WO2021120228A1

WO2021120228A1 - Data processing method and device

Info

Publication number: WO2021120228A1
Application number: PCT/CN2019/127223
Authority: WO
Inventors: 傅培森; 郑文琛; 杨强
Original assignee: 深圳前海微众银行股份有限公司
Priority date: 2019-12-20
Filing date: 2019-12-20
Publication date: 2021-06-24

Abstract

Provided are a data processing method and a device, wherein the method is: acquiring click data of the resource to be processed (101); querying conversion data of the resource according to the click data and outputting a privacy level setting interface (102); receiving the privacy level set by the privacy level setting interface, and querying the privacy protection mechanism corresponding to the privacy level (103); performing privacy processing on the conversion data using the queried privacy protection mechanism, and sending the conversion data after privacy processing to a resource recommendation platform for analysis model establishment (104). When the above method is applied to Fintech, because the privacy level set by the privacy level setting interface can be received, the queried privacy protection mechanism is used to perform privacy processing on the conversion data, which can flexibly perform conversion processing on user data, and seek a balance between privacy protection and the estimated effect of the analysis model.

Description

Data processing method and device

Technical field

The present invention relates to the field of financial technology (Fintech) and computer software, in particular to a data processing method and device.

Background technique

With the development of computer technology, more and more technologies (big data, distributed, Blockchain, artificial intelligence, etc.) are applied in the financial field, and the traditional financial industry is gradually changing to Fintech. At present, information recommendation in the field of financial technology is particularly important, and it is the main channel for expanding financial product customers. If the resource provider wants to improve the conversion effect of information recommended users into consumer users, the information recommendation source needs to transmit user data to the resource recommendation platform, and the resource recommendation platform builds an analysis model through the user data to estimate the user's conversion behavior. In practical applications, in order to avoid the disclosure of user privacy, the conversion data needs to be privately processed and then transmitted to the resource recommendation platform.

Currently, a fixed privacy protection mechanism is usually used to process conversion data. However, if the conversion data is processed in the above manner, the privacy treatment of the conversion data may be too strict or too loose. If the conversion data is subjected to too strict privacy treatment, the information will be extremely distorted, leading to the estimation of the analysis model The accuracy is low. If the privacy treatment of the conversion data is too loose, the privacy protection of the conversion data will be weaker, and the security of the conversion data will be lower.

Summary of the invention

The present application provides a data processing method and device, which solves the problem of seeking a balance between privacy protection and the prediction effect of an analysis model in the prior art.

In the first aspect, this application provides a data processing method: obtaining click data of a resource to be processed; querying conversion data of the resource based on the click data and outputting a privacy level setting interface; receiving information set through the privacy level setting interface Privacy level, and query the privacy protection mechanism corresponding to the privacy level, where different privacy levels correspond to different privacy protection mechanisms; use the query privacy protection mechanism to perform privacy processing on the converted data, and convert the privacy processing The data is sent to the resource recommendation platform for analysis model establishment.

Optionally, said using the privacy protection mechanism of the query to perform privacy processing on the conversion data includes: determining the noise variance corresponding to the privacy protection mechanism of the query; Noisy, the conversion data is subjected to noise processing; according to the preset public key and the preset homomorphic encryption algorithm, the conversion data after the noise processing is encrypted, so as to complete the privacy processing of the conversion data.

Optionally, said using the privacy protection mechanism of the query to perform privacy processing on the converted data includes: determining the noise variance corresponding to the privacy protection mechanism of the query; and the conversion according to a preset public key and a preset homomorphic encryption algorithm The data is encrypted; the encrypted converted data is added with noise corresponding to the noise variance, and the encrypted converted data is noise-added to complete the privacy processing of the converted data.

Optionally, the privacy protection mechanism of the query is used to perform privacy processing on the conversion data, and the converted data after the privacy processing is sent to the resource recommendation platform for analysis model establishment. Before, the method further includes: according to a preset The hybrid algorithm performs mixed processing on the click data and the conversion data to obtain mixed data; the privacy protection mechanism of the query is used to perform privacy processing on the conversion data, and the privacy processing conversion data is sent to the resource recommendation platform The establishment of the analysis model includes: using the privacy protection mechanism of the query to perform privacy processing on the mixed data, and sending the mixed data after the privacy processing to the resource recommendation platform to establish the analysis model.

Optionally, before the acquiring click data of the resource to be processed, the method further includes: acquiring different privacy protection mechanisms; dividing the different privacy protection mechanisms according to different privacy levels; establishing and saving the different privacy protection mechanisms; The mapping relationship between the privacy protection mechanism and the different privacy protection mechanisms.

