WO2023092428A1

WO2023092428A1 - Data value evaluation method and related product

Info

Publication number: WO2023092428A1
Application number: PCT/CN2021/133297
Authority: WO
Inventors: 吕清
Original assignee: 华为技术有限公司
Priority date: 2021-11-25
Filing date: 2021-11-25
Publication date: 2023-06-01

Abstract

Disclosed in embodiments of the present application are a data value evaluation method and a related product. The method comprises: a data evaluation device sending first model information to a first data processing device, the first model information representing a second encryption service model obtained by training a first encryption service model by a second data processing device using first local data thereof, and a second service model obtained by decrypting the second encryption service model being the same as a service model after using the first local data to train the first service model; obtaining a first performance index according to first index information from the first data processing device; and evaluating, according to the first performance index and a second performance index, the value of the first local data for training the first service model, the second performance index representing the performance of the first service model. In the embodiments of the present application, in the premise that local data of a seller is completely invisible to a buyer, the buyer is helped to train the first service model, so as to evaluate the value of the first local data for training the first service model.

Description

Data value assessment methods and related products

technical field

The embodiments of the present application relate to the field of data value evaluation, and in particular, to a data value evaluation method and related products.

Background technique

At present, a large number of historical and real-time operation data of new energy vehicles are stored on major automobile OEMs and government supervision platforms at all levels. Using these data to develop relevant data business models, such as vehicle fault diagnosis model, power battery thermal runaway warning model, and power battery residual value evaluation model, will greatly promote the safety and performance efficiency of new energy vehicles.

At present, although automobile OEMs and government regulatory platforms at all levels store a large amount of new energy vehicle operation data, the quality of these data is generally not high, and there are a lot of missing and wrong problems. For the development of data business models, different vehicle data have different values. For example, using the running data of vehicle 1 to train a certain data service model can obviously improve the performance of the data service model, but using the data of vehicle 2 to train the data service model can hardly improve the performance of the data service model. For the development of the data business model, it is necessary to accurately evaluate the value of different vehicle data, so that the vehicle data with greater value can be used to train the data business model with better performance. Therefore, it is necessary to study how to accurately evaluate the value of vehicle data for developing data business models.

Contents of the invention

The embodiment of the present application discloses a data value evaluation method and related products, which can accurately evaluate the value of vehicle data for training data business models.

In the first aspect, the embodiment of the present application provides a data value evaluation method, the method includes: the data evaluation device sends the first model information to the first data processing device; the first model information indicates that the second data processing device uses its first model information A second encrypted business model obtained by training the first encrypted business model with local data; the second encrypted business model obtained by decrypting the second encrypted business model is the same as the business model after using the first local data to train the first business model, The first business model is an unencrypted business model corresponding to the first encrypted business model; the data evaluation device obtains a first performance index according to the first index information from the first data processing device; the The first performance index represents the performance of the business model after using the first local data to train the first business model, and the first index information is that the first data processing device uses the first model information and the The test data corresponding to the second encrypted business model is obtained; the data evaluation device evaluates the value of the first local data for training the first business model according to the first performance index and the second performance index; The second performance indicator characterizes the performance of the first business model.

In the embodiment of the present application, the data evaluation device evaluates the value of the first local data for training the first business model according to the first performance index and the second performance index; it can accurately evaluate the value of the first local data for training the first business model . The data evaluation device sends the first model information to the first data processing device. Since the first model information represents the second encrypted service model, the first data processing device can only obtain the second encrypted service model according to the first model information, but cannot obtain the first local data of the second data processing device. That is to say, the first local data of the second data processing device does not leave the local, and the use of the first local data is limited to training the first encrypted service model. Therefore, in the embodiment of the present application, on the premise that the first local data of the second data processing device (seller) is completely invisible to the first data processing device (buyer), the first data processing device is helped to train the first business model, Then evaluate the value of the first local data for training the first business model. In addition, the second data processing device uses its first local data to train the first encrypted business model, so that the first data processing device will not expose its business model and model training method.

In a possible implementation manner, the first index information is obtained by the first data processing device testing the second encrypted service model by using encrypted test data, and the encrypted test data is obtained by encrypting the test data.

In this implementation, the first index information is obtained by the first data processing device using the encrypted test data to test the second encrypted business model, so that the data evaluation device decrypts the first index information to obtain the first performance index.

In a possible implementation manner, before the data evaluation device sends the first model information to the first data processing device, the method further includes: the data evaluation device receives the second model from the second data processing device information; the data evaluation device obtains the first model information according to the second model information.

In this implementation manner, the data evaluation device obtains the first model information according to the second model information; leakage of the first model information can be avoided, and data security can be improved.

In a possible implementation manner, before the data evaluation device receives the second model information from the second data processing device, the method further includes: the data evaluation device sending the second data processing device Send third model information; the third model information is used to train the first encrypted service model.

In this implementation, the data evaluation device sends the third model information to the second data processing device, which can not only enable the second data processing device to train the first encrypted business model, but also avoid the model training method or program of the first data processing device .

In a possible implementation manner, before the data evaluation device sends the first model information to the first data processing device, the method further includes: the data evaluation device sends the first encrypted information to the second data processing device ; The first encrypted information is used by the second data processing device to encrypt the first local data, and the second encrypted business model is used by the second data processing device to use the first encrypted data and the third model The information is obtained by training the first encrypted business model, and the first encrypted data is obtained by encrypting the first local data with the first encrypted information.

In this implementation, the data evaluation device sends the first encryption information to the second data processing device, so that the second data processing device uses the first encryption information to encrypt its first local data to obtain the first encryption information for training the first encrypted business model. - encrypted data.

In a possible implementation manner, before the data evaluation device sends the third model information to the second data processing device, the method further includes: the data evaluation device receives information from the first data processing device The third model information of .

In a possible implementation manner, before the data evaluation device receives the third model information from the first data processing device, the method further includes: the data evaluation device sends the first data The processing device sends first encrypted information; the first encrypted information is used by the first data processing device to encrypt fourth model information to obtain the third model information, and the fourth model information is used to train the first business model.

In this implementation, the data evaluation device sends the first encrypted information to the first data processing device, so that the first data processing device uses the first encrypted information to encrypt the fourth model information to obtain the third model information; the exposure of the fourth model can be avoided information.

In a possible implementation manner, the obtaining the first performance index by the data evaluation device according to the first index information from the first data processing device includes: the data evaluation device decrypting the first index information processing to obtain the first performance index.

In this implementation manner, the data evaluation device performs decryption processing on the first index information, so as to accurately obtain the first performance index.

In a possible implementation manner, after the data evaluation device evaluates the value of the first local data for training the first business model according to the first performance index and the second performance index, the method further includes: The data evaluation device sends a decryption private key to the first data processing device; the decryption private key is used by the first data processing device to decrypt the second encrypted business model.

In this implementation, the data evaluation device sends a decryption private key to the first data processing device, so that the first data processing device decrypts the second encrypted business model to obtain required model parameters.

In a possible implementation manner, after the data evaluation device evaluates the value of the first local data for training the first business model according to the first performance index and the second performance index, the method further includes: The data evaluation device receives third model information from the first data processing device; the third model information is used to train the first encrypted business model; the data evaluation device sends the second data processing device Send the third model information.

In this implementation, the data evaluation device sends the third model information to the second data processing device, so that the second data processing device uses the third model information to assist the first data processing device in model training.

In a second aspect, the present application provides another data value evaluation method, including: the first data processing device receives the first model information from the data evaluation device; the first model information indicates that the second data processing device utilizes its first The second encrypted business model obtained by training the first encrypted business model with local data; the second encrypted business model obtained by decrypting the second encrypted business model is the same as the business model after using the first local data to train the first business model, so The first business model is an unencrypted business model corresponding to the first encrypted business model; the first data processing device sends the test data corresponding to the first model information and the second encrypted business model to the The data evaluation device sends first index information; the first index information is used by the data evaluation device to obtain a first performance index; the first performance index indicates that the first business model is trained using the first local data The performance of the subsequent business model.

In the embodiment of the present application, the first data processing device sends the first index information to the data evaluation device according to the first model information and the test data corresponding to the second encrypted business model; the data evaluation device can accurately evaluate the first local data for The value of training the first business model. The first data processing device receives the first model information from the data evaluation device, and the first data processing device cannot obtain the first local data of the second data processing device according to the first model information, which can prevent the first local data from being exposed to A first data processing device.

In a possible implementation manner, the first data processing device sends the first indicator information to the data evaluation device according to the first model information and the test data corresponding to the second encrypted service model, including: the The first data processing device uses the encrypted test data to test the second encrypted business model to obtain the first index information; the encrypted test data is obtained by encrypting the test data with the first encrypted information; the first data processing The means sends the first indicator information to the data evaluation means.

In this implementation manner, the first data processing device uses the encrypted test data to test the second encrypted service model, and can test to obtain the first index information for obtaining the first performance index.

In a possible implementation manner, before the first data processing device receives the first model information from the data evaluation device, the method further includes: the first data processing device sends a third model information to the data evaluation device Model information, the third model information is used to train the first encrypted service model.

In this implementation, the first data processing device sends the third model information to the data evaluation device, which not only enables the second data processing device to train the first encrypted business model, but also prevents its model training method or program from being exposed.

In a possible implementation manner, before the first data processing device sends the third model information to the data evaluation device, the method further includes: the first data processing device receives from the data evaluation device first encrypted information; the first data processing device uses the first encrypted information to encrypt fourth model information to obtain the third model information, and the fourth model information is used to train the first service Model.

In this implementation, the first data processing device uses the first encryption information to encrypt the fourth model information to obtain the third model information; it can prevent the fourth model information from being exposed to the second data processing device.

In a possible implementation manner, the method further includes: the first data processing device sends a second performance index to the data evaluation device, the second performance index represents the performance of the first business model, The second performance indicator is used by the data evaluation device to evaluate the value of the first local data for training the first business model.

In this implementation manner, the first data processing device sends the second performance index to the data evaluation device, which may enable the data evaluation device to evaluate the value of the first local data for training the first business model more quickly.

In a possible implementation manner, after sending the first indicator information to the data evaluation device, the method further includes: the first data processing device receives a decryption private key from the data evaluation device; The first data processing device uses the decryption private key to decrypt the second encrypted business model to obtain the second business model.

In this implementation, the first data processing device uses the decryption private key to decrypt the second encrypted business model to obtain the second business model; the required business model can be accurately obtained, and the second business model is prevented from being exposed to a second data processing device.

In a possible implementation manner, before the first data processing device uses the decryption private key to decrypt the second encrypted business model to obtain the second business model, the method further includes: the The first data processing device sends third model information to the data evaluation device; the third model information is used to train the first encrypted business model; the first data processing device according to the information from the second data processing device The second model information to obtain the second encrypted business model.

In this implementation manner, the first data processing device obtains the second encrypted business model according to the second model information from the second data processing device, so as to decrypt the second encrypted business model to obtain the second business model.

In the third aspect, the embodiment of the present application provides another data value evaluation method, including: the second data processing device receives the third model information from the data evaluation device; the third model information is used to train the first encrypted business model, The first encrypted business model is the same as the business model obtained by encrypting the first business model belonging to the first data processing device; the second data processing device sends second model information to the data evaluation device; the The second model information is used by the data evaluation device to obtain the first model information, and the first model information represents the second encrypted service obtained by the second data processing device using its first local data to train the one encrypted service model Model; the second business model obtained by decrypting the second encrypted business model is the same as the business model after training the first business model with the first local data.

In the embodiment of the present application, the second data processing device sends the second model information to the data evaluation device, which can not only avoid exposing its own first local data, but also enable the data evaluation device to evaluate the first local data for training the first business model the value of.

In a possible implementation manner, after the second data processing device receives the third model information from the data evaluation device and before the second data processing device sends the second model information to the data evaluation device, the The method further includes: the second data processing device uses the first encrypted data and the third model information to train the first encrypted business model to obtain the second encrypted business model; The first local data is obtained through encryption processing; the second data processing device obtains the second model information according to the second encrypted service model.

In this implementation, the second data processing device uses the first encrypted data and the third model information to train the first encrypted business model to obtain the second encrypted business model; the use of the first local data can be limited to the training data business model , that is, the first local data of the second data processing device does not come out of the local.

In a possible implementation manner, the second data processing device uses the first encrypted data and the third model information to train the first encrypted business model, and before obtaining the second encrypted model, the method further includes: The second data processing device receives the first encrypted information from the data evaluation device; the second data processing device encrypts the first local data by using the first encrypted information to obtain the First encrypt the data.

In this implementation manner, the second data processing device uses the first encrypted information to encrypt the first local data, so as to obtain encrypted data for training the first encrypted business model.

In a possible implementation manner, after the second data processing device sends the second model information to the data evaluation device, the method further includes: the second data processing device receives information from the data evaluation device Third model information; the third model information is used to train the first encrypted business model; the second data processing device uses the first local data to train the first encrypted business model to obtain the second An encrypted business model; the second data processing device sends second model information to the first data processing device; the second model information is used by the first data processing device to decrypt to obtain parameters of the second business model.

In this implementation manner, the second data processing device uses the first local data to train the first encrypted service model to obtain the second encrypted service model; it can help the first data processing device to implement model training.

