Data processing method, device, system and equipment

Abstract

The embodiments of the present specification provide a data processing method, apparatus, system and equipment, wherein the method includes: determining pseudo sample data based on feature information of first sample data; training a preset first risk identification model based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data to obtain a trained first risk identification model, and determining the risk identification accuracy of the trained first risk identification model based on third sample data; if it is determined that the pseudo sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo sample data is sent to a target device, and the pseudo sample data is used to trigger the target device to train a second risk identification model based on the pseudo sample data, and perform risk identification processing on the user based on the trained second risk identification model.

Description

Technical Field

[0001] This document relates to the field of data processing technology, and in particular to a data processing method, device, system and equipment. Background Art

[0002] With the rapid development of computer technology, the number of users of enterprises that provide resource transfer services to users, such as online banking, has increased dramatically, and the data risks faced by enterprises are also increasing. In order to reduce data risks, enterprises can conduct risk identification by building risk identification models. For example, enterprises can train risk identification models based on locally stored user sample data and risk labels, and conduct risk identification based on the trained risk identification model.

[0003] However, enterprises such as online banks may have a small amount of user sample data and risk labels and poor data quality, which will lead to poor risk identification effect and low risk identification accuracy of the trained risk identification model. Therefore, a solution is needed to improve the risk identification effect and risk identification accuracy of the risk identification model. Summary of the invention

[0004] The purpose of the embodiments of this specification is to provide a data processing method, device, system and equipment to provide a solution that can improve the risk identification effect and risk identification accuracy of the risk identification model.

[0005] In order to implement the above technical solution, the embodiments of this specification are implemented as follows:

[0006] In a first aspect, an embodiment of the present specification provides a data processing method, comprising: determining pseudo-sample data based on feature information of first sample data, wherein data obtained by restoring the pseudo-sample data is different from the first sample data; training a preset first risk identification model based on the pseudo-sample data and second sample data determined by the data amount of the pseudo-sample data to obtain a trained first risk identification model, and determining the risk identification accuracy of the trained first risk identification model based on third sample data; if it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, sending the pseudo-sample data to a target device, the pseudo-sample data being used to trigger the target device to train a second risk identification model based on the pseudo-sample data, and performing risk identification processing on the user based on the trained second risk identification model.

[0007] In a second aspect, an embodiment of the present specification provides a data processing method, including: receiving pseudo-sample data sent by a server, the pseudo-sample data being sample data determined based on feature information of first sample data and sent by the server when the server determines that the pseudo-sample data meets preset risk identification requirements based on the risk identification accuracy of a trained first risk identification model, the trained first risk identification model being trained based on the pseudo-sample data and second sample data determined by the data amount of the pseudo-sample data, wherein data obtained by restoring the pseudo-sample data is different from the first sample data; based on the pseudo-sample data, training a preset second risk identification model to obtain a pre-trained second risk identification model, so as to perform risk identification processing on users based on the pre-trained second risk identification model.

[0008] In a third aspect, an embodiment of the present specification provides a data processing system, comprising a server and a target device, wherein: the server is used to determine pseudo-sample data based on feature information of first sample data, wherein data obtained by restoring the pseudo-sample data is different from the first sample data; a preset first risk identification model is trained based on the pseudo-sample data and second sample data determined by the data volume of the pseudo-sample data to obtain a trained first risk identification model, and a risk identification accuracy rate of the trained first risk identification model is determined based on third sample data; if the pseudo-sample data is determined to meet the preset risk identification requirements based on the risk identification accuracy rate of the trained first risk identification model, the pseudo-sample data is sent to the target device; the target device is used to train a preset second risk identification model based on the pseudo-sample data to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model.

[0009] In a fourth aspect, an embodiment of the present specification provides a data processing device, comprising: a data determination module, for determining pseudo-sample data based on feature information of first sample data, wherein data obtained by restoring the pseudo-sample data is different from the first sample data; a model training module, for training a preset first risk identification model based on the pseudo-sample data and second sample data determined by the data amount of the pseudo-sample data to obtain a trained first risk identification model, and determining the risk identification accuracy of the trained first risk identification model based on third sample data; a data sending module, for sending the pseudo-sample data to a target device if it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, wherein the pseudo-sample data is used to trigger the target device to train a second risk identification model based on the pseudo-sample data, and perform risk identification processing on the user based on the trained second risk identification model.

[0010] In a fifth aspect, an embodiment of the present specification provides a data processing device, comprising: a data receiving module, for receiving pseudo-sample data sent by a server, the pseudo-sample data being sample data determined based on feature information of the first sample data and sent by the server when the server determines that the pseudo-sample data meets preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the trained first risk identification model is trained based on the pseudo-sample data and second sample data determined by the data amount of the pseudo-sample data, wherein data obtained by restoring the pseudo-sample data is different from the first sample data; a model training module, for training a preset second risk identification model based on the pseudo-sample data to obtain a pre-trained second risk identification model, so as to perform risk identification processing on users based on the pre-trained second risk identification model.

[0011] In a sixth aspect, an embodiment of the present specification provides a data processing device, comprising: a processor; and a memory arranged to store computer executable instructions, wherein the executable instructions, when executed, cause the processor to: determine pseudo sample data based on feature information of first sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data; train a preset first risk identification model based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data to obtain a trained first risk identification model, and determine the risk identification accuracy of the trained first risk identification model based on third sample data; if it is determined that the pseudo sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo sample data is sent to a target device, and the pseudo sample data is used to trigger the target device to train a second risk identification model based on the pseudo sample data, and perform risk identification processing on the user based on the trained second risk identification model.

[0012] In a seventh aspect, an embodiment of the present specification provides a data processing device, comprising: a processor; and a memory arranged to store computer executable instructions, wherein the executable instructions, when executed, cause the processor to: receive pseudo-sample data sent by a server, wherein the pseudo-sample data is sample data determined based on feature information of first sample data and sent by the server when the server determines that the pseudo-sample data meets preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, wherein the trained first risk identification model is trained based on the pseudo-sample data and second sample data determined by the data amount of the pseudo-sample data, wherein data obtained by restoring the pseudo-sample data is different from the first sample data; and based on the pseudo-sample data, a preset second risk identification model is trained to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model.