Optionally, after the privacy-processed conversion data is sent to the resource recommendation platform for analysis model establishment, the method further includes: receiving privacy model parameters sent by the resource recommendation platform; using a preset privacy protection mechanism, Perform deprivation processing on the privacy model parameters to obtain deprivacy model parameters; perform click conversion analysis on the resource based on the deprivacy model parameters to obtain the click conversion rate of the resource.

Optionally, the use of a preset de-privacy protection mechanism to perform de-privacy processing on the privacy model parameters to obtain the de-privacy model parameters includes: if the converted data is encrypted by adding noise and a preset homomorphic encryption algorithm To perform privacy processing in the manner of, obtain the preset private key corresponding to the preset homomorphic encryption algorithm; use the preset private key to decrypt the privacy model parameters, and complete the privacy processing of the privacy model parameters , Get the deprivation model parameters.

Optionally, the obtaining click data of the resource to be processed includes: sending a click data download request to the resource recommendation platform, where the click data download request carries characteristic information of the resource; and receiving the resource recommendation platform according to The click data queried by the characteristic information.

In a second aspect, this application provides a data processing device, including: an acquisition module for acquiring click data of a resource to be processed; a query module for querying conversion data of the resource based on the click data and outputting privacy level settings Interface; receiving the privacy level set through the privacy level setting interface, and querying the privacy protection mechanism corresponding to the privacy level, wherein different privacy levels correspond to different privacy protection mechanisms; processing module used to use query privacy protection The mechanism performs privacy processing on the converted data, and sends the converted data after privacy processing to the resource recommendation platform for analysis model establishment.

Optionally, the processing module is specifically configured to: determine the noise variance corresponding to the privacy protection mechanism of the query; add noise to the conversion data by adding noise corresponding to the noise variance to the conversion data ; According to the preset public key and the preset homomorphic encryption algorithm, the converted data after the noise processing is encrypted, so as to complete the privacy processing of the converted data.

Optionally, the processing module is specifically configured to: determine the noise variance corresponding to the privacy protection mechanism of the query; perform encryption processing on the converted data according to the preset public key and the preset homomorphic encryption algorithm; Noise corresponding to the variance of the noise is added to the data, and noise is added to the encrypted converted data to complete the privacy processing of the converted data.

Optionally, the processing module is further configured to: perform mixing processing on the click data and the conversion data according to a preset mixing algorithm to obtain mixed data; and perform privacy processing on the mixed data by using the privacy protection mechanism of the query, And send the mixed data after privacy processing to the resource recommendation platform for analysis model establishment.

Optionally, the acquisition module is further used to: acquire different privacy protection mechanisms; divide the different privacy protection mechanisms according to different privacy levels; establish and save the different privacy protection mechanisms and the different privacy protection mechanisms. The mapping relationship between protection mechanisms.

Optionally, the acquisition module is further configured to: receive privacy model parameters sent by the resource recommendation platform; and the processing module is further configured to: use a preset privacy protection mechanism to perform privacy removal processing on the privacy model parameters , Obtain de-privacy model parameters; perform click conversion analysis on the resource based on the de-privacy model parameters to obtain the click conversion rate of the resource.

Optionally, the processing module is specifically configured to: if the converted data is subjected to privacy processing by means of noise addition and encryption with a preset homomorphic encryption algorithm, obtaining a preset corresponding to the preset homomorphic encryption algorithm Private key; use the preset private key to decrypt the privacy model parameters, complete the deprivacy processing of the privacy model parameters, and obtain deprivacy model parameters.

Optionally, the acquisition module is specifically configured to: send a click data download request to the resource recommendation platform, where the click data download request carries characteristic information of the resource; and receive the resource recommendation platform according to the characteristic information The click data of the query.

In a third aspect, the present application provides a computer device including a program or instruction. When the program or instruction is executed, the following steps are realized: obtaining click data of the resource to be processed; querying the conversion of the resource according to the click data Data and output the privacy level setting interface; receive the privacy level set through the privacy level setting interface, and query the privacy protection mechanism corresponding to the privacy level, where different privacy levels correspond to different privacy protection mechanisms; use the queried privacy The protection mechanism performs privacy processing on the converted data, and sends the converted data after privacy processing to the resource recommendation platform for analysis model establishment.

In a fourth aspect, the present application provides a storage medium including a program or instruction. When the program or instruction is executed, the following steps are implemented: obtain click data of the resource to be processed; query the conversion of the resource according to the click data Data and output the privacy level setting interface; receive the privacy level set through the privacy level setting interface, and query the privacy protection mechanism corresponding to the privacy level, where different privacy levels correspond to different privacy protection mechanisms; use the queried privacy The protection mechanism performs privacy processing on the converted data, and sends the converted data after privacy processing to the resource recommendation platform for analysis model establishment.