In a fourth aspect, the embodiment of the present application provides a data usage right delivery method, the method includes: the data delivery device receives fifth model information from the third data processing device; the fifth model information is used to train the third encryption The business model, the business model obtained by decrypting the third encrypted business model is the same as the unencrypted third business model that the third data processing device needs the assistance of the fourth data processing device to train; the data delivery device sends the The fourth data processing device sends the fifth model information.

In this embodiment of the present application, the data delivery device sends fifth model information to the fourth data processing device, so that the fourth data processing device uses its local data and the fifth model information to train the third encrypted service model. Since the fifth model information is used to train the third encrypted business model, the fourth data processing device (which cannot successfully decrypt the third encrypted business model) cannot acquire an unencrypted business model according to the fifth model information. The fourth data processing device uses its local data to train the third encrypted business model. On the one hand, the third data processing device will not expose its business model and model training method, and on the other hand, the local data of the fourth data processing device will not leave the local . Therefore, in the embodiment of the present application, the third data processing device (buyer) can be helped to train the second business model on the premise that the local data of the fourth data processing device (seller) is completely invisible to the third data processing device (buyer).

In a possible implementation manner, the method further includes: the data delivery device sending a decryption private key to the third data processing device; the decryption private key is used by the third data processing device to The encrypted business model is decrypted, the fourth encrypted business model is obtained by the fourth data processing device using the second encrypted data to train the third encrypted business model, and the second encrypted data is obtained by encrypting the fourth data The second local data of the processing device is obtained, and the fourth business model obtained by decrypting the fourth encrypted business model is the same as the business model after training the third business model by using the second local data.

In this implementation manner, the data delivery device sends the decryption private key to the third data processing device, so that the third data processing device uses the decryption private key to decrypt the fourth encrypted business model.

In a possible implementation manner, the method further includes: the data delivery device sends first encrypted information to the third data processing device, and the first encrypted information is used by the third data processing device to generate The fifth model information.

In this implementation, the data delivery device sends the first encrypted information to the third data processing device, so that the third data processing device generates fifth model information, which can prevent the third data processing device from exposing its model training method or program.

In a possible implementation manner, the method further includes: the data delivery device sends first encrypted information to the fourth data processing device, and the first encrypted information is used by the fourth data processing device to encrypt Its second local data.

In this implementation, the data delivery device sends the first encrypted information to the fourth data processing device, so that the fourth data processing device uses the first encrypted information to encrypt its second local data to obtain the first encryption information for training the third encrypted business model. Two encrypted data.

In a possible implementation manner, the sending, by the data delivery device, the fifth model information to the fourth data processing device includes: when detecting that there is no security problem in the fifth model information, the The data delivering means sends the fifth model information to the fourth data processing means.

In this implementation manner, it is possible to avoid sending model information with security issues to the fourth data processing device.

In the fifth aspect, the embodiment of the present application provides another data usage right delivery method, the method includes: a third data processing device receives a decryption private key from the data delivery device; the third data processing device uses the decryption private key Decrypt the fourth encrypted business model to obtain the fourth business model; the fourth encrypted business model is obtained by the fourth data processing device using the second encrypted data to train the third encrypted business model, and the second encrypted data is obtained by the The fourth data processing device encrypts its second local data, and the fourth business model is the same as the business model after using the second local data to train the third business model, and the third business model is the third The encrypted business model corresponds to an unencrypted business model, and is a business model that the third data processing device needs assistance from the fourth data processing device for training.

In the embodiment of the present application, the third data processing device uses the decryption private key to decrypt the fourth encrypted business model to obtain the fourth business model; without obtaining the local data of the fourth data processing device, it can use The local data of the fourth data processing device implements model training.

In a possible implementation manner, before the third data processing device uses the decryption private key to decrypt the fourth encrypted business model to obtain the fourth business model, the method further includes: the third The data processing device sends fifth model information to the data delivery device; the fifth model information is used to train the third encrypted business model; the third data processing device according to the fourth data processing device Six model information to obtain the fourth encrypted service model.

In this implementation manner, the third data processing device sends the fifth model information to the data delivery device, which can prevent its model training method or program from being exposed. The third data processing device obtains the fourth encrypted business model according to the sixth model information from the fourth data processing device; the business model that can be used by the fourth data processing device to help it train.

In a possible implementation manner, before the third data processing device sends the fifth model information to the data delivery device, the method further includes: the third data processing device acquires the fifth model information from the data delivery device Encrypted information; the third data processing device encrypts the seventh model information according to the first encrypted information to obtain the fifth model information; the seventh model information is used to train the third business model .

In this implementation, the third data processing device encrypts the seventh model information according to the first encrypted information; it can not only enable the fourth data processing device to train the third encrypted business model, but also prevent the model training method or program from being exposed .

In a possible implementation manner, the third data processing device uses the decryption private key to decrypt the fourth encrypted business model, and after obtaining the fourth business model, the method further includes: the third data The processing means updates its local business model with parameters of said fourth business model. In this application, obtaining a business model may refer to obtaining parameters of the business model.

In this implementation manner, the third data processing device uses the parameters of the fourth service model to update its local service model; the fourth data processing device can be used to help it implement model training.

In the sixth aspect, the embodiment of the present application provides another data usage right delivery method, the method includes: the fourth data processing device receives the fifth model information from the data delivery device; the fifth model information is used to train the third encryption business model; the fourth data processing device uses the second encrypted data to train the third encrypted business model to obtain a fourth encrypted business model; the second encrypted data is encrypted by the fourth data processing device in its second local The data is obtained, and the fourth business model obtained by decrypting the fourth encrypted business model is the same as the business model after using the second local data to train the third business model, and the third business model is the third encrypted business model The corresponding unencrypted business model, which is the business model that the third data processing device needs the assistance of the fourth data processing device to train; the fourth data processing device sends the sixth model to the third data processing device Information; the sixth model information represents the fourth encrypted business model; the fourth encrypted business model is used by the third data processing device to decrypt to obtain the required model update parameters.

In the embodiment of the present application, the fourth data processing device receives the fifth model information from the data delivery device; the fourth data processing device uses the second encrypted data to train the third encrypted business model to obtain the fourth encrypted business model; The business models and model training methods of the three data processing devices were leaked. The fourth data processing device sends the sixth model information to the third data processing device, which can prevent the third data processing device from obtaining the local data of the fourth data processing device. That is to say, on the premise that the local data of the fourth data processing device is completely invisible to the third data processing device, it can help the third data processing device to realize the training of the service model.

In a possible implementation manner, the fourth data processing device uses the second encrypted data to train the third encrypted business model, and before obtaining the fourth encrypted business model, the method further includes: the fourth data processing The device encrypts the second local data by using the first encryption information from the data delivery device to obtain the second encrypted data.

In this implementation manner, the fourth data processing device uses the first encrypted information from the data delivery device to encrypt the second local data, so as to obtain the second encrypted data for training the third encrypted business model.

In a seventh aspect, the embodiments of the present application provide a data processing device, and the data processing device has a function of implementing the behaviors in the method embodiments of the first aspect above. The functions described above may be implemented by hardware, or may be implemented by executing corresponding software on the hardware. The hardware or software includes one or more modules or units corresponding to the above functions. In a possible implementation manner, it includes a transceiver module and a processing module, wherein: the transceiver module is configured to send first model information to the first data processing device; the first model information indicates that the second data processing device utilizes The second encrypted business model obtained by training the first encrypted business model with the first local data; the second business model obtained by decrypting the second encrypted business model and the business model after using the first local data to train the first business model Similarly, the first business model is an unencrypted business model corresponding to the first encrypted business model; the processing module is configured to obtain the first performance according to the first index information from the first data processing device Index; the first performance index represents the performance of the business model after using the first local data to train the first business model, and the first index information is that the first data processing device uses the first model Information and test data corresponding to the second encrypted service model are obtained; the processing module is further configured to evaluate the effectiveness of the first local data for training the first service according to the first performance index and the second performance index The value of the model; the second performance indicator characterizes the performance of the first business model.

In a possible implementation manner, the transceiver module is further configured to receive second model information from the second data processing device; the processing module is further configured to obtain the Describe the first model information.

In a possible implementation manner, the transceiving module is further configured to send third model information to the second data processing device; the third model information is used to train the first encrypted service model.

In a possible implementation manner, the transceiver module is further configured to send first encrypted information to the second data processing device; the first encrypted information is used by the second data processing device to encrypt the first A local data, the second encrypted business model is obtained by the second data processing device using the first encrypted data and the third model information to train the first encrypted business model, the first encrypted data is obtained by the The first encryption information is obtained by encrypting the first local data.

In a possible implementation manner, the transceiving module is further configured to receive the third model information from the first data processing device.

In a possible implementation manner, the transceiver module is further configured to send first encrypted information to the first data processing device; the first encrypted information is used by the first data processing device to encrypt the fourth model information to obtain the third model information, and the fourth model information is used to train the first business model.

In a possible implementation manner, the processing module is specifically configured to decrypt the first indicator information to obtain the first performance indicator.

In a possible implementation manner, the transceiver module is further configured to send a decryption private key to the first data processing device; the decryption private key is used by the first data processing device to encrypt the second The business model is decrypted.

In a possible implementation manner, the transceiver module is further configured to receive third model information from the first data processing device; the third model information is used to train the first encrypted service model; The data evaluation means sends the third model information to the second data processing means.

Regarding the seventh aspect or the technical effects brought by various possible implementations of the seventh aspect, reference may be made to the introduction to the first aspect or the technical effects of various possible implementations of the first aspect.

In an eighth aspect, the embodiment of the present application provides another data processing device, the data processing device has the function of implementing the behavior in the method embodiment of the second aspect above. The functions described above may be implemented by hardware, or may be implemented by executing corresponding software on the hardware. The hardware or software includes one or more modules or units corresponding to the above functions. In a possible implementation, it includes a transceiver module and a processing module, wherein: the transceiver module is configured to receive the first model information from the data evaluation device; the first model information represents the model information used by the second data processing device The second encrypted business model obtained by training the first encrypted business model with the first local data; the second encrypted business model obtained by decrypting the second encrypted business model is the same as the business model obtained by using the first local data to train the first business model , the first business model is an unencrypted business model corresponding to the first encrypted business model; the processing module is configured to, according to the first model information and the test data corresponding to the second encrypted business model, Controlling the transceiver module to send first index information to the data evaluation device; the first index information is used by the data evaluation device to obtain a first performance index; the first performance index is characterized by using the first local data The performance of the business model after training the first business model.

In a possible implementation manner, the processing module is specifically configured to use encrypted test data to test the second encrypted business model to obtain the first index information; the encrypted test data is encrypted by using the first encrypted information. The above test data were obtained.

In a possible implementation manner, the transceiving module is further configured to send third model information to the data evaluation device, where the third model information is used for training the first encrypted service model.

In a possible implementation manner, the transceiver module is further configured to receive the first encrypted information from the data evaluation device; the processing module is further configured to use the first encrypted information to update the fourth model information Perform encryption processing to obtain the third model information, and the fourth model information is used to train the first service model.

In a possible implementation manner, the transceiver module is further configured to send a second performance indicator to the data evaluation device, the second performance indicator represents the performance of the first business model, and the second performance The indicator is used by the data evaluating means to evaluate the value of the first local data for training the first business model.

In a possible implementation manner, the transceiver module is further configured to receive a decryption private key from the data evaluation device; the first data processing device uses the decryption private key to encrypt the second encrypted business model Perform decryption processing to obtain the second service model.

In a possible implementation manner, the transceiving module is further configured to send third model information to the data evaluation device; the third model information is used to train the first encrypted business model; the transceiving module is further configured to obtain the second encrypted service model according to the second model information from the second data processing device.

Regarding the eighth aspect or the technical effects brought by various possible implementation manners of the eighth aspect, reference may be made to the introduction of the second aspect or the technical effects of various possible implementation manners of the second aspect.

In a ninth aspect, the embodiment of the present application provides another data processing device, the data processing device has the function of implementing the behavior in the method embodiment of the third aspect above. The functions described above may be implemented by hardware, or may be implemented by executing corresponding software on the hardware. The hardware or software includes one or more modules or units corresponding to the above functions. In a possible implementation manner, the data processing device includes a transceiver module, wherein: the transceiver module is configured to receive third model information from the data evaluation device; the third model information is used to train the first encrypted business model , the first encrypted business model is the same as the business model obtained by encrypting the first business model belonging to the first data processing device; the transceiver module is further configured to send second model information to the data evaluation device; The second model information is used by the data evaluation device to obtain first model information, and the first model information represents the second data obtained by the second data processing device using its first local data to train the one encrypted service model. An encrypted business model; the second business model obtained by decrypting the second encrypted business model is the same as the business model after training the first business model with the first local data.

In a possible implementation manner, the data processing device further includes a processing module; the processing module is configured to use the first encrypted data and the third model information to train the first encrypted business model to obtain the A second encrypted business model; the first encrypted data is obtained by encrypting the first local data; the processing module is further configured to obtain the second model information according to the second encrypted business model.

In a possible implementation manner, the transceiver module is further configured to receive the first encrypted information from the data evaluation device; the processing module is further configured to use the first encrypted information to process the The first local data is encrypted to obtain the first encrypted data.