[0013] In an eighth aspect, an embodiment of the present specification provides a storage medium, wherein the storage medium is used to store computer-executable instructions, and the executable instructions implement the following process when executed: based on the characteristic information of the first sample data, determine pseudo-sample data, wherein the data obtained by restoring the pseudo-sample data is different from the first sample data; based on the pseudo-sample data and the second sample data determined by the data amount of the pseudo-sample data, train a preset first risk identification model to obtain a trained first risk identification model, and determine the risk identification accuracy of the trained first risk identification model based on the third sample data; if it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo-sample data is sent to the target device, and the pseudo-sample data is used to trigger the target device to train the second risk identification model based on the pseudo-sample data, and perform risk identification processing on the user based on the trained second risk identification model.

[0014] In the ninth aspect, an embodiment of the present specification provides a storage medium, which is used to store computer-executable instructions, and the executable instructions implement the following process when being executed: receiving pseudo-sample data sent by a server, the pseudo-sample data is sample data determined based on feature information of the first sample data and sent by the server when it is determined that the pseudo-sample data meets preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the trained first risk identification model is trained based on the pseudo-sample data and second sample data determined by the data amount of the pseudo-sample data, wherein the data obtained by restoring the pseudo-sample data is different from the first sample data; based on the pseudo-sample data, the preset second risk identification model is trained to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0016] Figure 1A A flowchart of a data processing method embodiment of this specification;

[0017] Figure 1B A schematic diagram of a processing process of a data processing method in this specification;

[0018] Figure 2 A processing diagram of another data processing method of the present specification;

[0019] Figure 3 A schematic diagram of a pseudo sample data determination process of this specification;

[0020] Figure 4 This is a schematic diagram of the corresponding relationship between the risk identification accuracy and the sample authenticity probability in this manual;

[0021] Figure 5 A processing diagram of another data processing method of the present specification;

[0022] Figure 6 A schematic diagram of a pseudo sample data determination process of this specification;

[0023] Fig. 7A A flowchart of another data processing method embodiment of the present specification;

[0024] Figure 7B A processing diagram of another data processing method of the present specification;

[0025] Figure 8 A processing diagram of another data processing method of the present specification;

[0026] Fig. 9 A schematic diagram of the structure of another data processing device embodiment of the present specification;

[0027] Fig.10 A schematic diagram of the structure of another data processing device embodiment of the present specification;

[0028] Fig.11 This is a structural schematic diagram of a data processing device in this specification. DETAILED DESCRIPTION

[0029] The embodiments of this specification provide a data processing method, apparatus, system and device.

[0030] In order to enable those skilled in the art to better understand the technical solutions in this specification, the technical solutions in the embodiments of this specification will be clearly and completely described below in conjunction with the drawings in the embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of this specification.

[0031] Embodiment 1

[0032] like Figure 1A and1B As shown, the embodiment of this specification provides a data processing method, the execution subject of the method can be a server, and the server can be an independent server or a server cluster composed of multiple servers. The method can specifically include the following steps:

[0033] In S102 , pseudo sample data is determined based on the feature information of the first sample data.

[0034] Among them, the first sample data may include real user data composed of business data generated by a user when a certain business is triggered and the attribute data of the user. For example, the characteristic information of the first sample data may be the user's attribute information (such as information that can characterize one or more attributes of the user), behavior information (such as business data generated by triggering the execution of a certain business such as transfer behavior, call behavior, etc.). Specifically, the first sample data may be sample data constructed by the attribute information and resource transfer behavior information (such as resource transfer quantity, resource transfer time, resource transfer object, etc.) of users who have used the resource transfer service in the past month stored by the server. The data obtained by restoring the pseudo sample data is different from the first sample data.

[0035] In implementation, with the rapid development of computer technology, the number of users of enterprises that provide resource transfer services to users, such as online banks, has increased dramatically, and the data risks faced by enterprises are also increasing. In order to reduce data risks, enterprises can perform risk identification processing by constructing risk identification models. For example, an enterprise can train a risk identification model based on locally stored user sample data and risk labels, and perform risk identification processing based on the trained risk identification model. However, enterprises such as online banks may have a small amount of user sample data and risk labels, and poor data quality, which will result in poor risk identification effect of the risk identification model obtained by training, and low risk identification accuracy. Therefore, a solution that can improve the risk identification effect and risk identification accuracy of the risk identification model is needed. To this end, the embodiments of this specification provide a technical solution that can solve the above-mentioned problems. For details, please refer to the following content.

[0036] Taking the target device as a client used by an organization that can provide resource transfer services to users as an example, since different organizations have different user private data, and the user private data owned by each organization may have problems such as small data volume and poor data quality, the target device can interact with the server with larger data volume and stronger data processing capabilities to enable the risk identification model (i.e., the second risk identification model) used by the target device to achieve better risk identification effect.

[0037] The server can construct the first sample data based on the business data of multiple businesses stored locally. For example, the server can obtain real user data within a preset sample construction period, such as obtaining business data of multiple different businesses such as resource transfer business and instant messaging business in the past month, and then construct the first sample data based on the business data of these multiple businesses. Specifically, the service data corresponding to the user ID can be filtered out from multiple different business data according to the user ID, and the first sample data can be constructed based on the user ID and the filtered business data. Alternatively, the service data corresponding to the service ID and the user ID can be filtered out from multiple different business data according to the service ID and the user ID to construct the first sample data.

[0038] After determining the first sample data, the server can determine the pseudo sample data based on the feature information of the first sample data. For example, taking the first sample data as the user's image data (such as face image data, fingerprint image data, etc.) as an example, the user image data can be input into a pre-trained generation model to obtain pseudo sample data, wherein the generation model can be a model for generating a synthetic image based on image data and random perturbation data obtained by training a model constructed by a deep learning algorithm based on historical image data. In this way, the pseudo sample data is a synthetic image generated based on the first sample data, and therefore, the image obtained by restoring the synthetic image is different from the original image, that is, the data obtained by restoring the pseudo sample data is different from the first sample data. In addition, since the first sample data can be real user data, and the data obtained by restoring the pseudo sample data is different from the first sample data, that is, the real user cannot be determined based on the pseudo sample data. For example, when performing user recognition on the synthetic image generated above, the real user cannot be identified.

[0039] In addition, the above is an example in which the first sample data is image data. In actual application scenarios, there may be multiple data types for the first sample data, which may vary according to the actual application scenarios. Moreover, there may be multiple methods for determining the pseudo-sample data, which may be different according to the data type of the first sample data and the actual application scenarios. The embodiments of this specification do not specifically limit this.

[0040] In S104, the preset first risk identification model is trained based on the pseudo sample data and the second sample data determined by the data amount of the pseudo sample data to obtain the trained first risk identification model, and the risk identification accuracy of the trained first risk identification model is determined based on the third sample data.