Compared with the prior art using a fixed privacy protection mechanism to process conversion data, the data processing method and device provided in this application can query the conversion data of the resource based on the click data of the resource to be processed, and can output a The privacy level setting interface can be flexibly set according to specific needs, and the privacy protection mechanism corresponding to the privacy level can be queried. Since different privacy levels correspond to different privacy protection mechanisms, different privacy protection mechanisms can be used to perform different operations on the conversion data. Privacy processing of the privacy level, and sending the converted data after privacy processing to the resource recommendation platform for analysis model establishment, that is, it can flexibly select the privacy protection mechanism to process the converted data, which can be in the privacy protection and the estimated effect of the analysis model. Seeking a balance between the two, that is, it can ensure that the accuracy of the analysis model is improved while protecting user privacy.

Description of the drawings

FIG. 1 is a schematic flow diagram of the steps of a data processing method provided by an embodiment of this application;

2 is a schematic diagram of an applicable architecture of a data processing method provided by an embodiment of the application;

FIG. 3 is a schematic diagram of a corresponding interface of a data processing method provided by an embodiment of the application;

FIG. 4 is a schematic structural diagram of a data processing device provided by an embodiment of the application.

Detailed ways

In order to better understand the above technical solutions, the above technical solutions will be described in detail below with reference to the drawings and specific implementations of the specification. It should be understood that the embodiments of the application and the specific features in the embodiments are detailed to the technical solutions of the application. Note, rather than limiting the technical solution of the present application, the embodiments of the present application and the technical features in the embodiments can be combined with each other if there is no conflict.

During the operation of financial institutions (banking institutions, insurance institutions or securities institutions) (such as bank loan business, deposit business, etc.), information recommendation in the field of financial technology is particularly important and is the main channel for expanding financial product customers. If the resource provider wants to improve the conversion effect of the users recommended by the information, the information recommendation source needs to transfer the user conversion data to the resource recommendation platform, and the resource recommendation platform analyzes the user conversion data. However, in current methods, a fixed privacy protection mechanism is usually used to process user conversion data. This method may cause the privacy treatment of user conversion data to be too strict or too loose. On the one hand, if the privacy treatment of user conversion data is too strict, the original user conversion data will become severely distorted, resulting in lower estimation accuracy of the analysis model; on the other hand, if the privacy treatment of user conversion data is too loose Although there is a high accuracy of the analysis model, it will expose more user conversion data to become severely distorted, resulting in lower security of the user conversion data. Obviously, the existing technology cannot find a balance between privacy protection and the predicted effect of the analysis model. This situation does not meet the needs of banks and other financial institutions, and cannot guarantee the efficient operation of various businesses of financial institutions.

To this end, as shown in FIG. 1, an embodiment of the present application provides a data processing method, which includes the following steps:

Step 101: Obtain the click data of the resource to be processed.

Step 102: Query the conversion data of the resource according to the click data and output a privacy level setting interface.

Step 103: Receive the privacy level set through the privacy level setting interface, and query the privacy protection mechanism corresponding to the privacy level.

Among them, different privacy levels correspond to different privacy protection mechanisms.

Step 104: Use the privacy protection mechanism of the query to perform privacy processing on the converted data, and send the converted data after privacy processing to the resource recommendation platform for analysis model establishment.

It should be noted that in step 101 to step 104, users can be divided into three categories in chronological order: exposure users, click users, and converted users. Exposure users are users who have recommended resources for the resource recommendation platform; click users are users who have clicked on the resources recommended by the resource recommendation platform. Obviously, the click users must be exposed users; the converted users are users who have clicked on the resources recommended by the resource recommendation platform. For users who send conversion behaviors (such as registration behavior purchase behaviors) on the other end, obviously, the converted users must be clickers. The exposure data includes at least feature information and an exposure label value (for example, the exposure label value is represented by 0). Click data includes at least feature information and click tag value (for example, the click tag value is represented by 1). The conversion data includes at least the characteristic information and the conversion tag value (for example, the conversion tag value is represented by 2). It should be noted that, in the following description, unless otherwise specified, the conversion data is the original conversion data without privacy treatment, and the conversion data is also It can be called raw conversion data. Among them, the feature information can include multiple types of features, such as user features, resource features, and scene features. User characteristics refer to the basic attributes of users, such as age and gender. Resource characteristics refer to the basic attributes of recommended resources, such as the format and layout of recommended resources. Scenario features refer to scenarios in which recommended resources are recommended to users, such as recommended locations and user actions that trigger recommendations.