In a possible implementation manner, the transceiver module is further configured to receive third model information from the data evaluation device; the third model information is used to train the first encrypted business model; the processing The module is further configured to use the first local data to train the first encrypted business model to obtain the second encrypted business model; the transceiver module is further configured to send the second model to the first data processing device information; the second model information is used by the first data processing device to decrypt to obtain parameters of the second service model.

Regarding the ninth aspect or the technical effects brought by various possible implementation manners of the ninth aspect, reference may be made to the introduction of the third aspect or the technical effects of various possible implementation manners of the third aspect.

In a tenth aspect, the embodiment of the present application provides another data processing device, the data processing device has the functions in the method embodiment of the fourth aspect above. The functions described above may be implemented by hardware, or may be implemented by executing corresponding software on the hardware. The hardware or software includes one or more modules or units corresponding to the above functions. In a possible implementation manner, the data processing device includes a transceiver module, wherein: the transceiver module is configured to receive fifth model information from the third data processing device; the fifth model information is used to train the third encrypted business model The business model obtained by decrypting the third encrypted business model is the same as the unencrypted third business model that the third data processing device needs the assistance of the fourth data processing device to train; The fourth data processing device sends the fifth model information.

In a possible implementation manner, the transceiver module is further configured to send a decryption private key to the third data processing device; the decryption private key is used by the third data processing device to performing decryption processing, the fourth encrypted business model is obtained by the fourth data processing device using the second encrypted data to train the third encrypted business model, and the second encrypted data is obtained by encrypting the fourth data processing device The second local data is obtained, and the fourth business model obtained by decrypting the fourth encrypted business model is the same as the business model obtained by using the second local data to train the third business model.

In a possible implementation manner, the transceiver module is further configured to send first encrypted information to the third data processing device, where the first encrypted information is used by the third data processing device to generate the first Five model information.

In a possible implementation manner, the transceiver module is further configured to send first encrypted information to the fourth data processing device, and the first encrypted information is used by the fourth data processing device to encrypt its second local data.

In a possible implementation manner, the transceiving module is specifically configured to send the fifth model information to the fourth data processing device when it is detected that there is no security problem in the fifth model information .

Regarding the technical effects brought by the tenth aspect or various possible implementation manners of the tenth aspect, reference may be made to the introduction of the fourth aspect or the technical effects of various possible implementation manners of the fourth aspect.

In an eleventh aspect, the embodiment of the present application provides another data processing device, the data processing device has the function of implementing the behavior in the method embodiment of the fifth aspect above. The functions described above may be implemented by hardware, or may be implemented by executing corresponding software on the hardware. The hardware or software includes one or more modules or units corresponding to the above functions. In a possible implementation manner, the data processing device includes a transceiver module and a processing module, wherein: the transceiver module is configured to receive a decryption private key from the data delivery device; the processing module is configured to use the decryption private key key to decrypt the fourth encrypted business model to obtain the fourth business model; the fourth encrypted business model is obtained by the fourth data processing device using the second encrypted data to train the third encrypted business model, and the second encrypted data is obtained by The fourth data processing device obtains by encrypting its second local data, the fourth business model is the same as the business model after using the second local data to train the third business model, and the third business model is the first The three encrypted business models correspond to unencrypted business models, and are business models that the third data processing device needs the assistance of the fourth data processing device to train.

In a possible implementation manner, the transceiver module is further configured to send fifth model information to the data delivery device; the fifth model information is used to train the third encrypted service model; the processing module is further configured to obtain the fourth encrypted service model according to the sixth model information from the fourth data processing device.

In a possible implementation manner, the processing module is further configured to obtain first encrypted information from the data delivery device through the transceiver module; perform encryption processing on the seventh model information according to the first encrypted information, The fifth model information is obtained; the seventh model information is used to train the third service model.

In a possible implementation manner, the processing module is further configured to use parameters of the fourth service model to update its local service model.

For the technical effects of the eleventh aspect or various possible implementations of the eleventh aspect, reference may be made to the introduction of the fifth aspect or the technical effects of various possible implementations of the fifth aspect.

In a twelfth aspect, the embodiment of the present application provides another data processing device, which has the function of implementing the behavior in the method embodiment of the sixth aspect above. The functions described above may be implemented by hardware, or may be implemented by executing corresponding software on the hardware. The hardware or software includes one or more modules or units corresponding to the above functions. In a possible implementation manner, the data processing device includes a transceiver module, where: the transceiver module is configured to receive fifth model information from the data delivery device; the fifth model information is used to train the third encrypted business model The fourth data processing device uses the second encrypted data to train the third encrypted business model to obtain a fourth encrypted business model; the second encrypted data is obtained by encrypting its second local data by the fourth data processing device The fourth business model obtained by decrypting the fourth encrypted business model is the same as the business model after using the second local data to train the third business model, and the third business model corresponds to the third encrypted business model An unencrypted business model, which is a business model that the third data processing device needs the assistance of the fourth data processing device to train; the transceiver module is also configured to send sixth model information to the third data processing device The sixth model information represents the fourth encrypted business model; the fourth encrypted business model is used by the third data processing device to decrypt to obtain the required model update parameters.

In a possible implementation manner, the data processing device further includes a processing module; the processing module is configured to encrypt the second local data by using the first encryption information from the data delivery device to obtain The second encrypted data.

Regarding the technical effects brought about by the twelfth aspect or various possible implementation manners of the twelfth aspect, reference may be made to the introduction of the sixth aspect or the technical effects of various possible implementation manners of the sixth aspect.

In a thirteenth aspect, the present application provides a server, the server includes a processor, and the processor can be used to execute computer-executed instructions stored in the memory, so that the above-mentioned first aspect or any possible implementation of the first aspect The method shown in the above-mentioned second aspect or any possible implementation of the second aspect is executed, or the method shown in the above-mentioned third aspect or any possible implementation of the third aspect is executed. The method is executed, or the method shown in the fourth aspect or any possible implementation of the fourth aspect is executed, or the method shown in the fifth aspect or any possible implementation of the fifth aspect is executed Execute, or enable the method shown in the sixth aspect or any possible implementation manner of the sixth aspect to be executed.

In the embodiment of the present application, in the process of executing the above method, the process of sending information in the above method can be understood as the process of outputting information based on the instructions of the processor. In outputting information, the processor outputs the information to the transceiver for transmission by the transceiver. After the information is output by the processor, it may also need to undergo other processing before reaching the transceiver. Similarly, when the processor receives incoming information, the transceiver receives that information and inputs it to the processor. Furthermore, after the transceiver receives the information, the information may require other processing before being input to the processor.

For the sending and/or receiving operations involved in the processor, if there is no special description, or if it does not conflict with its actual function or internal logic in the relevant description, it can be generally understood as the instruction output based on the processor .

During implementation, the above-mentioned processor may be a processor dedicated to performing these methods, or may be a processor that executes computer instructions in a memory to perform these methods, such as a general-purpose processor. For example, the processor may also be used to execute a program stored in the memory. When the program is executed, the data processing device is made to execute the method as shown in the first aspect or any possible implementation manner of the first aspect. In a possible implementation manner, the memory is located outside the data processing device. In a possible implementation manner, the memory is located in the above data processing device.

In the embodiment of the present application, the processor and the memory may also be integrated into one device, that is, the processor and the memory may also be integrated together.

In a possible implementation manner, the data processing apparatus further includes a transceiver, where the transceiver is configured to receive a message or send a message, and the like.

In a fourteenth aspect, the present application provides a data processing device, the data processing device includes a processing circuit and an interface circuit, and the interface circuit is used to obtain data or output data; the processing circuit is used to perform the above-mentioned first aspect or the first aspect The corresponding method shown in any possible implementation manner of the second aspect, or the processing circuit is used to execute the corresponding method shown in the second aspect or any possible implementation manner of the second aspect, or the processing circuit is used to execute the corresponding method shown in the first aspect above The corresponding method shown in the third aspect or any possible implementation of the third aspect, or the processing circuit is used to execute the corresponding method shown in the fourth aspect or any possible implementation of the fourth aspect, or the processing circuit For performing the corresponding method as shown in the above fifth aspect or any possible implementation of the fifth aspect, or the processing circuit is used for performing the corresponding method as shown in the above sixth aspect or any possible implementation of the sixth aspect Methods.

In a fifteenth aspect, the present application provides a computer-readable storage medium, which is used to store a computer program, and when it is run on a computer, the above-mentioned first aspect or any possible implementation of the first aspect can be realized The method shown in the manner is executed, or the method shown in the second aspect or any possible implementation of the second aspect is executed, or the method shown in the third aspect or any possible implementation of the third aspect is executed. The method is executed, or the method shown in the fourth aspect or any possible implementation manner of the fourth aspect is executed, or the method shown in the fifth aspect or any possible implementation manner of the fifth aspect is executed, Or the method shown in the sixth aspect or any possible implementation manner of the sixth aspect is executed.

In a sixteenth aspect, the present application provides a computer program product, the computer program product includes a computer program or computer code, and when it is run on a computer, the above-mentioned first aspect or any possible implementation of the first aspect shows The method is executed, or the method shown in the second aspect or any possible implementation of the second aspect is executed, or the method shown in the third aspect or any possible implementation of the third aspect is executed , or make the above fourth aspect or the method shown in any possible implementation of the fourth aspect be executed, or cause the above fifth aspect or the method shown in any possible implementation of the fifth aspect to be executed, or make the above The method shown in the sixth aspect or any possible implementation manner of the sixth aspect is executed.

In the seventeenth aspect, the present application provides a data value evaluation system, including the data evaluation device described in the seventh aspect or any possible implementation of the seventh aspect, and the eighth aspect or any possible implementation of the eighth aspect The data processing device described in the manner and the data processing device described in the ninth aspect or any possible implementation of the ninth aspect.

In an eighteenth aspect, the present application provides a data usage right delivery system, including the data delivery device described in the above tenth aspect or any possible implementation of the tenth aspect, any of the above eleventh aspect or the eleventh aspect The data processing device described in a possible implementation manner and the data processing device described in the above twelfth aspect or any possible implementation manner of the twelfth aspect.

Description of drawings

In order to more clearly illustrate the technical solutions in the embodiment of the present application or the background art, the following will describe the drawings that need to be used in the embodiment of the present application or the background art.

FIG. 1 is a schematic diagram of a data value evaluation system provided by an embodiment of the present application;

FIG. 2 is a schematic diagram of a data usage right delivery system provided by an embodiment of the present application;

Fig. 3 is an interactive flowchart of a data value evaluation method provided by the embodiment of the present application;

Fig. 4 is an interactive flowchart of another data value evaluation method provided by the embodiment of the present application;

Fig. 5 is a flow chart of a data value evaluation method provided by the embodiment of the present application;

FIG. 6 is a flow chart of another data value evaluation method provided in the embodiment of the present application;

FIG. 7 is a flow chart of another data value evaluation method provided in the embodiment of the present application;

FIG. 8 is a flow chart of another data value evaluation method provided in the embodiment of the present application;

FIG. 9 is a flow chart of another data value evaluation method provided in the embodiment of the present application;

FIG. 10 is a flow chart of another data value evaluation method provided in the embodiment of the present application;

Fig. 11 is an interactive flowchart of a data usage right delivery method provided by the embodiment of the present application;

Fig. 12 is a flow chart of a data usage right delivery method provided by the embodiment of the present application;

FIG. 13 is a flow chart of another data usage right delivery method provided by the embodiment of the present application;

Fig. 14 is a flow chart of another data usage right delivery method provided by the embodiment of the present application;

Fig. 15 is an interactive flowchart of a data evaluation and use right delivery method provided by the embodiment of the present application;

FIG. 16 shows a schematic structural diagram of a data processing device 1600;

FIG. 17 is a schematic structural diagram of another data processing device 170 provided in the embodiment of the present application;

FIG. 18 is a schematic structural diagram of another data processing device 180 provided by an embodiment of the present application.

Detailed ways

The terms "first" and "second" in the description, claims, and drawings of the embodiments of the present application are only used to distinguish different objects, rather than to describe a specific order. Furthermore, the terms "comprising" and "having", as well as any variations thereof, are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally It also includes other steps or units inherent to these processes, methods, products, or devices.

Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least one embodiment of the present application. The occurrences of this phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is understood explicitly and implicitly by those skilled in the art that the embodiments described herein can be combined with other embodiments.

The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to limit the present application. As used in the specification and appended claims of this application, the singular expressions "a", "an", "said", "above", "the" and "this" are intended to also Plural expressions are included unless the context clearly indicates otherwise. It should also be understood that the term "and/or" as used in this application refers to and includes any and all possible combinations of one or more of the listed items. For example, "A and/or B" can mean: there are only A, only B, and both A and B, where A and B can be singular or plural. The term "plurality" as used in this application means two or more.

The terms used in the embodiments of the present application are firstly introduced below.