[0041] Among them, the preset first risk identification model can be a model for risk identification built based on a preset deep learning algorithm. There may or may not be an intersection between the third sample data, the second sample data and the first sample data, that is, the third sample data and the second sample data may contain the same sample data as the first sample data, or may contain different sample data. In addition, the second sample data and the third sample data may be real user data stored on the server.

[0042] In implementation, since the pseudo-sample data is "non-real" data determined based on the characteristic information of the first sample data (i.e., it is impossible to identify real users based on the pseudo-sample data), if only the pseudo-sample data is used to train the preset first risk identification model, it may cause the trained first risk identification model to be unable to meet the usage requirements of the actual application scenario. Therefore, the second sample data can be selected based on the data amount of the pseudo-sample data, and the first risk model can be trained based on the pseudo-sample data and the second sample data, so that the trained first risk identification model can meet the usage requirements of the actual application scenario.

[0043] For example, the corresponding second sample data can be selected based on the preset sample selection ratio and the data volume of the pseudo sample data. Specifically, if the data volume of the pseudo sample data is 100 and the preset sample selection ratio is 2:1, the server can select 50 sample data as the second sample data (e.g., 50 sample data can be selected from the first sample data as the second sample data).

[0044] After obtaining the trained first risk identification model, the risk identification accuracy of the trained first risk identification model may be determined based on the third sample data.

[0045] Among them, the server can select corresponding business data as the first sample data, the second sample data and the third sample data according to the business scenario corresponding to the second risk identification model of the target device. For example, if the business scenario corresponding to the second risk identification model of the target device is a resource transfer scenario, the server can obtain the business data of the resource transfer business stored locally within the past month, and use n% of the obtained business data of the resource transfer business as the first sample data to determine the pseudo-sample data. Then, based on the data volume of the determined pseudo-sample data, m% is selected from the business data of the resource transfer business as the second sample data, and the preset first risk identification model is trained based on the pseudo-sample data and the second sample data to obtain the trained first risk identification model. Finally, x% is selected from the business data of the resource transfer business as the third sample data, and the third sample data is input into the trained first risk identification model to obtain the risk identification accuracy of the trained first risk identification model, wherein 0 <n%,m%,x%<1。

[0046] In S106, if it is determined that the pseudo sample data meets the preset risk identification requirement based on the risk identification accuracy of the trained first risk identification model, the pseudo sample data is sent to the target device.

[0047] Among them, the pseudo-sample data can be used to trigger the target device to train the second risk identification model based on the pseudo-sample data, and perform risk identification processing on the user based on the trained second risk identification model. The model structure of the second risk identification model can be the same as the first risk identification model or different. For example, the first risk identification model can be a model built based on a decision tree algorithm, the second risk identification model can also be a model built based on a decision tree algorithm, or the second risk identification model can also be a model built based on a neural network algorithm.

[0048] In implementation, for example, it is possible to determine whether the pseudo-sample data meets the risk identification requirements based on a preset accuracy threshold and the risk identification accuracy of the trained first risk identification model. Specifically, if the risk identification accuracy of the trained first risk identification model is greater than the preset accuracy threshold, it can be determined that the pseudo-sample data meets the preset risk identification requirements, and the server can send the pseudo-sample data to the target device.

[0049] In addition, there can be multiple target devices, and the business scenarios corresponding to the second risk identification models of different target devices may also be different. Therefore, the server can select different sample data (i.e., first sample data, second sample data, and third sample data) according to the different business scenarios corresponding to the second risk identification model of each target device, to determine different pseudo-sample data, and then send the determined pseudo-sample data corresponding to the business scenario to the corresponding target device based on the device identification of the target device.

[0050] An embodiment of the present specification provides a data processing method, which determines pseudo sample data based on feature information of first sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data, and a preset first risk identification model is trained based on the pseudo sample data and second sample data determined by the data amount of the pseudo sample data to obtain a trained first risk identification model, and the risk identification accuracy of the trained first risk identification model is determined based on third sample data; if it is determined that the pseudo sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo sample data is sent to a target device, and the pseudo sample data can be used to trigger the target device to train a second risk identification model based on the pseudo sample data, and perform risk identification processing on the user based on the trained second risk identification model. In this way, since the pseudo-sample data is determined based on the characteristic information of the first sample data, and the data obtained by restoring the pseudo-sample data is different from the first sample data, that is, the pseudo-sample data is both related to the first sample data and the first sample data cannot be obtained based on the pseudo-sample data, the private data of the server can be kept within the domain and the security of the private data can be guaranteed. At the same time, since the pseudo-sample data meets the preset risk identification requirements, the target device can train the second risk identification model based on the pseudo-sample data to improve the risk identification effect and risk identification accuracy of the second risk identification model obtained based on the training.

[0051] Embodiment 2

[0052] The embodiment of this specification provides a data processing method, the execution subject of the method can be a server, the server can be an independent server, or a server cluster composed of multiple servers. The method can specifically include the following steps:

[0053] There are many ways to determine pseudo sample data, such as Figure 2 As shown, pseudo sample data may be determined according to the following S202 to S208.

[0054] In S202 , a risk label of the first sample data is obtained, and a pseudo label is determined based on the risk label.

[0055] Among them, the first sample data may have multiple risk labels. For example, the risk label of the first sample data may be a 0 or 1 label. That is, if the risk label of the first sample data is 0, it can be indicated that the first sample data is risk-free sample data. If the risk label of the first sample data is 1, it can be indicated that the first sample data is risky sample data. In addition, the risk label of the first sample data may also have multiple different types, which may vary according to different actual application scenarios. The embodiments of this specification do not specifically limit the type of risk label of the first sample data.

[0056] In implementation, the risk label of the first sample data may be determined as a pseudo label. For example, if the risk label of the first sample data is a 0 or 1 label, the pseudo label may also be a 0 or 1 label.

[0057] In S204, random perturbation data is generated, and the random perturbation data and the pseudo label are input into the generator to obtain adversarial sample data.

[0058] The random disturbance data may be disturbance data consisting of random numbers generated based on a preset random generation algorithm.

[0059] In implementation, the generated random perturbation data and pseudo labels can be input into the generator to obtain multiple adversarial sample data corresponding to the pseudo samples.

[0060] In S206, the adversarial sample data and the first sample data are input into the discriminator, so that the discriminator performs true and false sample discrimination processing based on the feature information of the first sample data and the feature information of the adversarial sample data to obtain the sample authenticity probability.