The execution subject in step 101 to step 104 may be the resource provider, and the resource provider will entrust the resource recommendation platform to perform resource recommendation. Since the user’s conversion behavior occurs on the resource provider side, the resource provider side will have a large amount of user conversion data. Because the processing resources recommended to the user are executed by the resource recommendation platform, and the resource recommendation platform will respond when the user clicks on the processing resource Therefore, the resource recommendation platform will have a large amount of user exposure data and user click data.

In an optional implementation manner before step 101, obtain different privacy protection mechanisms; divide the different privacy protection mechanisms according to different privacy levels; establish and save the different privacy protection mechanisms and the different privacy The mapping relationship between protection mechanisms.

It should be noted that the specific methods of the privacy protection mechanism are not limited, as long as the privacy protection effect of the converted data is achieved. For example, change the conversion data according to certain rules, such as randomly changing the label value of the conversion data, and then cover the conversion data according to certain rules, such as encrypting the conversion data through a symmetric encryption algorithm. Each privacy protection mechanism can correspond to a degree of data change. The degree of data change is the degree of difference between the changed converted data and the original converted data. As for the mapping relationship between the privacy level and the privacy protection mechanism, it can be set to be positively correlated with the degree of data change of the privacy protection mechanism, or it can be set to be negatively correlated, which is not limited here. It can also be set without the degree of data change of the privacy protection mechanism. For example, directly establish a mapping relationship between privacy level one and privacy protection mechanism one, and establish a mapping relationship between privacy level two and privacy protection mechanism two, and so on.

In an optional implementation manner of step 101, the specific manner may be:

Sending a click data download request to the resource recommendation platform, where the click data download request carries characteristic information of the resource; and receiving the click data that the resource recommendation platform queries according to the characteristic information.

When the resource recommendation platform recommends processing resources to users, it can record which users are recommended which resources and record the characteristic information of the resources. Then, the resource provider can query which click data puts the resource with corresponding characteristic information according to the characteristic information of the resource, so as to query the corresponding click data.

For example, if the download request is q, q includes the characteristic information t of the resource, and the click data of the characteristic information resource of t placed in the resource platform includes 10 items such as a1, a2,..., a10, then a1, a2 are received ,..., a10 and other 10 click data.

In an implementation of step 102, determine the noise variance corresponding to the privacy protection mechanism of the query; add noise to the conversion data by adding noise corresponding to the noise variance to the conversion data; It is assumed that the public key and the preset homomorphic encryption algorithm perform encryption processing on the converted data after the noise processing, so as to complete the privacy processing of the converted data.

Among them, the specific noise type is not limited. For example, the noise is Laplace noise. Since the noise variance characterizes the degree of fluctuation of the noise, noise with different fluctuation degrees corresponding to the noise variance can be added to the conversion data. Obviously, the greater the noise variance, the stronger the uncertainty of adding noise, and the greater the transformation of the noise-added conversion data from the original conversion data. Therefore, noise acts as a confusing information. The noise variance can be used to flexibly control the degree of transformation of the converted data after adding noise, that is, control the degree of privacy. Furthermore, it can continue to encrypt the converted data after the noise processing according to the preset public key and the preset homomorphic encryption algorithm, and further cover the converted data after the noise processing, thereby providing a noise-based fluctuation The extent to which the conversion data is changed and covered up.

In another embodiment, it is also possible to encrypt the conversion data first, and then add noise to the conversion data, specifically as follows:

Determine the noise variance corresponding to the queried privacy protection mechanism; perform encryption processing on the converted data according to the preset public key and preset homomorphic encryption algorithm; add noise corresponding to the noise variance to the encrypted converted data, Noise addition processing is performed on the encrypted converted data to complete the privacy processing of the converted data.

In this implementation manner, the transformed data is covered first, and then the encrypted transformed data is obfuscated, and the degree of transformation of the transformed data after adding noise can also be flexibly controlled through the noise variance.

It should be noted that the homomorphic encryption algorithm of the original conversion data can be decrypted and restored to the original conversion data by the corresponding homomorphic decryption algorithm after encryption. Because the encryption is homomorphic, the homomorphic encryption will not predict the effect of the analysis model. The conversion data added with noise is a transformation of the conversion data, and this transformation will have an impact on the estimated effect of the analysis model.

In an optional implementation, before step 101, the click data and the conversion data are mixed according to a preset mixing algorithm to obtain mixed data; in step 104, the query privacy protection mechanism is used to perform the mixed processing on the mixed data. The data is subjected to privacy processing, and the mixed data after privacy processing is sent to the resource recommendation platform for analysis model establishment. In this embodiment, the mixed data that mixes the click data and the conversion data is subjected to privacy processing, so that the click data is also subjected to privacy processing, and the mixed data can also seek a balance between privacy protection and model prediction effect .