Homomorphic encryption (homomorphic encryption): processing the homomorphically encrypted data to obtain an output, and decrypting this output, the result is the same as the output obtained by processing the unencrypted original data in the same way. Homomorphic encryption is a cryptographic technique based on computational complexity theory of mathematical puzzles. In essence, homomorphic encryption refers to such an encryption function, which performs ring addition and multiplication operations on the plaintext and re-encrypts, and performs corresponding operations on the ciphertext after encryption, and the result is equivalent. Due to this good nature, people can entrust a third party to process the data without disclosing the information. An encryption function with homomorphic properties means that two plaintexts a and b satisfy

Encryption function, where En is the encryption operation, Dec is the decryption operation, ⊙,

Corresponding to operations on plaintext and ciphertext domains, respectively. when

When represents addition, the encryption is called additive homomorphic encryption: when

When stands for multiplication, the encryption is called multiplicative homomorphic encryption. Fully homomorphic encryption refers to an encryption function that satisfies both the homomorphic and homomorphic properties of addition and multiplication, and can perform any number of addition and multiplication operations. Expressed in mathematical formulas, namely Dec(f(En(m1), En(m2),...,En(mk)))=f(m1, m2,...,mk), or written as: f(En(m1) , En(m2),..., En(mk))=En(f(m1, m2,...,mk)), if f is an arbitrary function, it is called fully homomorphic encryption. The fully homomorphic encryption system can perform any operation on the encrypted data that can be performed on the plaintext without decryption, so that the encrypted information can still be deeply and infinitely analyzed without affecting its confidentiality.

The key to developing a high-quality new energy vehicle data business model lies in high-quality vehicle operation data. At present, although automobile OEMs and government regulatory platforms at all levels store a large amount of operating data of new energy vehicles, the quality of these data is generally not high, there are a lot of missing and wrong problems, and these data usually lack vehicle status labels. For the development of data business models, the real value is the data of those abnormal vehicles. However, the amount of data generated for these abnormal vehicles is small, and determining the abnormal type of the vehicle requires human labeling. Therefore, these abnormal vehicle data with manual annotations belong to the confidential data of major automobile OEMs. They are strictly controlled and have almost no liquidity, which makes the rare but high-value abnormal vehicle data scattered among major automobile OEMs. Form a data island. The value of these valuable vehicle data has not been fully tapped, resulting in a huge waste of resources.

At present, the way of developing new energy vehicle business models (an example of data business models) using data cannot open up the data channel between various automobile OEMs, that is, the problem of data mobility has not been fundamentally resolved, resulting in the development of new energy vehicle business models. It is difficult for automobile OEMs or third-party auto parts suppliers to meet the demand for high-quality vehicle data. For automobile OEMs or third-party auto parts suppliers, because they cannot accurately evaluate the value of different vehicle data for their data business model development, they do not know which vehicle data to use to develop their data business models, nor It makes sense to know how much you will be paid for using this vehicle data. It can be seen that one of the obstacles to the fundamental solution to the mobility of data lies in the inability to accurately evaluate the value of vehicle data for the development of data business models. Another obstacle that has not fundamentally solved the problem of data liquidity is how to complete the transaction between the buyer and the seller without exposing the data business model developed by the buyer (or the method of developing the data business model) and the seller without exposing the data. data delivery.

In order to solve the problem of data mobility, it is necessary to study how to accurately evaluate the value of vehicle data for the development of data business models and how to complete the data delivery between buyers and sellers. The data value evaluation method provided in the embodiment of the present application can accurately evaluate the value of vehicle data for the training data business model. The data value evaluation method provided by the embodiment of this application can not only be used to evaluate the value of vehicle data for training data business models, but also can evaluate arbitrary data (such as new energy vehicle charging pile data, data with similar data structures between different organizations in the same industry, etc. ) for any object (such as a model to be trained), that is, the value of the right to use the data. Using the data value evaluation method provided by the embodiment of the present application to evaluate the value of data use rights has the following advantages: the buyer does not expose the data business model or the method of developing the data business model, such as data processing and model training methods, and On the premise that the seller does not disclose its data details, output the value of the seller's data to the buyer. Further, the embodiment of the present application also provides a data usage right delivery method, by which the data usage right can be traded, and the data security of both transaction parties can be guaranteed. Using the data usage right delivery method provided in this application to deliver the value of the data usage right has the following advantages: what is delivered is the data usage right, the seller's data does not leave the local area, and the use of the data is limited to the development of the buyer's data business model. On the premise that the detailed value of the seller's data is completely invisible to the buyer, it helps the buyer realize the development of the data business model. Under the premise of satisfying the privacy and security protection of the seller's data, the potential value of the seller's data can be brought into play. The delivery method of data usage rights provided by this application is the usage rights of transaction data rather than data ownership, and the use of data is limited to training data business models, and the buyer purchases the seller's data to help it train the service of the model. After the buyer purchases the right to use the data, a buyer-led federated learning architecture can be built based on the data trading system. In this architecture, the buyer can obtain the latest parameters of the seller's data after each round of model training without providing the seller with an updated model. The latest parameters of . The following uses evaluating the value of vehicle data for the training data business model as an example to introduce the data value evaluation method provided by this application; taking the delivery of vehicle data usage rights as an example to introduce the data usage right delivery method provided by this embodiment of the application.

Fig. 1 is a schematic diagram of a data value evaluation system provided by an embodiment of the present application. As shown in Figure 1, the data value evaluation system includes: a data evaluation device (which can be regarded as a server running a data transaction system), one or more buyers' data processing devices (only the first data processing device shown in Figure 1 device as an example of a buyer's data processing device) and one or more seller's data processing devices (the second data processing device shown in FIG. 1 as an example of a seller's data processing device). In this application, for the sake of simplicity, the data processing device of the buyer may be referred to as the buyer for short, and the data processing device of the seller may be referred to as the seller for short. The buyer's data processing device refers to the device used by the buyer to train the data business model, and the buyer has the right to use or own the data processing device. The seller's data processing device refers to the device used by the seller to store local data and realize data processing functions, and the seller has the right to use or ownership of its data processing device. Both the first data processing device and the second data processing device may be devices or devices capable of data processing and storage, such as servers. Since the data evaluation device provided in the embodiment of the present application can provide a data value evaluation function, the data evaluation device can be regarded as a data value evaluation platform product. The buyer can use the data value evaluation platform products to evaluate the value of the seller's data for its data business model development, such as the value of training any data business model.

In some embodiments, the data value evaluation platform is characterized by: input one: the buyer's data processing and training model method or program, and the buyer's initial data model performance indicators; input two: the seller uses the data processing and training provided by the buyer The method or program for training the model, and the parameters of the trained model. Output: The model parameters trained by the buyer using the seller's data, after updating the parameters of the initial data model, the percentage of model performance improvement and the value of the seller's data.

FIG. 2 is a schematic diagram of a data usage right delivery system provided by an embodiment of the present application. As shown in Figure 2, the data usage right delivery system includes: a data delivery device (which can be regarded as a server running the data usage right delivery system), one or more buyer's data processing devices (only the first one shown in Figure 2 three data processing devices as an example of a buyer's data processing device) and one or more seller's data processing devices (a fourth data processing device shown in FIG. 2 as an example of a seller's data processing device). Both the third data processing device and the fourth data processing device may be devices or devices capable of data processing and storage, such as servers. Since the data delivery device provided in this application can provide the delivery function of the data usage right (that is, the function of delivering the seller's data usage right to the buyer), the data delivery device can be regarded as a data usage right delivery platform product. The buyer can purchase the right to use the seller's data through the delivery platform product of the right to use the data, that is, the right to use the seller's data.

In some embodiments, the data usage rights delivery platform is characterized by providing a federated learning framework for joint modeling of buyers and sellers. In this framework, the buyer only buys the seller's data to participate in the buyer-led federated learning framework, uses the seller's data to train the model and passes the model parameters to the buyer to update the buyer's model. In this process, the seller's data does not come out locally, and the detailed values of the seller's data are completely invisible to the buyer. The use of the seller's data is limited to helping the buyer train the model in the buyer-led federated learning architecture.

The data usage right delivery system in Figure 2 and the data value evaluation system in Figure 1 can be two independent systems, or they can be the same system. In other words, the data usage right delivery platform and the data value evaluation platform in this application can be two independent platforms, or they can be the same platform. It should be understood that if the data usage right delivery system in FIG. 2 and the data value evaluation system in FIG. 1 are the same system, then the data evaluation device in FIG. 1 and the data delivery device in FIG. 2 are the same device.

The interaction process of a data value evaluation method provided by the embodiment of the present application is described below with reference to the accompanying drawings.

Fig. 3 is an interactive flowchart of a data value evaluation method provided by the embodiment of the present application. As shown in Figure 3, the method includes:

301. The data evaluation device sends first encrypted information to a first data processing device and a second data processing device respectively.

The data evaluation device may store or obtain a decryption private key corresponding to the first encrypted information. For example, the first encryption information is a homomorphic encryption program for implementing homomorphic encryption of data, and the data evaluation device may store or obtain a decryption private key corresponding to the homomorphic encryption program. The data evaluation device can use the decryption private key to decrypt encrypted data obtained by encrypting arbitrary data with the first encrypted information.

The first data processing device is a data processing device of the buyer, and the second data processing device is a data processing device of the seller. In some embodiments, the data evaluation device, the first data processing device and the second data processing device are all servers, such as cloud servers. The first data processing device may use the first encrypted information to encrypt the fourth model information used to train the first business model to obtain the third model information. The third model information can be used to train the first encrypted business model. Encrypting the first business model can obtain the first encrypted business model. In other words, the first encrypted business model can be decrypted to obtain the first business model. The fourth model information may be a data processing and model training program or code for training the first service model. The second data processing device may use the first encrypted information to encrypt the first local data to obtain the first encrypted data. Wherein, the first local data is the data to be evaluated.

The above-mentioned first encryption information may be information for realizing homomorphic encryption of data. For example, the first encryption information is a homomorphic encryption program, an encryption function capable of realizing homomorphic encryption of data, etc., wherein the homomorphic encryption program refers to a program capable of realizing homomorphic encryption of data. In a possible implementation, the first encrypted information is a homomorphic encryption program; the first data processing device uses the homomorphic encryption program to encrypt the fourth model information used to train the first business model to obtain the third Model information, the third model information can be used to train the first encrypted business model; the second data processing device uses the homomorphic encryption program to encrypt its first local data to obtain the first encrypted data, and uses the first encrypted The data and the third model information train the first encrypted business model to obtain the second encrypted business model. Neither the first data processing device nor the second data processing device stores and cannot obtain the decryption private key of the homomorphic encryption program. The first encrypted business model can be regarded as obtained by the first data processing device by using a homomorphic encryption program to encrypt the first business model. In other words, the above-mentioned first business model is an unencrypted business model corresponding to the first encrypted business model. It can be seen from the characteristics of homomorphic encryption that the service model obtained by decrypting the second encrypted service model is the same as the service model after training the first service model with the first local data. Since the second data processing device obtains and uses the first encrypted data and the third model information to train the first encrypted business model, it does not store and cannot obtain the decryption private key of the homomorphic encryption program, so the second data processing device cannot obtain the Fourth model information for training the first business model. That is to say, in this implementation, the buyer's fourth model information (such as data processing and model training methods) will not be exposed to the seller. The first data processing device may obtain first model information representing the second encrypted service model. Since the first data processing device only provides the first model information to the first data processing device instead of the first local data, the first data processing device cannot obtain the first local data. That is, the seller's data details are not exposed to the buyer.

302. The first data processing device sends the initial performance index and third model information obtained by encrypting fourth model information by using the first encrypted information to the data evaluation device.

The fourth model information is used to train the first business model, and the first business model is a business model that needs to be trained for the first data processing. The method flow in FIG. 3 can be regarded as a flow of evaluating the value of the first local data of the second data processing device for training the first business model. The initial performance index (or called the second performance index) represents the performance of the first service model. Initial performance metrics can include precision, recall, etc. It should be understood that different service models have different performance indexes, which are not limited in this application. The third model information is used to train the first encrypted business model. The first encrypted business model may be obtained by the first data processing device performing encryption processing on the first business model by using the first encrypted information.

In some possible embodiments, the first encryption information is a homomorphic encryption program, the fourth model information includes a data processing and model training program for training the first business model, and the third model information includes the data processing and model training program Encrypt the processed program with a homomorphic encryption program; the first data processing device sends the program (corresponding to the fourth model information) encrypted with the data processing and model training program to the data evaluation device, and Initial performance indicators characterizing the performance of the first business model are sent to the data evaluation means.

303. The data evaluation device sends third model information to the second data processing device.

A possible implementation manner of step 303 is as follows: after the data evaluation device detects that the third model information has no security problem, it sends the third model information to the second data processing device. The way for the data evaluation device to detect whether the third model information has a security problem may include: detecting whether the data processing and model training program included in the third model information is encrypted with a homomorphic encryption program. Check if the program is complete, scan the program for viruses, etc. The first data processing device can implement integrity protection on the program through a signature algorithm. It should be understood that if it is detected that the encryption algorithm used by the program complies with general security specifications, and passes the integrity check and virus scanning (that is, no virus is detected by the program), then it is determined that the program has no security problems.

304. The second data processing device sends second model information to the data evaluating device.

The second model information is used by the data evaluation device to obtain the first model information. The above-mentioned first model information represents the second encrypted service model obtained by the second data processing device using its first local data to train the first encrypted service model. The second business model obtained by decrypting the second encrypted business model is the same as the business model after training the first business model with the first local data.