[0061] Among them, Figure 3 As shown, the generator and the discriminator can be two components of a pre-trained adversarial neural network model, the generator can be used to generate simulated data (i.e., adversarial data), and the discriminator can be used to determine whether the simulated data generated by the generator is real data or simulated data, that is, the discriminator can be used to identify the simulated data generated by the generator as much as possible, so that after alternating training of the generator and the discriminator, pseudo sample data that is real enough to be mistaken for the real thing can be generated.

[0062] In implementation, Figure 3 As shown, the adversarial sample data generated by the generator and the first sample data can be input into the discriminator, so that the discriminator performs true and false sample discrimination processing based on the feature information of the first sample data and the feature information of the adversarial sample data to obtain the sample authenticity probability.

[0063] In S208, when the sample authenticity probability is greater than a preset probability threshold, the adversarial sample data is determined to be pseudo sample data.

[0064] In S104, the preset first risk identification model is trained based on the pseudo sample data and the second sample data determined by the data amount of the pseudo sample data to obtain the trained first risk identification model, and the risk identification accuracy of the trained first risk identification model is determined based on the third sample data.

[0065] In S106, if it is determined that the pseudo sample data meets the preset risk identification requirement based on the risk identification accuracy of the trained first risk identification model, the pseudo sample data is sent to the target device.

[0066] In implementation, since the quality of pseudo samples generated by the generator is different, in order to balance the impact of the increase in sample size and the noise brought by the pseudo samples on the first risk identification model, the pseudo sample data can be segmented according to the sample authenticity probability output by the discriminator, and then the pseudo sample data corresponding to different sample authenticity probabilities and the corresponding second sample data are input into the first risk identification model for training, and then the risk identification accuracy of the trained first risk identification model is determined based on the third sample data.

[0067] For example, the first risk identification model is trained according to the number of pseudo samples selected according to the sample authenticity probability output by the discriminator, and the risk identification accuracy of the trained first risk identification model is determined based on the third sample data, so that the following can be obtained: Figure 4 The correspondence between the sample authenticity probability (λ) and the risk identification accuracy (such as the area under the ROC curve and the coordinate axis (Area Under Curve, AUC)) shown in the figure is based on the performance of the trained first risk identification model on the third sample data. It can be seen that the pseudo-sample data corresponding to different sample authenticity probabilities have different improvements in the training of the first risk identification model.

[0068] based on Figure 4 It can be seen from the authenticity probability (λ) of the training samples and the risk identification accuracy that when λ is 0.7, the AUC is the largest, that is, the pseudo sample data with a sample authenticity probability of 0.7 can be determined as pseudo sample data that meets the preset risk identification requirements, and the pseudo sample data can be sent to the target device.

[0069] In addition to the above S202-208, if Figure 5 As shown, before S104 to S106, pseudo sample data may also be determined according to the following S502 to S506.

[0070] In S502, the first sample data is input into an encoder so that the encoder performs characterization learning on feature information of the first sample data to obtain a feature vector of the first sample data.

[0071] Among them, the encoder can be the part of the autoencoder used for representation learning of input information. The autoencoder (AE) is a type of artificial neural network (ANNs) used in semi-supervised learning and unsupervised learning. Its function is to perform representation learning on the input information by taking the input information as the learning target. The autoencoder may include an encoder and a decoder.

[0072] In implementation, the first sample data may be input into an encoder to obtain a feature vector of the first sample data.

[0073] In S504, sample enhancement processing is performed on the feature vector of the first sample data to obtain a processed feature vector.

[0074] In implementation, random perturbation data may be generated, and a processed feature vector may be generated based on the random perturbation data and the feature vector of the first sample data. For example, the feature vector of the first sample data may be graph vector data, and the random perturbation data may be added to the graph vector data to obtain the processed feature vector.

[0075] In addition, when the first sample data includes multiple data, the feature vector of the first sample data can also be weighted averaged based on the preset weight of the feature vector of the first sample data to obtain the processed feature vector. For example, assuming that the preset weight of the feature vector of the first sample data 1 is 0.8, the feature vector of the first sample data 1 is w1, the preset weight of the feature vector of the first sample data 2 is 0.2, and the feature vector of the first sample data 2 is w2, then the result of (0.8*w1+0.2*w2) / 2 can be used to determine the processed feature vector of the first sample data 1, wherein the risk labels of the first sample data 1 and the first sample data 2 can be the same.

[0076] The above-mentioned sample enhancement processing method for the feature vector of the first sample data is an optional and feasible processing method. In actual application scenarios, there may be a variety of different processing methods. Different processing methods can be selected according to different actual application scenarios. The embodiments of this specification do not make specific limitations on this.

[0077] In S506, the processed feature vector is input into the decoder for reconstruction processing to obtain reconstructed first sample data, and the reconstructed first sample data is used as pseudo sample data.

[0078] In implementation, Figure 6 As shown, the processed feature vector can be input into the decoder for reconstruction processing to obtain the reconstructed first sample data, and the reconstructed first sample data is used as the pseudo sample data. Since the feature vector input into the decoder is the processed feature vector, even if the pseudo sample data is restored, the obtained data is different from the first sample data. In this way, the pseudo sample data can retain a part of the feature information of the first sample data, but the first sample data cannot be restored based on the pseudo sample data. While improving the security of the privacy data, the pseudo sample data can be used to train the second risk identification model to improve the identification effect of the second risk identification model.

[0079] An embodiment of the present specification provides a data processing method, which determines pseudo sample data based on feature information of first sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data, and a preset first risk identification model is trained based on the pseudo sample data and second sample data determined by the data amount of the pseudo sample data to obtain a trained first risk identification model, and the risk identification accuracy of the trained first risk identification model is determined based on third sample data; if it is determined that the pseudo sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo sample data is sent to a target device, and the pseudo sample data can be used to trigger the target device to train a second risk identification model based on the pseudo sample data, and perform risk identification processing on the user based on the trained second risk identification model. In this way, since the pseudo-sample data is determined based on the characteristic information of the first sample data, and the data obtained by restoring the pseudo-sample data is different from the first sample data, that is, the pseudo-sample data is both related to the first sample data and the first sample data cannot be obtained based on the pseudo-sample data, the private data of the server can be kept within the domain and the security of the private data can be guaranteed. At the same time, since the pseudo-sample data meets the preset risk identification requirements, the target device can train the second risk identification model based on the pseudo-sample data to improve the risk identification effect and risk identification accuracy of the second risk identification model obtained based on the training.