In an optional implementation after step 104, receiving privacy model parameters sent by the resource recommendation platform; using a preset privacy protection mechanism to perform privacy removal processing on the privacy model parameters to obtain privacy model parameters; based on The deprivation model parameter performs a click conversion analysis on the resource to obtain the click conversion rate of the resource.

Since in step 104 the converted data after privacy processing is sent to the resource recommendation platform for analysis model establishment, the analysis model parameters obtained by the resource recommendation platform are also privacy model parameters with privacy protection; the direct use of the privacy model parameters will reduce the analysis model Therefore, the preset privacy protection mechanism can be used to deprivacy processing the privacy model parameters to obtain the privacy model parameters; based on the privacy model parameters, click conversion analysis of the resources to obtain more Accurate click conversion rate of the resource.

In the above-mentioned optional implementation manner, a preset privacy protection mechanism is used to perform privacy removal processing on the privacy model parameters, and the specific method for obtaining the privacy model parameters may be as follows:

If the converted data is subjected to privacy processing by adding noise and encrypting with a preset homomorphic encryption algorithm, obtain the preset private key corresponding to the preset homomorphic encryption algorithm; The privacy model parameters are decrypted, the privacy removal processing of the privacy model parameters is completed, and the privacy model parameters are obtained.

Since the addition of noise changes the conversion data, this part of the change will be reflected in the training of the analysis model. The encryption algorithm used is the preset homomorphic encryption algorithm, so the encrypted data can be trained to get the same effect as the data before encryption. Only the preset private key corresponding to the preset homomorphic encryption algorithm is required to perform the privacy model parameters. Decryption, completes the de-privacy processing of the privacy model parameters, and obtains the de-privacy model parameters. Therefore, this method can not only protect the privacy of the converted data, but also obtain more accurate privacy model parameters.

The data processing method shown in Fig. 1 will be described in detail below in conjunction with Fig. 2 and Fig. 3.

The data processing method shown in Figure 1 can provide tools for the resource provider to set the privacy level. The core is that the resource provider can balance the privacy level and model prediction effect by itself. For example, the privacy level and the privacy protection mechanism are positively correlated, that is, the higher the privacy level, the greater the degree of modification to the conversion data, and the higher the degree of encryption of the conversion data. When the resource provider sets the privacy level to the lowest, the resource provider focuses more on pursuing the effect of information recommendation; when the resource provider sets the privacy level to the highest, the resource provider focuses more on pursuing user privacy protection. When the privacy level is set between the lowest and the highest, you can freely control the degree of privacy protection and obtain different effects. The specific process is as follows:

Step 1: The resource provider obtains click data from the resource recommendation platform.

Step 2: The resource provider mixes user click data and conversion data.

Step 3: The resource provider uses tools to set different privacy levels (for example, the higher the privacy level, the higher the degree of data conversion) is equivalent to adding noise to the data. As shown in Figure 3, this step of operation is visual and measurable for the resource provider.

Specific ways to add noise to the conversion data can include:

The first type: Randomly convert the label value of the conversion data to 0 or randomly to 1. The resource provider reverses the tag value 0, 1 of the user conversion data with a certain probability; the higher the probability of reversal, the stronger the degree of privacy protection, but the worse the estimated effect.

The second type: the resource provider adds a noise conforming to the Laplacian distribution, for example, adding Laplace noise conforming to L(0,0.1) on the basis of 0 and 1. The greater the variance of the noise, the stronger the privacy protection.

It should be noted that the degree of privacy protection and the estimated effect need to be balanced. Specifically, machine learning methods can be used to estimate the subsequent effects of different privacy protections based on the effects of different levels of privacy protection in the past. Generally speaking, this kind of privacy leakage of the machine learning model is contradictory to the effect of the model. Therefore, a corresponding level of privacy protection can be applied according to the effect of the model and the similarity of the model. Specifically, for a model with extremely high accuracy or extremely sensitive to the estimated value of each individual, a large amount of noise is applied to provide a certain level of privacy protection. On the other hand, for models with low accuracy or insensitive to each individual estimation, a smaller noise can be applied to provide a certain level of privacy protection. In addition, in addition to the effect of the machine learning model and personal sensitivity, the level of protection that the resource provider hopes to provide to its users also affects the degree of noise imposed. Specifically, when the machine learning model is fixed, larger noise addition generally represents stricter privacy protection.

Step 4: The resource provider encrypts the converted data.

The specific encryption method can be a homomorphic encryption method, which is mainly used to protect the security during data transmission.