In some embodiments, the second data processing device may perform the following operations before performing step 304: the second data processing device encrypts its first local data by using the first encrypted information to obtain the first encrypted data; The data processing device uses the first encrypted data and the third model information to train the first encrypted business model to obtain the second encrypted business model; generates the first model information according to the parameters of the second encrypted business model; performs encryption processing on the first model information to get the second model information. The first model information contains parameters characterizing the second encrypted business model or contains parameters of the second encrypted business model. The second model information includes parameters obtained by encrypting parameters representing the second encrypted service model or parameters obtained by encrypting parameters of the second encrypted service model. In these embodiments, the second data processing device may encrypt the first model information using an encryption method pre-agreed with the data evaluation device to obtain the second model information. The data evaluation device decrypts the second model information to obtain the first model information. It should be noted that the purpose of encrypting the first model information by the second data processing device is to ensure the security of the data transmitted between the second data processing device and the data evaluation device, and avoid sending the first model information directly to cause the first model information to be encrypted. Give way.

In some embodiments, the second data processing device may perform the following operations before performing step 304: the second data processing device encrypts its first local data by using the first encrypted information to obtain the first encrypted data; The data processing device uses the first encrypted data and the third model information to train the first encrypted business model to obtain a second encrypted business model; generates the first model information according to the parameters of the second encrypted business model. The first model information contains parameters characterizing the second encrypted business model or contains parameters of the second encrypted business model. In these embodiments, the first model information and the second model information are the same. The second data processing device sends the second model information to the data evaluation device, that is to say the first model information to the data evaluation device.

305. The data evaluation device sends the first model information to the first data processing device according to the second model information.

A possible implementation of step 305 is as follows: the data evaluation device decrypts the second model information to obtain the first model information; after detecting that the first model information has no security issues, it sends the first model information to the first data processing device .

The method for the data evaluation device to detect whether the first model information has a security problem may include: detecting whether the encryption algorithm used by the first model information complies with general security specifications, detecting whether the first model information is complete, and performing virus detection on the first model information. scan etc.

A possible implementation manner of step 305 is as follows: after detecting that the second model information has no security problem, sending the second model information to the first data processing device; wherein, the second model information is the same as the first model information.

306. The first data processing device sends the first indicator information to the data evaluation device according to the first model information and the test data corresponding to the second encrypted service model.

The above-mentioned first index information is used for the above-mentioned data evaluation device to obtain the first performance index. The above-mentioned first performance index represents the performance of the service model after the first service model is trained by using the first local data.

A possible implementation of step 306 is as follows: the first data processing device encrypts its local test data (that is, the test data corresponding to the second encrypted business model) by using the first encrypted information to obtain encrypted test data; Test the second encryption business model to obtain the first indicator information. The encrypted test data and the second encrypted business model can be obtained by using the same homomorphic encryption program, so the encrypted test data can be used to test the second encrypted business model. The first encryption information may be a homomorphic encryption program, and the first index information may be a group of homomorphic encryption performance indicators corresponding to the second encryption business model.

307. The data evaluation device decrypts the first index information by using the decryption private key corresponding to the first encrypted information to obtain the first performance index.

308. The data evaluation device evaluates the value of the first local data for training the first service model according to the initial performance index and the first performance index.

A possible implementation manner of step 308 is as follows: the data evaluation device evaluates the value of the first local data for training the first business model by comparing the initial performance index with the first performance index.

In some embodiments, the data evaluation device compares the initial performance index with the first performance index, and evaluates the percentage of the first local data for the performance improvement of the buyer model (ie, the first business model) and the data reference value (ie, the first local data For training the value of the first business model); and sending the first local data to the first data processing device and the second data processing device respectively to the percentage and/or the first local data of the buyer's model (ie the first business model) performance improvement Value for training the first business model.

The main principle of the embodiment of the present application is: the data evaluation device uses homomorphic encryption technology to hide the buyer's data processing and model training methods from the seller, and hide the seller's data from the buyer; output a set of buyer model performance after updating the buyer's model based on the seller's data. The percentage of lift metrics. It should be understood that the value of any data of the seller for training any business model of the buyer can be evaluated by adopting the method flow in FIG. 3 , and it is ensured that the data of the seller and the model training program or code of the buyer are not exposed. For the buyer, the value of any data of the seller to the buyer's training business model can be evaluated, so that the buyer can choose whether to purchase the right to use the seller's data and the right to use the seller's data according to the value of the seller's data to its training business model Fair remuneration. One problem solved by the data value evaluation method provided in this application is to evaluate the value of the seller's data to the buyer without exposing the buyer's data processing and model training methods and the seller's data. The technical means adopted are: adopt homomorphic encryption technology, hide the buyer's data processing and model training methods from the seller before the transaction is concluded, and hide the seller's data from the buyer; The percentage of model performance improvement metrics. Another problem solved by the data value evaluation method provided in this application is: to meet the privacy and security protection requirements of the seller’s data: the seller’s data cannot leave the seller’s control area, the detailed value of the seller’s data is invisible to the buyer, and the use of the data can only be It is used to train the data business model, and meets the buyer's demand for using the seller's data to train the data business model under this restriction. The technical means adopted: improve the federated learning architecture, and the buyer purchases the seller's data to participate in the federated learning architecture led by it. Achieving technical effect: On the premise of meeting the seller's data privacy and security protection requirements, the data buyer's training data business model needs are met, and the potential value of the originally immobile data can be brought into play.

In the embodiment of the present application, the second data processing device (seller) sends the second model information to the data evaluation device, and neither the data evaluation device nor the first data processing device (buyer) can obtain the second data through the second model information Process the local data of the device; it can meet the seller's data privacy and security protection requirements. That is to say, the seller's data (such as the first local data) will not leave the seller's control area, the detailed value of the seller's data is invisible to the buyer, and the use of the data can only be used to train the data business model. Buyer needs to train data business models using seller data. In the embodiment of the present application, the first data processing device sends the initial performance index and the third model information obtained by encrypting the fourth model information by using the first encrypted information to the data evaluation device; the first data processing device ( Buyer) data processing and model training methods to assess the value of seller data to buyers.

Fig. 4 is an interactive flowchart of another data value evaluation method provided by the embodiment of the present application. The method flow in FIG. 4 is a possible implementation of the method described in FIG. 3 . As shown in Figure 4, the method includes:

401. The data evaluation device sends a homomorphic encryption program to the first data processing device and the second data processing device respectively.

The data evaluation device holds the private decryption key for the homomorphic encryption program.

402. The first data processing device sends an initial performance index and a homomorphically encrypted data processing and model training program to the data evaluation device.

The data processing and model training program is used to train the first business model. The first data processing device is to evaluate the value of the first local data of the second data processing device for training the first business model. After receiving the initial performance index, the data evaluation device may record the initial performance index. The data processing and model training program of homomorphic encryption refers to the data processing and model training program encrypted by homomorphic encryption program.

403. After detecting that the data processing and model training program of homomorphic encryption has no security problem, the data evaluation device sends the data processing and model training program of homomorphic encryption to the second data processing device.

Detecting whether there are security issues in the data processing and model training program of homomorphic encryption may include: checking whether the encryption algorithm used by the program complies with general security specifications, checking whether the program is complete, and scanning the program for viruses, etc.

404. The second data processing device uses a homomorphic encryption program to encrypt its first local data to obtain the first encrypted data, and then uses the homomorphic encryption data processing and model training program to train with the first encrypted data to obtain a second encrypted business model, And send the second encrypted business model to the data evaluation device.

405. After detecting that the second encrypted service model has no security problem, the data evaluation device sends the second encrypted service model to the second data processing device.

Detecting whether the second encrypted business model has a security problem may include: detecting whether the encryption algorithm used by the second encrypted business model complies with general security specifications, detecting whether the second encrypted business model is complete, and performing virus scanning on the second encrypted business model wait.

406. The first data processing device uses a homomorphic encryption program to encrypt its local test data to obtain encrypted test data; uses the encrypted test data to test a second encrypted service model to obtain first index information.

407. The data evaluation device decrypts the first index information by using the decryption private key corresponding to the homomorphic encryption program to obtain the first performance index.

408. The data evaluation device evaluates the value of the first local data for training the first service model according to the initial performance index and the first performance index.

Step 408 can refer to step 308 .

In this embodiment of the application, the second data processing device (seller) sends the second encrypted business model to the data evaluation device, and neither the data evaluation device nor the first data processing device (buyer) can obtain the second encrypted business model through the second encrypted business model. 2. The local data of the data processing device; it can meet the seller's data privacy and security protection requirements. That is to say, the seller's data (such as the first local data) will not leave the seller's control area, the detailed value of the seller's data is invisible to the buyer, and the use of the data can only be used to train the data business model. Buyer needs to train data business models using seller data. In the embodiment of the present application, the first data processing device sends the initial performance index and the data processing and model training program encrypted by the homomorphic encryption program to the data evaluation device; Model training methods assess the value of seller data to buyers.

Fig. 3 and Fig. 4 introduce and describe the interaction process of the data value evaluation method in which the data evaluation device, the first data processing device and the second data processing device jointly participate. The data value evaluation methods performed by the data evaluation device, the first data processing device and the second data processing device are respectively described below.

Fig. 5 is a flow chart of a data value evaluation method provided by the embodiment of the present application. Fig. 5 is a flow chart of the data value evaluation method performed by the data evaluation device. As shown in Figure 5, the method includes:

501. The data evaluation device sends first model information to a first data processing device.

The above-mentioned first model information represents the second encrypted service model obtained by the second data processing device using its first local data to train the first encrypted service model. The first index information may be obtained by the first data processing device testing the second encrypted service model with encrypted test data, and the encrypted test data is obtained by encrypting the test data. The second business model obtained by decrypting the second encrypted business model is the same as the business model after training the first business model with the first local data. The first business model is an unencrypted business model corresponding to the first encrypted business model.

In some embodiments, the data evaluation device may perform the following operations before performing step 501: receiving the second model information from the second data processing device; and obtaining the first model information according to the second model information. Obtaining the first model information according to the second model information may be: the data evaluation device decrypts the second model information using a decryption method pre-agreed with the second data processing device to obtain the first model information. The second model information may be obtained by the second data processing device encrypting the first model information using an encryption method pre-agreed with the data evaluation device.

In some embodiments, the data evaluation device may perform the following operation before performing step 501: receive the first model information from the above-mentioned second data processing device.

502. The data evaluation device obtains a first performance index according to the first index information from the first data processing device.

The above-mentioned first performance index represents the performance of the business model after the above-mentioned first business model is trained by using the above-mentioned first local data, and the above-mentioned first index information is the The corresponding test data are obtained. For step 502, refer to step 407.

503. The data evaluation device evaluates the value of the first local data for training the first service model according to the first performance index and the second performance index.

The above-mentioned second performance index represents the performance of the above-mentioned first business model.

In the embodiment of the present application, the data evaluation device sends the first model information to the first data processing device, and the data of the first data processing device can be limited to training the first business model, so as to achieve the purpose of not leaving the seller's data locally. The data evaluation device evaluates the value of the first local data for training the first business model according to the first performance index and the second performance index; it can accurately evaluate the value of the first local data for training the first business model.

Fig. 6 is a flow chart of another data value evaluation method provided by the embodiment of the present application. The method flow in FIG. 6 is a possible implementation of the method described in FIG. 5 . As shown in Figure 6, the method includes:

601. The data evaluation device sends first encrypted information to a first data processing device and a second data processing device respectively.

The first encryption information is used by the first data processing device to encrypt the fourth model information to obtain the third model information. The fourth model information is used to train the first service model. The first encryption information is used by the second data processing device to encrypt its first local data.

602. The data evaluation device receives third model information and a second performance index from the first data processing device.

The above third model information is used to train the first encrypted service model. The second performance indicator characterizes the performance of the first business model. The first business model is an unencrypted business model corresponding to the first encrypted business model. In some embodiments, the first encryption information may be a homomorphic encryption program; the third model information is that the first data processing device uses a homomorphic encryption program to encrypt the data processing and model training program used to train the first business model dealt with.

603. The data evaluation device sends third model information to the second data processing device.

604. The data evaluation device receives second model information from a second data processing device.

The above-mentioned second model information is used for the above-mentioned data evaluation device to obtain the first model information. The first model information represents the second encrypted service model obtained by the second data processing device using its first local data to train the first encrypted service model.

605. The data evaluation device obtains first model information according to the second model information.

Step 605 is optional but not necessary. In some embodiments, the data evaluation device receives the second model information from the second data processing device, and directly sends the first model information to the first data processing device, that is, executes step 606 . That is, in some embodiments, the first model information and the second model information are the same. In some embodiments, step 605 is implemented by: the data evaluation device decrypts the second model information to obtain the first model information.

606. The data evaluation device sends the first model information to the first data processing device.

607. The data evaluation device decrypts the first index information from the first data processing device to obtain the first performance index.

The above-mentioned first performance index represents the performance of the service model after the first service model is trained by using the first local data. The first index information is obtained by the first data processing device by using the first model information and test data corresponding to the second encrypted service model.

608. The data evaluation device evaluates the value of the first local data for training the first service model according to the first performance index and the second performance index.