[0080] Embodiment 3

[0081] like Fig. 7A and Figure 7B As shown, the embodiment of this specification provides a data processing method, the execution subject of the method can be a target device, and the target device can be a server or a terminal device, wherein the server can be an independent server or a server cluster composed of multiple servers, and the terminal device can be a device such as a personal computer, or a mobile terminal device such as a mobile phone or a tablet computer. The method can specifically include the following steps:

[0082] In S702, pseudo sample data sent by the server is received.

[0083] Among them, the pseudo-sample data can be the sample data determined based on the characteristic information of the first sample data and sent by the server when it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model. The trained first risk identification model can be trained based on the pseudo-sample data and the second sample data determined by the data amount of the pseudo-sample data, wherein the data obtained by restoring the pseudo-sample data is different from the first sample data.

[0084] In S704, the preset second risk identification model is trained based on the pseudo sample data to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model.

[0085] Among them, the model structure of the second risk identification model can be the same as or different from the first risk identification model. For example, the first risk identification model can be a model constructed based on a decision tree algorithm, and the second risk identification model can also be a model constructed based on a decision tree algorithm, or the second risk identification model can also be a model constructed based on a neural network algorithm.

[0086] During implementation, when training the preset second risk identification model, in order to make the trained second risk identification model meet the risk identification requirements of the target device, the local fourth sample data and the received pseudo sample data can also be input into the preset second risk identification model for training to obtain a pre-trained second risk identification model.

[0087] An embodiment of the present specification provides a data processing method, which receives pseudo-sample data sent by a server, wherein the pseudo-sample data is sample data determined based on feature information of first sample data and sent by the server when the server determines that the pseudo-sample data meets preset risk identification requirements based on the risk identification accuracy of a trained first risk identification model, wherein the trained first risk identification model is trained based on the pseudo-sample data and second sample data determined by the data volume of the pseudo-sample data, wherein data obtained by restoring the pseudo-sample data is different from the first sample data, and a preset second risk identification model is trained based on the pseudo-sample data to obtain a pre-trained second risk identification model, so as to perform risk identification processing on users based on the pre-trained second risk identification model. In this way, since the pseudo-sample data is determined by the server based on the characteristic information of the first sample data, and the data obtained by restoring the pseudo-sample data is different from the first sample data, that is, the pseudo-sample data is both related to the first sample data and the first sample data cannot be obtained based on the pseudo-sample data, the private data of the server can be kept within the domain and the security of the private data can be guaranteed. At the same time, since the pseudo-sample data meets the preset risk identification requirements, the target device can train the second risk identification model based on the pseudo-sample data to improve the risk identification effect and risk identification accuracy of the second risk identification model obtained based on the training.

[0088] Embodiment 4

[0089] like Figure 8As shown, the embodiment of this specification provides a data processing method, the execution subject of the method can be a target device, and the target device can be a server or a terminal device, wherein the server can be an independent server or a server cluster composed of multiple servers, and the terminal device can be a device such as a personal computer, or a mobile terminal device such as a mobile phone or a tablet computer. The method can specifically include the following steps:

[0090] In S702, pseudo sample data sent by the server is received.

[0091] Among them, the pseudo-sample data can be the sample data determined based on the characteristic information of the first sample data and sent by the server when it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model. The trained first risk identification model can be trained based on the pseudo-sample data and the second sample data determined by the data amount of the pseudo-sample data, wherein the data obtained by restoring the pseudo-sample data is different from the first sample data.

[0092] In S704, the preset second risk identification model is trained based on the pseudo sample data to obtain a pre-trained second risk identification model.

[0093] In S802, when it is detected that the target user triggers the execution of the target service, characteristic data of the target user is obtained.

[0094] Among them, the target service can be any service that the target device can provide to the user. For example, the target service can be a resource transfer service, an identity authentication service, etc. The characteristic data of the target user may include the target user's attribute information (such as information that can characterize one or more attributes of the user's personal attributes), behavior information (such as transfer behavior, call behavior, etc.), etc.

[0095] In implementation, taking the target device as a client used by an organization that can provide resource transfer services to users as an example, the target service can be a resource transfer service. The target user can trigger the start of a resource transfer application installed in the client and trigger the start of the resource transfer service (i.e., the target service). When the client detects that the target user triggers the execution of the resource transfer service, it can obtain the target user's attribute information and resource transfer behavior information (such as resource transfer quantity, resource transfer time, resource transfer object, etc.) characteristic data.

[0096] In S804, the characteristic data of the target user is input into a pre-trained second risk identification model to obtain a target risk identification result of the target user.

[0097] In implementation, the target user's attribute information and resource transfer behavior information (such as resource transfer quantity, resource transfer time, resource transfer object, etc.) and other characteristic data can be input into the pre-trained second risk identification model to obtain the target risk identification result of the target user.

[0098] In S806 , based on the target risk identification result, it is determined whether there is a risk in triggering the execution of the target business.

[0099] In implementation, if it is determined that there is a risk in triggering the execution of the target business based on the target risk identification result, a preset alarm message can be output to the target user and the execution of the target business can be stopped.

[0100] An embodiment of the present specification provides a data processing method, which receives pseudo-sample data sent by a server, wherein the pseudo-sample data is sample data determined based on feature information of first sample data and sent by the server when the server determines that the pseudo-sample data meets preset risk identification requirements based on the risk identification accuracy of a trained first risk identification model, wherein the trained first risk identification model is trained based on the pseudo-sample data and second sample data determined by the data volume of the pseudo-sample data, wherein data obtained by restoring the pseudo-sample data is different from the first sample data, and a preset second risk identification model is trained based on the pseudo-sample data to obtain a pre-trained second risk identification model, so as to perform risk identification processing on users based on the pre-trained second risk identification model. In this way, since the pseudo-sample data is determined by the server based on the characteristic information of the first sample data, and the data obtained by restoring the pseudo-sample data is different from the first sample data, that is, the pseudo-sample data is both related to the first sample data and the first sample data cannot be obtained based on the pseudo-sample data, the private data of the server can be kept within the domain and the security of the private data can be guaranteed. At the same time, since the pseudo-sample data meets the preset risk identification requirements, the target device can train the second risk identification model based on the pseudo-sample data to improve the risk identification effect and risk identification accuracy of the second risk identification model obtained based on the training.

[0101] Embodiment 5

[0102] The embodiment of the present specification provides a data processing system, which includes: a server and a target device, wherein:

[0103] The server can be used to determine pseudo-sample data based on feature information of the first sample data, wherein data obtained by restoring the pseudo-sample data is different from the first sample data; a preset first risk identification model is trained based on the pseudo-sample data and second sample data determined by the data volume of the pseudo-sample data to obtain the trained first risk identification model; and the risk identification accuracy of the trained first risk identification model is determined based on the third sample data; if it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo-sample data is sent to the target device.