Step 5: The resource provider transmits the user's encrypted data to the resource recommendation platform. The resource recommendation platform establishes a conversion estimation model based on the conversion data after privacy processing. Used to estimate click conversion rate.

Step 6: The resource provider downloads the privacy model parameters and decrypts to obtain the privacy model parameters.

As shown in FIG. 4, the present application provides a data processing device, which includes: an acquisition module 401, configured to acquire click data of a resource to be processed; and a query module 402, configured to query conversion data of the resource according to the click data and Output the privacy level setting interface; receive the privacy level set through the privacy level setting interface, and query the privacy protection mechanism corresponding to the privacy level, where different privacy levels correspond to different privacy protection mechanisms; processing module 403 for The privacy protection mechanism of the query is used to perform privacy processing on the converted data, and the converted data after the privacy processing is sent to the resource recommendation platform for analysis model establishment.

In an optional implementation manner, the processing module 403 is specifically configured to: determine the noise variance corresponding to the privacy protection mechanism of the query; The data is subjected to noise processing; the converted data after the noise processing is encrypted according to the preset public key and the preset homomorphic encryption algorithm, so as to complete the privacy processing of the converted data.

In an optional implementation manner, the processing module 403 is specifically configured to: determine the noise variance corresponding to the privacy protection mechanism of the query; perform encryption processing on the converted data according to the preset public key and the preset homomorphic encryption algorithm; Noise corresponding to the variance of the noise is added to the encrypted conversion data, and noise is added to the encrypted conversion data to complete the privacy processing of the conversion data.

In an optional implementation manner, the processing module 403 is further configured to: perform mixing processing on the click data and the conversion data according to a preset mixing algorithm to obtain mixed data; The data is subjected to privacy processing, and the mixed data after privacy processing is sent to the resource recommendation platform for analysis model establishment.

In an optional implementation manner, the acquisition module 401 is further configured to: acquire different privacy protection mechanisms; divide the different privacy protection mechanisms according to different privacy levels; establish and save the different privacy protection mechanisms and The mapping relationship between the different privacy protection mechanisms.

In an optional implementation manner, the obtaining module 401 is further configured to: receive privacy model parameters sent by the resource recommendation platform; the processing module 403 is further configured to: use a preset privacy protection mechanism to protect the privacy The model parameters are subjected to de-privacy processing to obtain de-privacy model parameters; based on the de-privacy model parameters, click conversion analysis is performed on the resource to obtain the click conversion rate of the resource.

In an optional implementation manner, the processing module 403 is specifically configured to: if the converted data is subjected to privacy processing by means of noise addition and encryption with a preset homomorphic encryption algorithm, then obtain the data with the preset homomorphic encryption The preset private key corresponding to the algorithm; decrypt the privacy model parameters by using the preset private key, complete the deprivacy processing of the privacy model parameters, and obtain the deprivacy model parameters.

In an optional implementation manner, the acquisition module 401 is specifically configured to: send a click data download request to the resource recommendation platform, where the click data download request carries characteristic information of the resource; and receive the resource recommendation platform The click data queried according to the characteristic information.

The embodiment of the present application provides a computer device including a program or instruction. When the program or instruction is executed, the following steps are implemented: obtaining click data of a resource to be processed; querying conversion data of the resource according to the click data and Output the privacy level setting interface; receive the privacy level set through the privacy level setting interface, and query the privacy protection mechanism corresponding to the privacy level, where different privacy levels correspond to different privacy protection mechanisms; use the query privacy protection mechanism Perform privacy processing on the converted data, and send the converted data after privacy processing to a resource recommendation platform for analysis model establishment.

Optionally, the use of the privacy protection mechanism of the query to perform privacy processing on the converted data includes: determining the noise variance corresponding to the privacy protection mechanism of the query; and the conversion according to a preset public key and a preset homomorphic encryption algorithm The data is encrypted; the encrypted converted data is added with noise corresponding to the noise variance, and the encrypted converted data is noise-added to complete the privacy processing of the converted data.

The embodiment of the present application provides a storage medium that includes a program or instruction. When the program or instruction is executed, the following steps are implemented: obtain the click data of the resource to be processed; query the conversion data of the resource according to the click data and Output the privacy level setting interface; receive the privacy level set through the privacy level setting interface, and query the privacy protection mechanism corresponding to the privacy level, where different privacy levels correspond to different privacy protection mechanisms; use the query privacy protection mechanism Perform privacy processing on the converted data, and send the converted data after privacy processing to a resource recommendation platform for analysis model establishment.

Finally, it should be noted that those skilled in the art should understand that the embodiments of the present application can be provided as methods, systems, or computer program products. Therefore, this application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, optical storage, etc.) containing computer-usable program codes.