In some embodiments, after performing step 608, the data evaluation device may further perform the following operation: the data evaluation device sends the decryption private key to the above-mentioned first data processing device. The decryption private key is used by the first data processing device to decrypt the second encrypted business model.

Fig. 7 is a flow chart of another data value evaluation method provided by the embodiment of the present application. Fig. 7 is a flow chart of the data value evaluation method performed by the first data processing device (buyer). As shown in Figure 7, the method includes:

701. A first data processing device receives first model information from a data evaluation device.

The above-mentioned first model information represents the second encrypted service model obtained by the second data processing device using its first local data to train the first encrypted service model. The second business model obtained by decrypting the above-mentioned second encrypted business model is the same as the business model after training the first business model with the first local data. The first business model is an unencrypted business model corresponding to the first encrypted business model.

702. The first data processing device sends the first indicator information to the data evaluation device according to the first model information and the test data corresponding to the second encrypted service model.

The above-mentioned first index information is used for the above-mentioned data evaluation device to obtain the first performance index. The above-mentioned first performance index represents the performance of the service model after the above-mentioned first service model is trained by using the above-mentioned first local data.

Fig. 8 is a flow chart of another data value evaluation method provided by the embodiment of the present application. The method flow in FIG. 8 is a possible implementation of the method described in FIG. 7 . As shown in Figure 8, the method includes:

801. A first data processing device receives first encrypted information from a data evaluation device.

802. The first data processing device encrypts fourth model information by using the first encrypted information to obtain third model information.

The above fourth model information is used to train the first service model. The third model information is used to train the first encrypted business model. In some embodiments, the first encryption information is a homomorphic encryption program, the fourth model information is a data processing and model training program used to train the first business model, and the third model information is the use of a homomorphic encryption program to encrypt the fourth model Homomorphically encrypted data processing and model training programs obtained by encrypting information. The first encrypted business model can be regarded as obtained by encrypting the first business model by using a homomorphic encryption program.

803. The first data processing device sends the third model information and the second performance index to the data evaluation device.

The above-mentioned second performance index represents the performance of the first business model. The above-mentioned second performance index is used by the data evaluation device to evaluate the value of the first local data of the second data processing device for training the first business model.

804. The first data processing device receives first model information from the data evaluation device.

For step 804, refer to step 701.

805. The first data processing device encrypts the test data used for testing the first service model by using the first encryption information to obtain encrypted test data.

806. The first data processing device uses the encrypted test data to test the second encrypted service model to obtain the first indicator information.

The first index information is used by the data evaluation device to obtain a first performance index; the first performance index represents the performance of the service model after the first service model is trained using the first local data.

807. The first data processing device sends the first indicator information to the data evaluation device.

In some embodiments, the first data processing device may also perform the following operations: the first data processing device decrypts the second encrypted business model by using the decryption private key to obtain the second business model. The second business model is the same as the business model after training the first business model with the above-mentioned first local data.

In the embodiment of the present application, the first data processing device sends the first index information, the third model information and the second performance index to the data evaluation device, so that the data evaluation device can accurately evaluate the value of the first local data for training the first business model . The first data processing device sends the first index information, the third model information, and the first business model whose second performance index is not exposed and a program for training the first business model to the data evaluation device. The first data processing device receives the first model information from the data evaluation device, and the first data processing device cannot obtain the first local data of the second data processing device according to the first model information, which can prevent the first local data from being exposed to A first data processing device.

FIG. 9 is a flow chart of another data value evaluation method provided by the embodiment of the present application. FIG. 9 is a flow chart of the data value evaluation method performed by the second data processing device (seller). As shown in Figure 9, the method includes:

901. The second data processing device receives third model information from the data evaluation device.

The above third model information is used to train the first encrypted service model. The above-mentioned first encrypted business model is the same as the business model obtained by encrypting the first business model belonging to the first data processing device.

902. The second data processing device sends second model information to the data evaluating device.

The above-mentioned second model information is used by the data evaluation device to obtain the first model information. The above-mentioned first model information represents the second encrypted service model obtained by the second data processing device using its first local data to train the first encrypted service model. The second business model obtained by decrypting the above-mentioned second encrypted business model is the same as the business model after training the first business model with the first local data.

Fig. 10 is a flowchart of another data value evaluation method provided by the embodiment of the present application. The method flow in FIG. 10 is a possible implementation of the method described in FIG. 9 . As shown in Figure 10, the method includes:

1001. A second data processing device receives first encrypted information from a data evaluation device.

For step 1001, refer to step 801.

1002. The second data processing device encrypts the first local data by using the first encryption information to obtain the first encrypted data.

1003. The second data processing device receives third model information from the data evaluation device.

For step 1003, refer to step 901. The sequence of step 1003 and step 1002 is not limited.

1004. The second data processing device uses the first encrypted data and the third model information to train the first encrypted service model to obtain a second encrypted service model.

1005. The second data processing device obtains the second model information according to the second encrypted service model.

The second model information is used by the data evaluation device to obtain the first model information. The above-mentioned first model information represents the second encrypted service model obtained by the second data processing device using its first local data to train the first encrypted service model.

1006. The second data processing device sends second model information to the data evaluation device.

In the embodiment of the present application, the second data processing device sends the second model information to the data evaluation device, which can not only avoid exposing its own local data, but also enable the data evaluation device to evaluate the value of the first local data for training the first business model .

The interaction process of a data usage right delivery method provided by the embodiment of the present application is described below with reference to the accompanying drawings.

FIG. 11 is an interactive flow chart of a data usage right delivery method provided by an embodiment of the present application. As shown in Figure 11, the method includes:

1101. The third data processing device acquires a homomorphic encryption program and a decryption private key from the data delivery device.

Step 1101 can be understood as: the data delivery device sends the homomorphic encryption program and the decryption private key to the third data processing device. The decryption private key is used to decrypt data encrypted using a homomorphic encryption program.

1102. The third data processing device uses a homomorphic encryption program to encrypt the data processing and model training program to obtain a homomorphic encrypted data processing and model training program.

The data processing and model training program is used to train the business model that the third data processing device needs the assistance of the fourth data processing device to train.

1103. The third data processing device sends a homomorphically encrypted data processing and model training program to the data delivery device.

1104. After detecting that the data processing and model training program of homomorphic encryption has no security problem, the data delivery device sends the data processing and model training program of homomorphic encryption to the fourth data processing device.

In practical applications, the data delivery device may send the homomorphically encrypted data processing and model training program to one or more sellers after detecting that the homomorphically encrypted data processing and model training program has no security problems. It should be understood that the fourth data processing device is only an example of a seller. That is to say, the buyer can purchase the right to use the data of multiple sellers through the method flow in FIG. 11 .

1105. The fourth data processing device encrypts its second local data using a homomorphic encryption program to obtain second encrypted data, and then uses a homomorphic encryption data processing and model training program to train with the second encrypted data to obtain a fourth encrypted business model.

The data processing and model training program can be used to train the third business model, and the data processing and model training program of homomorphic encryption is used to train the third encrypted business model, that is, the third business model of homomorphic encryption. The fourth encryption business model is a homomorphic encryption business model. Using the decryption private key corresponding to the homomorphic encryption program to encrypt the fourth business model can obtain the unencrypted fourth business model. The fourth business model is equal to the business model trained by using the data processing and model training program and the second local data.

In some embodiments, the data delivery device sends the homomorphic encryption program to the fourth data processing device while sending the homomorphic encrypted data processing and model training program to the fourth data processing device. In some embodiments, the data delivery device may send the homomorphic encrypted data processing and model training program and the homomorphic encryption program to the fourth data processing device through different messages.

1106. The fourth data processing device sends the fourth encrypted service model to the third data processing device.

A possible implementation manner of step 1106 is: the fourth data processing device sends the sixth model information to the third data processing device. The sixth model information characterizes the above-mentioned fourth encrypted business model. Alternatively, the sixth model information includes parameters of the fourth encrypted service model.

1107. The third data processing device decrypts the fourth encrypted business model by using the decryption private key to obtain the fourth business model.

The fourth service model is the same as the service model after using the second local data to train the third service model. The above-mentioned third business model is an unencrypted business model corresponding to the third encrypted business model, and is a business model for which the third data processing device requires the assistance of the fourth data processing device for training.

1108. The third data processing device updates its local service model by using the parameters of the fourth service model.

The method flow in FIG. 11 and the method flow in FIG. 4 may be independent method flows, or may be two associated flows. In some embodiments, the data delivery device may be a data evaluation device, the third data processing device may be a first data processing device, and the fourth data processing device may be a second data processing device; the data evaluation device, the first data processing device And the second data processing device can first evaluate the value of some local data of the seller for the buyer's training business model through the method flow in Figure 4, and then use the method flow in Figure 11 to purchase these local data or include the use of these local data right. For example, the second data processing device is to sell the right to use a certain piece of local data, and the data volume of the local data is 100. The data evaluation device, the first data processing device and the second data processing device can first pass through the The method flow evaluates any part of the local data with a data volume of 1; after the buyer and the seller conclude a transaction, the right to use the local data is delivered by executing the method flow in Figure 11 . By evaluating a portion of the entire local dataset, the time spent on evaluation can be reduced.

In the embodiment of the present application, the third data processing device uses the decryption private key to decrypt the fourth encrypted business model to obtain the fourth business model; without obtaining the local data of the fourth data processing device, it can use The local data of the fourth data processing device implements model training. In the embodiment of this application, what is delivered is the right to use the data, the seller's data does not leave the local, and the use of the data is limited to training the data business model. On the premise that the detailed value of the seller's data is completely invisible to the buyer, it helps the buyer to realize the training of the model. Under the premise of satisfying the privacy and security protection of the seller's data, the potential value of the seller's data can be brought into play.

FIG. 11 introduces and describes the interaction process of the data value evaluation method in which the data delivery device, the third data processing device and the fourth data processing device participate together. The data usage right delivery methods performed by the data delivery device, the third data processing device, and the fourth data processing device are respectively described below.

FIG. 12 is a flow chart of a data usage right delivery method provided by an embodiment of the present application. Fig. 12 is a data usage right delivery process executed by the data delivery device. As shown in Figure 12, the method includes:

1201. The data delivery device receives fifth model information from a third data processing device.

The fifth model information is used to train the third encrypted service model. The business model obtained by decrypting the third encrypted business model is the same as the unencrypted third business model trained by the third data processing device with assistance from the fourth data processing device.

In a possible implementation manner, before performing step 1201, the data delivery device may send first encrypted information to the third data processing device, where the first encrypted information is used by the third data processing device to generate fifth model information. In some embodiments, the first encrypted information is a homomorphic encryption program, and the third data processing device uses the homomorphic encryption program to encrypt the data processing and model training program to obtain a homomorphic encrypted data processing and model training program ( i.e. fifth model information).

1202. The data delivery device sends fifth model information to the fourth data processing device.

A possible implementation of step 1202 is as follows: in the case of detecting that there is no security problem in the fifth model information, the data delivery device sends the fifth model information to the fourth data processing device.

Detecting whether the fifth model information has a security problem may include: detecting whether the encryption algorithm used by the fifth model information complies with general security regulations, detecting whether the fifth model information is complete, performing virus scanning on the fifth model information, and so on.

In a possible implementation manner, the data delivery device may also perform the following operations: the above-mentioned data delivery device sends a decryption private key to the above-mentioned third data processing device; the above-mentioned decryption private key is used by the above-mentioned third data processing device to encrypt the fourth The business model is decrypted. The fourth encrypted service model is obtained by the fourth data processing device using the second encrypted data to train the third encrypted service model. The above-mentioned second encrypted data is obtained by encrypting the second local data of the above-mentioned fourth data processing device. The fourth service model obtained by decrypting the fourth encrypted service model is the same as the service model after training the third service model by using the second local data.

In a possible implementation manner, the data delivery device may further perform the following operation: send first encrypted information to the fourth data processing device, where the first encrypted information is used by the fourth data processing device to encrypt its second local data. The fourth data processing device encrypts its second local data by using the first encryption information to obtain second encrypted data.

FIG. 13 is a flow chart of another data usage right delivery method provided by the embodiment of the present application. Fig. 13 is a data usage right delivery process executed by the third data processing device (buyer). As shown in Figure 13, the method includes:

1301. The third data processing device receives the decryption private key from the data delivery device.

1302. The third data processing device decrypts the fourth encrypted business model by using the decryption private key to obtain the fourth business model.

The fourth encrypted business model is obtained by the fourth data processing device using the second encrypted data to train the third encrypted business model. The above-mentioned second encrypted data is obtained by encrypting its second local data by the above-mentioned fourth data processing device. The fourth service model is the same as the service model after the third service model is trained by using the second local data. The third business model is an unencrypted business model corresponding to the third encrypted business model, and is a business model for which the third data processing device requires assistance from the fourth data processing device for training.

In a possible implementation manner, before the third data processing device performs step 1302, the following operations may be performed: the third data processing device sends fifth model information to the data delivery device; the fifth model information is used to train the third An encrypted business model; a fourth encrypted business model is obtained according to the sixth model information from the fourth data processing device. Before the third data processing device sends the fifth model information to the data delivery device, the following operations may be performed: the third data processing device acquires the first encrypted information from the data delivery device; and encrypts the seventh model information according to the first encrypted information , to obtain the fifth model information; the seventh model information is used to train the third service model.