[0104] The target device can be used to train a preset second risk identification model based on pseudo sample data to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model.

[0105] For example, there may be multiple servers and target devices in a data processing system. Taking risk user identification in a resource transfer scenario as an example, assuming that the target device and multiple servers respectively store characteristic data of several users (such as resource transfer behavior information, etc.), if the federated learning mechanism is used to train the preset second risk identification model of the target device, there are two ways: Method 1, a third-party trusted environment can be established, and multiple servers and target devices can upload their respective sample data to the third-party trusted environment through an encrypted data channel, and train the second risk identification model in the third-party trusted environment based on the sample data sent by the server and the target device; Method 2, multiple servers and target devices use distributed training to perform back propagation in the form of remote synchronous transmission of model gradients, and update the model parameters of the second risk identification model.

[0106] First, both of the above methods require that the risk identification models of multiple servers and target devices are the same, that is, the first risk identification model of the server and the second risk identification model of the target device are the same model. In actual application scenarios, the probability that the risk identification models adopted by the server and the target device have the same model structure is small, and the feasibility of the above two training methods is low. Secondly, method one requires the server to transfer sample data to a third-party trusted environment. Since the sample data is the user's private data, there are compliance issues when the private data is out of the domain. Even if there is no compliance issue when the private data is out of the domain, there is a risk of the sample data being stolen during the transmission process. Therefore, method one has the problem of poor data security. Finally, the distributed training framework of method two requires a lot of manpower and material resources, and the implementation process is relatively difficult.

[0107] By adopting the data processing system provided in the specification of the present application, first, pseudo-sample data can be used to replace real sample data for data output, avoiding reliance on complex message encryption transmission and distributed training processes. The pseudo-sample data can be sent to the target device after being produced on the server, and the target device can perform model training on the local second risk identification model based on the pseudo-sample data. There is no need to make special requirements on whether the model structure of the first risk identification model on the server and the second risk identification model of the target device are the same.

[0108] Secondly, the pseudo-sample data is determined based on the characteristic information of the first sample data, and the data obtained by restoring the pseudo-sample data is different from the first sample data. Therefore, it can help the target device to perform enhanced training of the model without leaking privacy data and improving the security of privacy data. It can also ensure that the sample data can be transmitted without relying on complex encryption technology (that is, since the pseudo-sample data cannot be restored to the first sample data, the pseudo-sample data can be sent directly to the target device during the data transmission process, and there is no problem of privacy data leakage), reducing data processing pressure and improving data processing efficiency.

[0109] In addition, the process of determining pseudo-sample data and the model training process of the target device are decoupled, and there is no need to develop a complex distributed training framework, which saves resources and improves data processing efficiency.

[0110] The embodiments of the present specification provide a data processing system. Since the pseudo-sample data is determined by the server based on the characteristic information of the first sample data, and the data obtained by restoring the pseudo-sample data is different from the first sample data, that is, the pseudo-sample data is related to the first sample data, but the first sample data cannot be obtained based on the pseudo-sample data. In this way, the private data of the server does not leave the domain, and the security of the private data is guaranteed. At the same time, since the pseudo-sample data meets the preset risk identification requirements, the target device can train the second risk identification model based on the pseudo-sample data to improve the risk identification effect and risk identification accuracy of the second risk identification model obtained based on the training.

[0111] Embodiment 6

[0112] The above is a data processing method provided in the embodiment of this specification. Based on the same idea, the embodiment of this specification also provides a data processing device, such as Fig. 9 shown.

[0113] The data processing device includes: a data determination module 901, a model training module 902 and a data sending module 903, wherein:

[0114] A data determination module 901 is used to determine pseudo sample data based on feature information of first sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data;

[0115] A model training module 902 is used to train a preset first risk identification model based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data to obtain a trained first risk identification model, and determine the risk identification accuracy of the trained first risk identification model based on third sample data;

[0116] The data sending module 903 is used to send the pseudo-sample data to the target device if it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model. The pseudo-sample data is used to trigger the target device to train the second risk identification model based on the pseudo-sample data, and perform risk identification processing on the user based on the trained second risk identification model.

[0117] In the embodiment of this specification, the data determination module 901 is used to:

[0118] Acquire a risk label of the first sample data, and determine a pseudo label based on the risk label;

[0119] Generate random perturbation data, and input the random perturbation data and the pseudo label into the generator to obtain adversarial sample data;

[0120] Inputting the adversarial sample data and the first sample data into the discriminator, so that the discriminator performs true and false sample discrimination processing based on feature information of the first sample data and feature information of the adversarial sample data to obtain a sample authenticity probability;

[0121] When the sample authenticity probability is greater than a preset probability threshold, the adversarial sample data is determined as the pseudo sample data.

[0122] In the embodiment of this specification, the data determination module 901 is used to:

[0123] Inputting the first sample data into an encoder so that the encoder performs characterization learning on feature information of the first sample data to obtain a feature vector of the first sample data;

[0124] Performing sample enhancement processing on the feature vector of the first sample data to obtain a processed feature vector;

[0125] The processed feature vector is input into a decoder for reconstruction processing to obtain reconstructed first sample data, and the reconstructed first sample data is used as the pseudo sample data.

[0126] In the embodiment of this specification, the data determination module 901 is used to:

[0127] Generate random disturbance data, and generate the processed feature vector based on the random disturbance data and the feature vector of the first sample data.

[0128] In the embodiment of this specification, the first sample data includes a plurality of data, and the data determination module 901 is used to:

[0129] Based on the preset weight of the feature vector of the first sample data, weighted averaging processing is performed on the feature vector of the first sample data to obtain the processed feature vector.

[0130] An embodiment of the present specification provides a data processing device, which determines pseudo sample data based on feature information of first sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data, and trains a preset first risk identification model based on the pseudo sample data and second sample data determined by the data amount of the pseudo sample data to obtain a trained first risk identification model, and determines the risk identification accuracy of the trained first risk identification model based on third sample data; if it is determined that the pseudo sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo sample data is sent to a target device, and the pseudo sample data can be used to trigger the target device to train a second risk identification model based on the pseudo sample data, and perform risk identification processing on the user based on the trained second risk identification model. In this way, since the pseudo-sample data is determined based on the characteristic information of the first sample data, and the data obtained by restoring the pseudo-sample data is different from the first sample data, that is, the pseudo-sample data is both related to the first sample data and the first sample data cannot be obtained based on the pseudo-sample data, the private data of the server can be kept within the domain and the security of the private data can be guaranteed. At the same time, since the pseudo-sample data meets the preset risk identification requirements, the target device can train the second risk identification model based on the pseudo-sample data to improve the risk identification effect and risk identification accuracy of the second risk identification model obtained based on the training.

