participant firstly carries out model training by using a local training sample to obtain local model parameters, homomorphic encryption is carried out on the local model parameters and the local model parameters are uploaded to a server, the server uses homomorphic encryption operation to aggregate the local model parameters of each participant to obtain a global model parameter ciphertext and sends the global model parameter ciphertext to each participant, each participant obtains the local model parameters by decryption according to the received global model parameter ciphertext and enters the next round of model training until the preset maximum iteration round is met. In the joint training, the sample size of each participant local data set is the contribution of the participant in the current round of training, which is marked as n, after the client training is completed, the local model parameter obtained in the current round of training needs to be multiplied by n, and n is placed in the last dimension of the current ciphertext vector, as shown in step 3.1.8, so that the privacy of each participant model parameter and the contribution (the size of the local data set) of the participant in the training process is ensured; in the iterative training, each participant trains the model parameters of the current round by using local training samples, updates the global model parameters received in the current round by using the received global model parameters, the local model parameters of the current round and the contribution degree of the participants, and homomorphically encrypts the local model parameters by using the updated global model parameters as the local model parameters of the participants and sends the local model parameters to the server.
Homomorphic Encryption (HE) Homomorphic Encryption is carried out on original data, and then a specific operation is carried out on a ciphertext to obtain a ciphertext calculation result; after homomorphic decryption is carried out, the obtained plaintext is equivalent to a data result obtained by directly carrying out the same calculation on the original plaintext data. Referring to fig. 2, in the embodiment of the present disclosure, in the joint training, CKKS homomorphic encryption may be used to ensure privacy of model parameters and contribution degrees of each participant in the training process; furthermore, the use of homomorphic multiplication with large time overhead can be avoided by optimizing the calculation process, and the high time overhead caused by introducing a cryptosystem can be relieved.
In the scheme, an algorithm of a Federal SVM (support vector machine) algorithm training model based on homomorphic encryption and privacy protection can comprise a client and a server, and the specific training process of the client and the server can be designed as the following steps:
a client:
step 3.1.1: initializing system parameters: learning rate, ciphertext vector length, number of communication rounds
Step 3.1.2: receiving initialization model parameters omega from a server if the current interaction between the client and the server is the first time 0
Step 3.1.3: otherwise, receiving ciphertext model parameters of the t-th iteration from the server side
Step 3.1.4: decrypting received ciphertext model parameters
Step 3.1.5: calculating to obtain global model parameters
Step 3.1.6: using local data sets D i Training to obtain the updated parameter g of the current round b
Step 3.1.7: furthermore, the utility modelNew model parameter omega t+1 ＝ω t -ηg b
Step 3.1.8: hiding contribution omega t+1 ＝nω t+1 ，
Step 3.1.9: for vector omega t+1 Carry out encryption c i ＝Enc pk (ω t+1 ) And sends to the server
Step 3.1.10: until reaching the communication round number
A server:
step 3.2.1: initializing system parameters: length of cipher text vector, number of communication rounds
Step 3.2.2: receiving cryptographic model parameters c of all participants i
Step 3.2.3: aggregating global model parameters C = Add (C, C) using homomorphic addition i )
Step 3.2.4: sending the aggregated ciphertext C to each participant
Further, based on the above method, an embodiment of the present invention further provides a privacy protection nonlinear federal support vector machine training system based on homomorphic encryption, including: a key negotiation module, a sample construction module and a joint training module, wherein,
the key negotiation module is used for acquiring a group key and a random seed for generating an original data mapping function by each participant through a key negotiation protocol;
the sample construction module is used for generating a mapping function by each participant by using the random seeds and mapping a local data set of each participant as original data to the same high-dimensional space by using the mapping function so as to obtain high-dimensional data corresponding to each participant;
and the joint training module is used for training local model parameters by using the high-dimensional data as training samples and using a privacy protection Federal SVM algorithm, aggregating the local model parameters by using homomorphic encryption operation through the server, and acquiring a final model through joint training.
To verify the validity of the protocol, the following further explanation is made with reference to the test data:
the simulation experiment is carried out by adopting two real data sets Ring and BCD and two generated data sets Moon and Circle, wherein the generated data set Moon used in the simulation is shown in 3,4. Fig. 5 shows the change of the model accuracy of the four data sets in the implementation algorithm and the feadvg algorithm without privacy protection according to the increase of the iteration times. From the results in fig. 5 it can be seen that: the PPNLFedSVM algorithm disclosed by the scheme can protect the privacy of data of participants under the condition that the model precision is not damaged.
Unless specifically stated otherwise, the relative steps, numerical expressions, and values of the components and steps set forth in these embodiments do not limit the scope of the present invention.