This application is described with reference to flowcharts and/or block diagrams of methods, equipment (systems), and computer program products according to this application. It should be understood that each process and/or block in the flowchart and/or block diagram, and the combination of processes and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing equipment to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing equipment are used to generate It is a device that realizes the functions specified in one process or multiple processes in the flowchart and/or one block or multiple blocks in the block diagram.

Obviously, those skilled in the art can make various changes and modifications to the application without departing from the scope of the application. In this way, if these modifications and variations of this application fall within the scope of the claims of this application and their equivalent technologies, then this application is also intended to include these modifications and variations.

Claims

A data processing method, characterized in that it comprises:

Get the click data of the resource to be processed;

Query the conversion data of the resource according to the click data and output a privacy level setting interface;

Receiving the privacy level set through the privacy level setting interface, and querying the privacy protection mechanism corresponding to the privacy level, where different privacy levels correspond to different privacy protection mechanisms;

The privacy protection mechanism of the query is used to perform privacy processing on the converted data, and the converted data after the privacy processing is sent to the resource recommendation platform for analysis model establishment.
The method according to claim 1, wherein said using the privacy protection mechanism of the query to perform privacy processing on the conversion data comprises:

Determine the noise variance corresponding to the queried privacy protection mechanism;

By adding noise corresponding to the variance of the noise to the conversion data, performing noise addition processing on the conversion data;

Encryption processing is performed on the converted data after noise processing according to the preset public key and the preset homomorphic encryption algorithm, so as to complete the privacy processing of the converted data.
The method according to claim 1, wherein said using the privacy protection mechanism of the query to perform privacy processing on the conversion data comprises:

Determine the noise variance corresponding to the queried privacy protection mechanism;

Perform encryption processing on the converted data according to the preset public key and the preset homomorphic encryption algorithm;

By adding noise corresponding to the variance of the noise to the encrypted conversion data, noise is added to the encrypted conversion data to complete the privacy processing of the conversion data.
The method of claim 1, wherein the privacy protection mechanism of the query is used to perform privacy processing on the conversion data, and the converted data after the privacy processing is sent to the resource recommendation platform for analysis model establishment, before, The method also includes:

Performing mixing processing on the click data and the conversion data according to a preset mixing algorithm to obtain mixed data;

Said using the privacy protection mechanism of the query to perform privacy processing on the conversion data, and send the converted data after privacy processing to the resource recommendation platform for analysis model establishment, including:

The privacy protection mechanism of the query is used to perform privacy processing on the mixed data, and the mixed data after the privacy processing is sent to the resource recommendation platform for analysis model establishment.
The method according to any one of claims 1 to 4, wherein before said obtaining the click data of the resource to be processed, the method further comprises:

Obtain different privacy protection mechanisms;

Divide the different privacy protection mechanisms according to different privacy levels;

Establish and save the mapping relationship between the different privacy protection mechanisms and the different privacy protection mechanisms.
The method according to any one of claims 1 to 4, characterized in that, after said sending the privacy-treated conversion data to a resource recommendation platform for analysis model establishment, the method further comprises:

Receiving privacy model parameters sent by the resource recommendation platform;

Using a preset de-privacy protection mechanism to perform de-privacy processing on the privacy model parameters to obtain de-privacy model parameters;

Perform click conversion analysis on the resource based on the deprivation model parameters to obtain the click conversion rate of the resource.
The method according to claim 6, wherein said using a preset de-privacy protection mechanism to perform de-privacy processing on said privacy model parameters to obtain de-privacy model parameters comprises:

If the converted data is subjected to privacy processing by means of adding noise and encryption with a preset homomorphic encryption algorithm, obtaining a preset private key corresponding to the preset homomorphic encryption algorithm;

Decrypt the privacy model parameters by using the preset private key, complete deprivacy processing of the privacy model parameters, and obtain deprivacy model parameters.
A data processing device, characterized in that it comprises:

The acquisition module is used to acquire the click data of the resource to be processed;

The query module is used to query the conversion data of the resource according to the click data and output the privacy level setting interface; receive the privacy level set through the privacy level setting interface, and query the privacy protection mechanism corresponding to the privacy level, Among them, different privacy levels correspond to different privacy protection mechanisms;

The processing module is used to perform privacy processing on the converted data using the privacy protection mechanism of the query, and send the converted data after privacy processing to the resource recommendation platform for analysis model establishment.
The device according to claim 8, wherein the processing module is specifically configured to:

Determine the noise variance corresponding to the queried privacy protection mechanism; add noise to the converted data by adding noise corresponding to the noise variance to the converted data;