In a possible implementation manner, after performing step 1302, the third data processing device may perform the following operation: the third data processing device updates its local service model by using the parameters of the fourth service model.

FIG. 14 is a flow chart of another method for delivering data usage rights provided by the embodiment of the present application. Fig. 14 is a data usage right transfer process executed by the fourth data processing device (seller). As shown in Figure 14, the method includes:

1401. The fourth data processing device receives fifth model information from the data delivery device.

The above fifth model information is used to train the third encrypted service model.

1402. The fourth data processing device uses the second encrypted data to train a third encrypted service model to obtain a fourth encrypted service model.

The above-mentioned second encrypted data is obtained by encrypting its second local data by the above-mentioned fourth data processing device. The fourth service model obtained by decrypting the fourth encrypted service model is the same as the service model after training the third service model by using the second local data. The third business model is an unencrypted business model corresponding to the third encrypted business model, and is a business model for which the third data processing device requires assistance from the fourth data processing device for training.

Before the fourth data processing device executes step 1402, the following operations may be performed: the fourth data processing device encrypts the second local data by using the first encryption information from the data delivery device to obtain the second encrypted data. In some embodiments, the first encryption information is a homomorphic encryption program, and the fifth model information is a homomorphic encryption data processing and model training program. The data processing and model training program of homomorphic encryption may be a program obtained by using the homomorphic encryption program (ie, the first encrypted information) to perform encryption processing on the data processing and model training program used to train the third business model.

1403. The fourth data processing device sends sixth model information to the third data processing device.

The sixth model information represents the fourth encrypted business model; the fourth encrypted business model is used by the third data processing device to decrypt to obtain required model update parameters.

In the embodiment of the present application, the fourth data processing device receives the fifth model information from the data delivery device; the fourth data processing device uses the second encrypted data to train the third encrypted business model to obtain the fourth encrypted business model; The business models and model training methods of the three data processing devices were leaked. The fourth data processing device sends the sixth model information to the third data processing device, which can prevent the third data processing device from obtaining the local data of the fourth data processing device. That is to say, on the premise that the local data of the fourth data processing device is completely invisible to the third data processing device, it can help the third data processing device to realize the training of the business model.

The above describes the scheme in which the buyer and the seller evaluate the value of the right to use the seller's data to the buyer through the data evaluation device (corresponding to the data value evaluation platform), and the buyer and the seller deliver the seller's data through the data delivery device (corresponding to the data usage right delivery platform). Right of use to Buyer's program. In practical applications, the data evaluation device and the data delivery device may be the same device or product, or different devices or products. That is to say, the data value evaluation platform and the data usage right delivery platform can be the same platform, or they can be two different platforms. In the foregoing embodiments, data value evaluation and data usage right delivery are described as two independent processes. The following takes the implementation of a data trading platform product that has both data value assessment and data usage rights delivery as an example to describe the solution of first evaluating data value and then delivering data usage rights.

Fig. 15 is an interaction flowchart of a data evaluation and use right delivery method provided by the embodiment of the present application. As shown in Figure 15, the method includes:

1501. The second data processing apparatus releases introduction information of its local data whose use rights are to be sold.

A possible implementation of step 1501 is as follows: the second data processing device (seller) publishes the source of the data it intends to sell the use right on the data transaction system (such as the vehicle data transaction system), the time when the data is generated, the data structure, and the field Data introduction information such as type and data label. The data transaction system can run on the data evaluation device (data delivery device). The second data processing device (seller) publishes presentation information of its local data for which the right to use is to be sold on the data transaction system operated by the data evaluation device.

1502. The first data processing device initiates value evaluation for the first local data to the data evaluation device.

The first local data may be a piece of local data for sale issued by the second data processing device, or may be a part of a piece of local data for sale issued by the second data processing device. The first local data may include two or more pieces of local data for sale issued by the second data processing device, for example, two pieces of local data with different data sources.

In some embodiments, after the buyer finds the data he is interested in from the data transaction system, he can initiate a value assessment for the data to the data transaction system.

1503. The data evaluation device calculates and outputs the value of the first local data for the buyer model based on the percentage of performance improvement after the buyer model is trained using the first local data.

The implementation manner of step 1503 may refer to FIG. 3 or FIG. 4 .

1504. The first data processing device and the second data processing device conduct a communication transaction based on the percentage of the first local data calculated by the data evaluation device for the performance improvement of the buyer model and the reference value of the data.

Step 1504 can be understood as: the buyer and the seller conduct a communication transaction based on the percentage of the first local data calculated by the data evaluation device for the performance improvement of the buyer's model and the data reference value.

If the transaction is concluded, execute step 1505; if the transaction is not concluded, the data transaction is terminated (or this process ends).

1505. The first data processing device pays the second data processing device to acquire the right to use the first local data.

Step 1505 can be understood as: the buyer pays the seller to obtain the right to use the first local data.

1506. The second data processing device participates in the federated learning architecture led by the first data processing device, and helps the first data processing device train the service model.

For step 1506, refer to the method flow in FIG. 11 . Step 1506 can be understood as: the first local data participates in the buyer-led federated learning framework to help the buyer train the business model. That is to say, after the transaction is concluded, the buyer obtains the right to use the seller's data, and the buyer can build a federated learning architecture based on the data transaction system, and use the seller's data to train its own data business model. In the embodiment of this application, the federated learning architecture is improved, and the buyer purchases the seller's data to participate in the federated learning architecture led by it. Achieving technical effect: On the premise of meeting the seller's data privacy and security protection requirements, the data buyer's training data business model needs are met, and the potential value of the originally immobile data can be brought into play.

In some embodiments, the buyer can purchase multiple data usage rights from multiple sellers, build a federated learning framework based on the data transaction system, and use the seller's data to train its own data business model.

The data value assessment method and the data usage right delivery method in the embodiment of this application are not strongly coupled. The data value assessment method and the data usage right delivery method can be developed as an independent product, or the functions of the two can be centralized A data trading platform is realized.

In the embodiment of the present application, the data evaluation device calculates and outputs the value of the first local data for the buyer model based on the percentage of performance improvement after the buyer model is trained using the first local data; it is possible for the buyer not to expose its data processing and model training methods, On the premise that the seller does not disclose its data details, output the percentage and data reference value of the seller's data to the buyer's model performance improvement. In addition, using the data trading system provided by the embodiment of this application, what is delivered is the right to use the data, the seller's data does not leave the local, and the use of the data is limited to training data business models. On the premise that the detailed value of the seller's data is completely invisible to the buyer, it helps the buyer to realize the training of the model. Under the premise of satisfying the privacy and security protection of the seller's data, the potential value of the seller's data can be brought into play.

FIG. 16 shows a schematic structural diagram of a data processing device 1600 . The data processing device 1600 can correspond to any one of the data evaluation device, the first data processing device, the second data processing device, the data delivery device, the third data processing device, and the fourth data processing device in the above-mentioned method embodiments. The functions or steps to be realized. The data processing device may include a processing module 1610 and a transceiver module 1620 . Optionally, a storage unit may also be included, and the storage unit may be used to store instructions (code or program) and/or data. The processing module 1610 and the transceiver module 1620 may be coupled with the storage unit, for example, the processing module 1610 may read instructions (code or program) and/or data in the storage unit to implement a corresponding method. Each of the above units can be set independently, or can be partially or fully integrated. For example, the transceiving module 1620 may include a sending module and a receiving module.

In some possible implementation manners, the data processing device 1600 can correspondingly implement the behaviors and functions of the data evaluation device in the foregoing method embodiments. For example, the data processing device 1600 may be a data evaluation device, or a component (such as a chip or a circuit) applied in the data evaluation device. The transceiver module 1620, for example, can be used to perform all the receiving or sending operations performed by the data evaluation device in the embodiments shown in FIG. 3, FIG. 4, FIG. 5, FIG. 6, and FIG. 15, such as the steps in the embodiment shown in FIG. The receiving or sending operations involved in steps 301 to 306 and the receiving or sending operations involved in steps 401 to 406 in the embodiment shown in FIG. 4 , and/or other processes for supporting the technology described herein. The processing module 1610 is used to execute all the operations performed by the data evaluation device in the embodiments of FIG. 3 , FIG. 4 , FIG. 5 , FIG. 6 , and FIG. 15 , except the transceiving operation.

In some possible implementation manners, the data processing device 1600 can correspondingly implement the behaviors and functions of the first data processing device in the foregoing method embodiments. For example, the data processing device 1600 may be a first data processing device, or may be a component (such as a chip or a circuit) applied in the first data processing device. The transceiver module 1620, for example, may be used to perform all the receiving or sending operations performed by the first data processing device in the embodiment shown in FIG. 3, FIG. 4, FIG. 7, FIG. 8 or FIG. Step 301 , step 302 , step 305 , step 306 and step 401 , step 402 , step 405 , step 406 in the embodiment shown in FIG. 4 , and/or other processes for supporting the technology described herein. The processing module 1610 is configured to execute all the operations performed by the first data processing device in the embodiment shown in FIG. 3 , FIG. 4 , FIG. 7 , FIG. 8 or FIG. 15 except the transceiving operation.

In some possible implementation manners, the data processing device 1600 can correspondingly implement the behaviors and functions of the second data processing device in the foregoing method embodiments. For example, the data processing device 1600 may be a second data processing device, or may be a component (such as a chip or a circuit) applied in the second data processing device. The transceiver module 1620, for example, may be used to perform all the receiving or sending operations performed by the second data processing device in the embodiment shown in FIG. 3, FIG. 4, FIG. 9, FIG. 10 or FIG. Step 301 , step 303 , step 304 and step 401 , step 403 , step 404 in the embodiment shown in FIG. 4 , and/or other processes for supporting the technology described herein. The processing module 1610 is configured to perform all the operations performed by the second data processing device in the embodiment shown in FIG. 3 , FIG. 4 , FIG. 9 , FIG. 10 or FIG. 15 , except the transceiving operation.

In some possible implementation manners, the data processing apparatus 1600 can correspondingly implement the behaviors and functions of the data delivery apparatus in the foregoing method embodiments. For example, the data processing device 1600 may be a data delivery device, or a component (such as a chip or a circuit) applied in the data delivery device. The transceiver module 1620, for example, can be used to perform all the receiving or sending operations performed by the data delivery device in the embodiment shown in FIG. 11 and FIG. Step 1201 and step 1202 in an embodiment, and/or other processes used to support the techniques described herein. The processing module 1610 is configured to perform all operations performed by the data delivery device in the embodiments shown in FIG. 11 and FIG. 12 except for the transceiving operation.

In some possible implementation manners, the data processing apparatus 1600 can correspondingly implement the behaviors and functions of the third data processing apparatus in the foregoing method embodiments. For example, the data processing device 1600 may be a third data processing device, or may be a component (such as a chip or a circuit) applied in the third data processing device. The transceiver module 1620, for example, may be used to perform all the receiving or sending operations performed by the third data processing device in the embodiment shown in FIG. 11 and FIG. And step 1301 in the embodiment shown in FIG. 13 , and/or other processes for supporting the techniques described herein. The processing module 1610 is configured to perform all the operations performed by the third data processing device in the embodiments shown in FIG. 11 and FIG. 13 except the transceiving operation.

In some possible implementation manners, the data processing apparatus 1600 can correspondingly implement the behavior and functions of the fourth data processing apparatus in the foregoing method embodiments. For example, the data processing device 1600 may be a fourth data processing device, or may be a component (such as a chip or a circuit) applied in the fourth data processing device. The transceiver module 1620, for example, may be used to perform all the receiving or sending operations performed by the fourth data processing device in the embodiment shown in FIG. 11 and FIG. Step 1401 and Step 1403 in the illustrated embodiment, and/or other processes used to support the techniques described herein. The processing module 1610 is configured to execute all the operations performed by the fourth data processing device in the embodiments shown in FIG. 11 and FIG. 14 except the transceiving operation.

FIG. 17 is a schematic structural diagram of another data processing device 170 provided in an embodiment of the present application. The data processing means in FIG. 17 may be the data evaluation means described above. The data processing device in FIG. 17 may be the first data processing device described above. The data processing device in FIG. 17 may be the second data processing device described above. The data processing device in FIG. 17 may be the data delivery device described above. The data processing device in FIG. 17 may be the third data processing device described above. The data processing device in FIG. 17 may be the fourth data processing device described above.

As shown in FIG. 17 , the data processing device 170 includes at least one processor 1720 and a transceiver 1710 .

In some other embodiments of the present application, the processor 1720 and the transceiver 1710 may be used to perform the functions or operations performed by the above-mentioned data evaluation device. The transceiver 1710 may perform all receiving or transmitting operations performed by the data evaluation device. The processor 1720 can, for example, execute all the operations performed by the data evaluation device in the embodiments of FIG. 3 , FIG. 4 , FIG. 5 , FIG. 6 , and FIG. 15 , except the transceiving operation.

In some other embodiments of the present application, the processor 1720 and the transceiver 1710 may be configured to perform the functions or operations performed by the above-mentioned first data processing device. The transceiver 1710 may perform all reception or transmission operations performed by the first data processing means. The processor 1720 may, for example, perform all the operations performed by the first data processing device in the embodiment shown in FIG. 3 , FIG. 4 , FIG. 7 , FIG. 8 or FIG. 15 except the transceiving operation.