[0131] Embodiment 7

[0132] Based on the same idea, the embodiment of this specification also provides a data processing device, such as Fig.10 shown.

[0133] The data processing device comprises: a data receiving module 1001 and a model training module 1002, wherein:

[0134] The data receiving module 1001 is used to receive pseudo sample data sent by the server, wherein the pseudo sample data is sample data determined based on feature information of the first sample data and sent by the server when the server determines that the pseudo sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, wherein the trained first risk identification model is trained based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data;

[0135] The model training module 1002 is used to train the preset second risk identification model based on the pseudo sample data to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model.

[0136] In the embodiment of this specification, the device further includes:

[0137] A data acquisition module, used to acquire characteristic data of the target user when it is detected that the target user triggers the execution of the target service;

[0138] A result acquisition module, used for inputting the characteristic data of the target user into the pre-trained second risk identification model to obtain a target risk identification result of the target user;

[0139] The risk identification module is used to determine whether there is a risk in triggering the execution of the target business based on the target risk identification result.

[0140] An embodiment of the present specification provides a data processing device, which receives pseudo-sample data sent by a server, wherein the pseudo-sample data is sample data determined based on feature information of first sample data and sent by the server when the server determines that the pseudo-sample data meets preset risk identification requirements based on the risk identification accuracy of a trained first risk identification model. The trained first risk identification model is obtained by training based on the pseudo-sample data and second sample data determined by the data volume of the pseudo-sample data, wherein data obtained by restoring the pseudo-sample data is different from the first sample data, and a preset second risk identification model is trained based on the pseudo-sample data to obtain a pre-trained second risk identification model, so as to perform risk identification processing on users based on the pre-trained second risk identification model. In this way, since the pseudo-sample data is determined by the server based on the characteristic information of the first sample data, and the data obtained by restoring the pseudo-sample data is different from the first sample data, that is, the pseudo-sample data is both related to the first sample data and the first sample data cannot be obtained based on the pseudo-sample data, the private data of the server can be kept within the domain and the security of the private data can be guaranteed. At the same time, since the pseudo-sample data meets the preset risk identification requirements, the target device can train the second risk identification model based on the pseudo-sample data to improve the risk identification effect and risk identification accuracy of the second risk identification model obtained based on the training.

[0141] Embodiment 8

[0142] Based on the same idea, the embodiment of this specification also provides a data processing device, such as Fig.11 shown.

[0143] The data processing device may have relatively large differences due to different configurations or performances, and may include one or more processors 1101 and memory 1102, and the memory 1102 may store one or more storage applications or data. Among them, the memory 1102 may be a short-term storage or a persistent storage. The application stored in the memory 1102 may include one or more modules (not shown in the figure), and each module may include a series of computer executable instructions in the data processing device. Furthermore, the processor 1101 may be configured to communicate with the memory 1102 and execute a series of computer executable instructions in the memory 1102 on the data processing device. The data processing device may also include one or more power supplies 1103, one or more wired or wireless network interfaces 1104, one or more input and output interfaces 1105, and one or more keyboards 1106.

[0144] Specifically in this embodiment, the data processing device includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer executable instructions in the data processing device, and the one or more programs are configured to be executed by one or more processors, including computer executable instructions for performing the following:

[0145] Determining pseudo sample data based on feature information of the first sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data;

[0146] Training a preset first risk identification model based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data to obtain a trained first risk identification model, and determining a risk identification accuracy rate of the trained first risk identification model based on third sample data;

[0147] If it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo-sample data is sent to the target device, and the pseudo-sample data is used to trigger the target device to train the second risk identification model based on the pseudo-sample data, and perform risk identification processing on the user based on the trained second risk identification model.

[0148] Optionally, determining the pseudo sample data based on the feature information of the first sample data includes:

[0149] Acquire a risk label of the first sample data, and determine a pseudo label based on the risk label;

[0150] Generate random perturbation data, and input the random perturbation data and the pseudo label into the generator to obtain adversarial sample data;

[0151] Inputting the adversarial sample data and the first sample data into the discriminator, so that the discriminator performs true and false sample discrimination processing based on feature information of the first sample data and feature information of the adversarial sample data to obtain a sample authenticity probability;

[0152] When the sample authenticity probability is greater than a preset probability threshold, the adversarial sample data is determined as the pseudo sample data.

[0153] Optionally, determining the pseudo sample data based on the feature information of the first sample data includes:

[0154] Inputting the first sample data into an encoder so that the encoder performs characterization learning on feature information of the first sample data to obtain a feature vector of the first sample data;

[0155] Performing sample enhancement processing on the feature vector of the first sample data to obtain a processed feature vector;

[0156] The processed feature vector is input into a decoder for reconstruction processing to obtain reconstructed first sample data, and the reconstructed first sample data is used as the pseudo sample data.

[0157] Optionally, performing sample enhancement processing on the feature vector of the first sample data to obtain a processed feature vector includes:

[0158] Generate random disturbance data, and generate the processed feature vector based on the random disturbance data and the feature vector of the first sample data.

[0159] Optionally, the first sample data includes a plurality of data, and the performing sample enhancement processing on the feature vector of the first sample data to obtain the processed feature vector includes:

[0160] Based on the preset weight of the feature vector of the first sample data, weighted averaging processing is performed on the feature vector of the first sample data to obtain the processed feature vector.

[0161] In addition, the data processing device includes a memory and one or more programs, wherein the one or more programs are stored in the memory, and the one or more programs may include one or more modules, and each module may include a series of computer-executable instructions for the data processing device, and the one or more programs configured to be executed by one or more processors include computer-executable instructions for performing the following:

[0162] Receive pseudo sample data sent by a server, the pseudo sample data being sample data determined based on feature information of first sample data and sent by the server when the server determines that the pseudo sample data meets a preset risk identification requirement based on the risk identification accuracy of the trained first risk identification model, the trained first risk identification model being trained based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data;

[0163] Based on the pseudo sample data, a preset second risk identification model is trained to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model.