Finally, it should be noted that: the above-mentioned embodiments are only specific embodiments of the present invention, which are used for illustrating the technical solutions of the present invention and not for limiting the same, and the protection scope of the present invention is not limited thereto, although the present invention is described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: any person skilled in the art can modify or easily conceive the technical solutions described in the foregoing embodiments or equivalent substitutes for some technical features within the technical scope of the present disclosure; such modifications, changes or substitutions do not depart from the spirit and scope of the embodiments of the present invention, and they should be construed as being included therein. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A privacy protection nonlinear federal support vector machine training method based on homomorphic encryption is characterized by comprising the following contents:
each participant acquires a group key and a random seed for generating an original data mapping function by using a key negotiation protocol;
each participant generates a mapping function by using the random seeds, and uses the mapping function to map the local data set of each participant as original data to the same high-dimensional space so as to obtain high-dimensional data corresponding to each participant;
each participant utilizes high-dimensional data as a training sample, local model parameters are trained by using a privacy protection Federal SVM algorithm, each local model parameter is aggregated by utilizing homomorphic encryption operation through a server, and a final model is obtained through joint training.
2. The privacy protection nonlinear federal support vector machine training method based on homomorphic encryption as claimed in claim 1, wherein each participant adopts a Burmester-Desmedt protocol as a key agreement protocol to obtain a public and private key pair; broadcasting the public key of the user to other participants; the group key is obtained by each participant participating in intermediate parameters of the key agreement.
3. The method for training the privacy protection nonlinear federal Support Vector Machine (SVM) based on homomorphic encryption as claimed in claim 1 or 2, wherein the group key is used as an input of a pseudo random number generator, and the pseudo random number generator is used to obtain a random seed for generating a raw data mapping function.
4. The method for privacy preserving nonlinear federal support vector machine training based on homomorphic encryption as claimed in claim 2, wherein a public and private key pair is generated by using a Burmester-Desmedt protocol, and the selected group is a 2048-bit multiplicative cyclic group in RFC 3526.
5. The privacy protection non-linear federated support vector machine training method based on homomorphic encryption as claimed in claim 3, characterized in that, the pseudo random number generator adopts a ChaCha pseudo random number generator.
6. The privacy protection nonlinear federal support vector machine training method as claimed in claim 1, wherein for the obtained random seeds, the original data of each participant is mapped to the same high-dimensional space by combining a random fourier feature algorithm.
7. The privacy protection nonlinear federal support vector machine training method based on homomorphic encryption as claimed in claim 6, wherein in the random Fourier feature algorithm, a random number generator is set according to random seeds; constructing a mapping function through a random number generator; and mapping the original data of the local data set of each participant by using a mapping function to obtain correspondingly mapped high-dimensional data.
8. The method for training the privacy protection nonlinear federal support vector machine based on homomorphic encryption according to claim 1, characterized in that in the joint training of the model by using the federal SVM algorithm for privacy protection, the server generates initial model parameters and broadcasts the initial model parameters to all participants; each participant carries out iterative training on a local model, and in the iterative training, each participant firstly carries out model training by using a local training sample to obtain local model parameters, homomorphic encryption is carried out on the local model parameters and the local model parameters are uploaded to a server, the server uses homomorphic encryption operation to aggregate the local model parameters of each participant to obtain a global model parameter ciphertext and sends the global model parameter ciphertext to each participant, each participant obtains the local model parameters by decryption according to the received global model parameter ciphertext and enters the next round of model training until the preset maximum iteration round is met.
9. The method for privacy protection nonlinear federal support vector machine training based on homomorphic encryption as claimed in claim 8, wherein in the joint training, the contribution degree of each participant is set according to the size of the local data set sample of each participant, each participant trains the model parameter of the current round by using the local training sample in the iterative training, updates the global model parameter received in the current round by using the received global model parameter, the local model parameter trained in the current round and the contribution degree of each participant, homomorphic encrypts the local model parameter by using the updated global model parameter as the local model parameter of the participant, and sends the local model parameter to the server.