In some other embodiments of the present application, the processor 1720 and the transceiver 1710 may be used to perform the functions or operations performed by the above-mentioned second data processing device. The transceiver 1710 may perform all reception or transmission operations performed by the second data processing means. The processor 1720 can, for example, execute all the operations performed by the second data processing device in the embodiment shown in FIG. 3 , FIG. 4 , FIG. 9 , FIG. 10 or FIG. 15 except the transceiving operation.

In some other embodiments of the present application, the processor 1720 and the transceiver 1710 may be configured to perform the functions or operations performed by the above data delivery device. The transceiver 1710 may perform all receive or transmit operations performed by the data delivery device. The processor 1720 may, for example, perform all the operations performed by the data delivery device in the embodiments of FIG. 11 and FIG. 12 except the transceiving operation.

In some other embodiments of the present application, the processor 1720 and the transceiver 1710 may be configured to perform the functions or operations performed by the above-mentioned third data processing device. The transceiver 1710 may perform all reception or transmission operations performed by the third data processing means. The processor 1720 may, for example, perform all the operations performed by the third data processing device in the embodiments of FIG. 11 and FIG. 13 except the transceiving operation.

In some other embodiments of the present application, the processor 1720 and the transceiver 1710 may be configured to perform the functions or operations performed by the fourth data processing device. The transceiver 1710 may perform all reception or transmission operations performed by the fourth data processing means. The processor 1720 may, for example, execute all the operations performed by the fourth data processing device in the embodiments of FIG. 11 and FIG. 14 except the transceiving operation.

Transceiver 1710 is used to communicate with other devices/devices over transmission media. The processor 1720 uses the transceiver 1710 to send and receive data and/or signaling, and is used to implement the methods in the foregoing method embodiments. The processor 1720 can realize the function of the processing module 1610 , and the transceiver 1710 can realize the function of the transceiver module 1620 .

Optionally, the data processing device 170 may further include at least one memory 1730 for storing program instructions and/or data. The memory 1730 is coupled to the processor 1720 . The coupling in the embodiments of the present application is an indirect coupling or a communication connection between devices, units or modules, which may be in electrical, mechanical or other forms, and is used for information exchange between devices, units or modules. Processor 1720 may cooperate with memory 1730 . Processor 1720 may execute program instructions stored in memory 1730 . At least one of the at least one memory may be included in the processor.

In this embodiment of the present application, a specific connection medium among the transceiver 1710, the processor 1720, and the memory 1730 is not limited. In the embodiment of the present application, in FIG. 17, the memory 1730, the processor 1720, and the transceiver 1710 are connected through a bus 1740. The bus is represented by a thick line in FIG. 17, and the connection between other components is only for schematic illustration. , is not limited. The bus can be divided into address bus, data bus, control bus and so on. For ease of representation, only one thick line is used in FIG. 17 , but it does not mean that there is only one bus or one type of bus.

In this embodiment of the application, the processor may be a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field programmable gate array or other programmable logic device, a discrete gate or transistor logic device, or a discrete hardware component, and may implement or Execute the methods, steps and logic block diagrams disclosed in the embodiments of the present application. A general purpose processor may be a microprocessor or any conventional processor or the like. The steps of the methods disclosed in connection with the embodiments of the present application may be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules in the processor.

FIG. 18 is a schematic structural diagram of another data processing device 180 provided by an embodiment of the present application. As shown in FIG. 18 , the data processing device shown in FIG. 18 includes a logic circuit 1801 and an interface 1802 . The processing module 1610 in FIG. 16 can be realized by a logic circuit 1801 , and the transceiver module 1620 in FIG. 16 can be realized by an interface 1802 . Wherein, the logic circuit 1801 may be a chip, a processing circuit, an integrated circuit or a system on chip (SoC) chip, etc., and the interface 1802 may be a communication interface, an input-output interface, or the like. In the embodiment of the present application, the logic circuit and the interface may also be coupled to each other. The embodiment of the present application does not limit the specific connection manner of the logic circuit and the interface.

In some embodiments of the present application, the logic circuit and interface can be used to perform the functions or operations performed by the above-mentioned data evaluation device.

In some other embodiments of the present application, the logic circuit and the interface may be used to perform the functions or operations performed by the above-mentioned first data processing device.

In some other embodiments of the present application, the logic circuit and the interface may be used to perform the functions or operations performed by the above-mentioned second data processing device.

In some embodiments of the present application, the logic circuit and interface may be used to perform the functions or operations performed by the above-mentioned data delivery device.

In some other embodiments of the present application, the logic circuit and the interface may be used to perform the functions or operations performed by the above-mentioned third data processing device.

In some other embodiments of the present application, the logic circuit and the interface may be used to perform the functions or operations performed by the above-mentioned fourth data processing device.

The present application also provides a computer-readable storage medium, where computer codes are stored in the computer-readable storage medium, and when the computer codes are run on the computer, the computer is made to execute the methods of the above-mentioned embodiments.

The present application also provides a computer program product, the computer program product includes computer code or computer program, and when the computer code or computer program is run on a computer, the methods in the above-mentioned embodiments are executed.

The present application also provides a data value evaluation system, including a data evaluation device, a first data processing device, and a second data processing device.

The present application also provides a data usage right delivery system, including a data delivery device, a third data processing device, and a fourth data processing device.

The above is only the specific implementation of the embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed in the present application. , should be covered within the protection scope of this application. Therefore, the protection scope of the present application should be based on the protection scope of the above claims.

Claims

A data value evaluation method, characterized in that it includes:

The data evaluation device sends the first model information to the first data processing device; the first model information represents the second encrypted business model obtained by the second data processing device using its first local data to train the first encrypted business model; the first The second business model obtained by decrypting the second encrypted business model is the same as the business model after using the first local data to train the first business model, and the first business model is the unencrypted business corresponding to the first encrypted business model Model;

The data evaluation device obtains a first performance index according to the first index information from the first data processing device; the first performance index represents the business after training the first business model with the first local data The performance of the model, the first index information is obtained by the first data processing device using the first model information and the test data corresponding to the second encrypted business model;

The data evaluation device evaluates the value of the first local data for training the first business model according to the first performance index and the second performance index; the second performance index represents the performance of the first business model performance.
The method according to claim 1, wherein the first index information is obtained by the first data processing device using encrypted test data to test the second encrypted business model, and the encrypted test data encrypts the test data obtained.
The method according to claim 1 or 2, wherein, before the data evaluation device sends the first model information to the first data processing device, the method further comprises:

The data evaluation device sends first encrypted information to the second data processing device; the first encrypted information is used by the second data processing device to encrypt the first local data, and the second encrypted business model is determined by The second data processing device uses the first encrypted data and the third model information to train the first encrypted business model, and the first encrypted data is obtained by encrypting the first local data with the first encrypted information .
The method according to any one of claims 1 to 3, wherein, before the data evaluation device receives the third model information from the first data processing device, the method further comprises:

The data evaluation device sends first encrypted information to the first data processing device; the first encrypted information is used by the first data processing device to encrypt fourth model information to obtain the third model information, the The fourth model information is used to train the first service model.
The method according to any one of claims 1 to 4, wherein the obtaining of the first performance index by the data evaluation device according to the first index information from the first data processing device includes:

The data evaluation device decrypts the first index information to obtain the first performance index.
The method according to any one of claims 1 to 5, wherein the data evaluation device evaluates the effectiveness of the first local data for training the first business model according to the first performance index and the second performance index After value, the method also includes:

The data evaluation device sends a decryption private key to the first data processing device; the decryption private key is used by the first data processing device to decrypt the second encrypted business model.
A data value evaluation method, characterized in that it includes:

The first data processing device receives the first model information from the data evaluation device; the first model information represents the second encrypted business model obtained by the second data processing device using its first local data to train the first encrypted business model; the The second business model obtained by decrypting the second encrypted business model is the same as the business model after using the first local data to train the first business model, and the first business model is the unencrypted corresponding to the first encrypted business model business model;

The first data processing device sends first index information to the data evaluation device according to the first model information and the test data corresponding to the second encrypted business model; the first index information is used for the data The evaluation device obtains a first performance index; the first performance index represents the performance of the business model after the first business model is trained by using the first local data.
The method according to claim 7, wherein the first data processing device sends the first index to the data evaluation device according to the first model information and the test data corresponding to the second encrypted business model Information includes:

The first data processing device uses the encrypted test data to test the second encrypted business model to obtain the first index information; the encrypted test data is obtained by encrypting the test data with the first encrypted information;

The first data processing means sends the first index information to the data evaluation means.
The method according to claim 8, wherein before the first data processing device sends the third model information to the data evaluation device, the method further comprises:

said first data processing means receives first encrypted information from said data evaluation means;

The first data processing device encrypts fourth model information by using the first encrypted information to obtain the third model information, and the fourth model information is used for training the first service model.
The method according to any one of claims 7 to 9, characterized in that, after sending the first indicator information to the data evaluation device, the method further comprises:

said first data processing means receives a decryption private key from said data evaluation means;

The first data processing device uses the decryption private key to decrypt the second encrypted business model to obtain the second business model.
A data value evaluation method, characterized in that it includes:

The second data processing device receives the third model information from the data evaluation device; the third model information is used to train the first encrypted business model, and the first encrypted business model is related to the first business belonging to the first data processing device The business model obtained by encrypting the model is the same;

The second data processing device sends second model information to the data evaluation device; the second model information is used by the data evaluation device to obtain first model information, and the first model information represents the second data The processing device uses its first local data to train the second encrypted business model obtained from the first encrypted business model; the second encrypted business model obtained by decrypting the second encrypted business model is the same as Business model after business model is the same.
The method according to claim 11, characterized in that after the second data processing device receives the third model information from the data evaluation device, the second data processing device sends the second model to the data evaluation device Before the information, the method also includes:

The second data processing device uses the first encrypted data and the third model information to train the first encrypted business model to obtain the second encrypted business model; the first encrypted data is provided by the first local The data is obtained through encryption;

The second data processing device obtains the second model information according to the second encrypted service model.
The method according to claim 12, wherein the second data processing device uses the first encrypted data and the third model information to train the first encrypted business model, and before obtaining the second encrypted model, the Methods also include:

said second data processing means receives said first encrypted information from said data evaluation means;

The second data processing device encrypts the first local data by using the first encryption information to obtain the first encrypted data.
The method according to any one of claims 11 to 13, wherein after the second data processing device sends the second model information to the data evaluation device, the method further comprises:

The second data processing device receives third model information from the data evaluation device; the third model information is used to train the first encrypted business model;

The second data processing device uses the first local data to train the first encrypted business model to obtain the second encrypted business model;

The second data processing device sends second model information to the first data processing device; the second model information is used by the first data processing device to decrypt to obtain parameters of the second service model.
A data value evaluation device, characterized in that it includes:

The transceiver module is configured to send the first model information to the first data processing device; the first model information represents the second encrypted business model obtained by the second data processing device using its first local data to train the first encrypted business model; The second business model obtained by decrypting the second encrypted business model is the same as the business model after using the first local data to train the first business model, and the first business model is the unencrypted version corresponding to the first encrypted business model. business model;

A processing module, configured to obtain a first performance index according to the first index information from the first data processing device; the first performance index represents the business after training the first business model with the first local data The performance of the model, the first index information is obtained by the first data processing device using the first model information and the test data corresponding to the second encrypted business model;

The processing module is further configured to evaluate the value of the first local data for training the first business model according to the first performance index and the second performance index; the second performance index characterizes the first Performance of the business model.
A data processing device, characterized in that it comprises:

The transceiver module is configured to receive the first model information from the data evaluation device; the first model information represents the second encrypted business model obtained by the second data processing device using its first local data to train the first encrypted business model; the The second business model obtained by decrypting the second encrypted business model is the same as the business model after using the first local data to train the first business model, and the first business model is the unencrypted corresponding to the first encrypted business model business model;

A processing module, configured to control the transceiver module to send first index information to the data evaluation device according to the first model information and the test data corresponding to the second encrypted business model; the first index information is used for The data evaluation device obtains a first performance index; the first performance index represents the performance of the business model after the first business model is trained by using the first local data.
A data processing device, characterized in that it comprises:

The transceiver module is used to receive the third model information from the data evaluation device; the third model information is used to train the first encrypted business model, and the first encrypted business model is related to the first business belonging to the first data processing device The business model obtained by encrypting the model is the same;

The transceiver module is further configured to send second model information to the data evaluation device; the second model information is used by the data evaluation device to obtain first model information, and the first model information represents the second model information. The data processing device uses its first local data to train the second encrypted business model obtained from the first encrypted business model; the second encrypted business model obtained by decrypting the second encrypted business model is the same as the second encrypted business model obtained by using the first local data to train the first The business model following a business model is the same.
A data value evaluation system, characterized by comprising: the data value evaluation device according to claim 15, the data processing device according to claim 16, and the data processing device according to claim 17.
A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, the computer program includes program instructions, and when the program instructions are executed by a processor, the processor executes The method according to any one of claims 1 to 14.