Encryption processing is performed on the converted data after noise processing according to the preset public key and the preset homomorphic encryption algorithm, so as to complete the privacy processing of the converted data.
The device according to claim 8, wherein the processing module is specifically configured to:

Determine the noise variance corresponding to the queried privacy protection mechanism; perform encryption processing on the converted data according to the preset public key and the preset homomorphic encryption algorithm;

By adding noise corresponding to the variance of the noise to the encrypted conversion data, noise is added to the encrypted conversion data to complete the privacy processing of the conversion data.
The device according to claim 8, wherein the processing module is further configured to:

Performing mixing processing on the click data and the conversion data according to a preset mixing algorithm to obtain mixed data;

The privacy protection mechanism of the query is used to perform privacy processing on the mixed data, and the mixed data after the privacy processing is sent to the resource recommendation platform for analysis model establishment.
The device according to any one of claims 8-11, wherein the acquisition module is further configured to:

Obtain different privacy protection mechanisms;

The different privacy protection mechanisms are divided according to different privacy levels; the mapping relationship between the different privacy protection mechanisms and the different privacy protection mechanisms is established and stored.
The device according to any one of claims 8-11, wherein the acquisition module is further configured to:

Receiving privacy model parameters sent by the resource recommendation platform;

The processing module is also used for:

Using a preset de-privacy protection mechanism, perform de-privacy processing on the privacy model parameters to obtain de-privacy model parameters; perform click conversion analysis on the resource based on the de-privacy model parameters to obtain the click conversion rate of the resource.
The device according to claim 13, wherein the processing module is specifically configured to:

If the converted data is subjected to privacy processing by means of adding noise and encryption with a preset homomorphic encryption algorithm, obtaining a preset private key corresponding to the preset homomorphic encryption algorithm;

Decrypt the privacy model parameters by using the preset private key, complete deprivacy processing of the privacy model parameters, and obtain deprivacy model parameters.
A computer device characterized by comprising a program or instruction, and when the program or instruction is executed, the following steps are implemented:

Get the click data of the resource to be processed;

Query the conversion data of the resource according to the click data and output a privacy level setting interface;

Receiving the privacy level set through the privacy level setting interface, and querying the privacy protection mechanism corresponding to the privacy level, where different privacy levels correspond to different privacy protection mechanisms;

The privacy protection mechanism of the query is used to perform privacy processing on the converted data, and the converted data after the privacy processing is sent to the resource recommendation platform for analysis model establishment.
The computer device according to claim 15, wherein said using the privacy protection mechanism of the query to perform privacy processing on the conversion data comprises:

Determine the noise variance corresponding to the queried privacy protection mechanism;

By adding noise corresponding to the variance of the noise to the conversion data, performing noise addition processing on the conversion data;

Encryption processing is performed on the converted data after noise processing according to the preset public key and the preset homomorphic encryption algorithm, so as to complete the privacy processing of the converted data.
15. The computer device according to claim 15, wherein said using the privacy protection mechanism of the query to perform privacy processing on the conversion data comprises:

Determine the noise variance corresponding to the queried privacy protection mechanism;

Perform encryption processing on the converted data according to the preset public key and the preset homomorphic encryption algorithm;

By adding noise corresponding to the variance of the noise to the encrypted conversion data, noise is added to the encrypted conversion data to complete the privacy processing of the conversion data.
A storage medium, characterized in that it includes a program or instruction, and when the program or instruction is executed, the following steps are implemented:

Get the click data of the resource to be processed;

Query the conversion data of the resource according to the click data and output a privacy level setting interface;

Receiving the privacy level set through the privacy level setting interface, and querying the privacy protection mechanism corresponding to the privacy level, where different privacy levels correspond to different privacy protection mechanisms;

The privacy protection mechanism of the query is used to perform privacy processing on the converted data, and the converted data after the privacy processing is sent to the resource recommendation platform for analysis model establishment.
18. The storage medium of claim 18, wherein said using the privacy protection mechanism of the query to perform privacy processing on the conversion data comprises:

Determine the noise variance corresponding to the queried privacy protection mechanism;

By adding noise corresponding to the variance of the noise to the conversion data, performing noise addition processing on the conversion data;

Encryption processing is performed on the converted data after noise processing according to the preset public key and the preset homomorphic encryption algorithm, so as to complete the privacy processing of the converted data.
18. The storage medium of claim 18, wherein said using the privacy protection mechanism of the query to perform privacy processing on the conversion data comprises:

Determine the noise variance corresponding to the queried privacy protection mechanism;

Perform encryption processing on the converted data according to the preset public key and the preset homomorphic encryption algorithm;

By adding noise corresponding to the variance of the noise to the encrypted conversion data, noise is added to the encrypted conversion data to complete the privacy processing of the conversion data.