10. A privacy protection non-linear Federal support vector machine training system based on homomorphic encryption is characterized by comprising: a key agreement module, a sample construction module and a joint training module, wherein,
the key negotiation module is used for acquiring a group key and a random seed for generating an original data mapping function by each participant through a key negotiation protocol;
the sample construction module is used for generating a mapping function by each participant by using the random seeds and mapping a local data set of each participant as original data to the same high-dimensional space by using the mapping function so as to obtain high-dimensional data corresponding to each participant;
and the joint training module is used for training local model parameters by using high-dimensional data as training samples and using a privacy protection Federal SVM (support vector machine) algorithm, aggregating the local model parameters by using homomorphic encryption operation through the server and acquiring a final model through joint training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210758385.2A CN115392487A (en) | 2022-06-30 | 2022-06-30 | Privacy protection nonlinear federal support vector machine training method and system based on homomorphic encryption |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210758385.2A CN115392487A (en) | 2022-06-30 | 2022-06-30 | Privacy protection nonlinear federal support vector machine training method and system based on homomorphic encryption |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115392487A true CN115392487A (en) | 2022-11-25 |
Family
ID=84116425
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210758385.2A Pending CN115392487A (en) | 2022-06-30 | 2022-06-30 | Privacy protection nonlinear federal support vector machine training method and system based on homomorphic encryption |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115392487A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115766295A (en) * | 2023-01-05 | 2023-03-07 | 成都墨甲信息科技有限公司 | Industrial internet data secure transmission method, device, equipment and medium |
| CN115865307A (en) * | 2023-02-27 | 2023-03-28 | 蓝象智联(杭州)科技有限公司 | Data point multiplication operation method for federal learning |
| CN116402169A (en) * | 2023-06-09 | 2023-07-07 | 山东浪潮科学研究院有限公司 | Federal modeling verification method, federal modeling verification device, federal modeling verification equipment and storage medium |
| CN117675411A (en) * | 2024-01-31 | 2024-03-08 | 智慧眼科技股份有限公司 | Global model acquisition method and system based on longitudinal XGBoost algorithm |
-
2022
- 2022-06-30 CN CN202210758385.2A patent/CN115392487A/en active Pending
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115766295A (en) * | 2023-01-05 | 2023-03-07 | 成都墨甲信息科技有限公司 | Industrial internet data secure transmission method, device, equipment and medium |
| CN115865307A (en) * | 2023-02-27 | 2023-03-28 | 蓝象智联(杭州)科技有限公司 | Data point multiplication operation method for federal learning |
| CN115865307B (en) * | 2023-02-27 | 2023-05-09 | 蓝象智联(杭州)科技有限公司 | Data point multiplication operation method for federal learning |
| CN116402169A (en) * | 2023-06-09 | 2023-07-07 | 山东浪潮科学研究院有限公司 | Federal modeling verification method, federal modeling verification device, federal modeling verification equipment and storage medium |
| CN116402169B (en) * | 2023-06-09 | 2023-08-15 | 山东浪潮科学研究院有限公司 | Federal modeling verification method, federal modeling verification device, federal modeling verification equipment and storage medium |
| CN117675411A (en) * | 2024-01-31 | 2024-03-08 | 智慧眼科技股份有限公司 | Global model acquisition method and system based on longitudinal XGBoost algorithm |
| CN117675411B (en) * | 2024-01-31 | 2024-04-26 | 智慧眼科技股份有限公司 | Global model acquisition method and system based on longitudinal XGBoost algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109684855B (en) | Joint deep learning training method based on privacy protection technology | |
| Li et al. | Privacy-preserving machine learning with multiple data providers | |
| Hao et al. | Efficient and privacy-enhanced federated learning for industrial artificial intelligence | |
| CN109951443B (en) | Set intersection calculation method and system for privacy protection in cloud environment | |
| CN115392487A (en) | Privacy protection nonlinear federal support vector machine training method and system based on homomorphic encryption | |
| CN110572253A (en) | Method and system for enhancing privacy of federated learning training data | |
| Hahn et al. | Versa: Verifiable secure aggregation for cross-device federated learning | |
| CN105577368A (en) | Two-way privacy protective system and method for inquiring medical diagnostic service | |
| CN107145792A (en) | Multi-user's secret protection data clustering method and system based on ciphertext data | |
| CN112383550B (en) | Dynamic authority access control method based on privacy protection | |
| Wei et al. | Lightweight federated learning for large-scale IoT devices with privacy guarantee | |
| CN104038493B (en) | Bilinear pairing-free cloud storage data security audit method | |
| CN114254386A (en) | Federated learning privacy protection system and method based on hierarchical aggregation and block chain | |
| CN112183767A (en) | Multi-key lower model aggregation federal learning method and related equipment | |
| CN105635135A (en) | Encryption system based on attribute sets and relational predicates and access control method | |
| CN117118617B (en) | Distributed threshold encryption and decryption method based on mode component homomorphism | |
| Yuan et al. | Novel threshold changeable secret sharing schemes based on polynomial interpolation | |
| CN116467736A (en) | Verifiable privacy protection federal learning method and system | |
| Tian et al. | DIVRS: Data integrity verification based on ring signature in cloud storage | |
| Ghavamipour et al. | Federated Synthetic Data Generation with Stronger Security Guarantees | |
| Zhang et al. | Efficient federated learning framework based on multi-key homomorphic encryption | |
| CN115865302A (en) | Multi-party matrix multiplication method with privacy protection attribute | |
| Zhou et al. | A survey of security aggregation | |
| CN112118257B (en) | Security-enhanced keyword search method based on public key encryption | |
| Agarwal | Cryptographic key generation using burning ship fractal |
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 |