[0164] Optionally, the method further comprises:

[0165] When it is detected that the target user triggers the execution of the target service, acquiring characteristic data of the target user;

[0166] Inputting the characteristic data of the target user into the pre-trained second risk identification model to obtain a target risk identification result of the target user;

[0167] Based on the target risk identification result, determine whether there is a risk in triggering the execution of the target business.

[0168] The embodiments of the present specification provide a data processing device. Since the pseudo-sample data is determined by the server based on the characteristic information of the first sample data, and the data obtained by restoring the pseudo-sample data is different from the first sample data, that is, the pseudo-sample data is related to the first sample data, but the first sample data cannot be obtained based on the pseudo-sample data, the private data of the server can be kept within the domain, thereby ensuring the security of the private data. At the same time, since the pseudo-sample data meets the preset risk identification requirements, the target device can train the second risk identification model based on the pseudo-sample data to improve the risk identification effect and risk identification accuracy of the second risk identification model obtained based on the training.

[0169] Embodiment 9

[0170] The embodiments of this specification also provide a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, each process of the above-mentioned data processing method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0171] The embodiments of the present specification provide a computer-readable storage medium. Since the pseudo-sample data is determined by the server based on the characteristic information of the first sample data, and the data obtained by restoring the pseudo-sample data is different from the first sample data, that is, the pseudo-sample data is associated with the first sample data, but the first sample data cannot be obtained based on the pseudo-sample data, the private data of the server can be kept within the domain, thereby ensuring the security of the private data. At the same time, since the pseudo-sample data meets the preset risk identification requirements, the target device can train the second risk identification model based on the pseudo-sample data to improve the risk identification effect and risk identification accuracy of the second risk identification model obtained based on the training.

[0172] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0173] In the 1990s, improvements to a technology could be clearly distinguished as hardware improvements (for example, improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the method flow). However, with the development of technology, many improvements to the method flow today can be regarded as direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using a hardware entity module. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to ask a chip manufacturer to design and produce a dedicated integrated circuit chip. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0174] The controller can be implemented in any appropriate manner, for example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of controllers include, but are not limited to, the following microcontrollers: ARC625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in a purely computer-readable program code manner, the controller can be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, this controller can be considered as a hardware component, and the devices included therein for implementing various functions can also be regarded as structures within the hardware component. Or even, the devices for implementing various functions can be regarded as both software modules for implementing the method and structures within the hardware component.

[0175] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0176] For the convenience of description, the above devices are described in terms of functions and are divided into various units. Of course, when implementing one or more embodiments of this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0177] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, one or more embodiments of this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, one or more embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0178] The embodiments of this specification are described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device produce a device for implementing the functions specified in one process or multiple processes in the flowchart and / or one box or multiple boxes in the block diagram.

[0179] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device that implements the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0180] These computer program instructions may also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0181] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0182] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0183] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0184] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0185] It should be understood by those skilled in the art that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, one or more embodiments of this specification may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, one or more embodiments of this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0186] One or more embodiments of the present specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. One or more embodiments of the present specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0187] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0188] The above description is only an embodiment of the present specification and is not intended to limit the present specification. For those skilled in the art, the present specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification shall be included in the scope of the claims of the present specification.

Claims

1. A data processing method, comprising: Determine pseudo sample data based on feature information of the first sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data, the first sample data includes real user data consisting of business data generated by the user when triggering the execution of the business and attribute data of the user, and the pseudo sample data cannot identify a real user; Training a preset first risk identification model based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data to obtain a trained first risk identification model, and determining a risk identification accuracy rate of the trained first risk identification model based on third sample data, wherein the second sample data and the third sample data are real user data stored on the server; If it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo-sample data is sent to the target device, and the pseudo-sample data is used to trigger the target device to train the second risk identification model based on the pseudo-sample data, and perform risk identification processing on the user based on the trained second risk identification model.

2. The method according to claim 1, wherein determining the pseudo sample data based on the feature information of the first sample data comprises: Acquire a risk label of the first sample data, and determine a pseudo label based on the risk label; Generate random perturbation data, and input the random perturbation data and the pseudo label into a generator to obtain adversarial sample data; Inputting the adversarial sample data and the first sample data into a discriminator, so that the discriminator performs true and false sample discrimination processing based on feature information of the first sample data and feature information of the adversarial sample data to obtain a sample authenticity probability; When the sample authenticity probability is greater than a preset probability threshold, the adversarial sample data is determined as the pseudo sample data.

3. The method according to claim 1, wherein determining the pseudo sample data based on the feature information of the first sample data comprises: Inputting the first sample data into an encoder so that the encoder performs characterization learning on feature information of the first sample data to obtain a feature vector of the first sample data; Performing sample enhancement processing on the feature vector of the first sample data to obtain a processed feature vector; The processed feature vector is input into a decoder for reconstruction processing to obtain reconstructed first sample data, and the reconstructed first sample data is used as the pseudo sample data.

4. The method according to claim 3, wherein the step of performing sample enhancement processing on the feature vector of the first sample data to obtain the processed feature vector comprises: Generate random disturbance data, and generate the processed feature vector based on the random disturbance data and the feature vector of the first sample data.

5. The method according to claim 3, wherein the first sample data comprises a plurality of samples, and the step of performing sample enhancement processing on the feature vector of the first sample data to obtain the processed feature vector comprises: Based on the preset weight of the feature vector of the first sample data, weighted averaging processing is performed on the feature vector of the first sample data to obtain the processed feature vector.

6. A data processing method, comprising: Receive pseudo sample data sent by a server, the pseudo sample data being sample data determined based on feature information of first sample data and sent by the server when the server determines that the pseudo sample data meets preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the trained first risk identification model being trained based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data, the first sample data including business data generated by the user when triggering the execution of a business and real user data consisting of attribute data of the user, the pseudo sample data cannot identify a real user, and the second sample data is real user data stored by the server; Based on the pseudo sample data, a preset second risk identification model is trained to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model.

7. The method according to claim 6, further comprising: When it is detected that the target user triggers the execution of the target service, acquiring characteristic data of the target user; Inputting the characteristic data of the target user into the pre-trained second risk identification model to obtain a target risk identification result of the target user; Based on the target risk identification result, determine whether there is a risk in triggering the execution of the target business.

8. A data processing system, comprising a server and a target device, wherein: The server is used to determine pseudo sample data based on feature information of the first sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data; train a preset first risk identification model based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data to obtain a trained first risk identification model, and determine the risk identification accuracy of the trained first risk identification model based on the third sample data; if it is determined that the pseudo sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo sample data is sent to the target device, the first sample data includes real user data composed of business data generated by the user when triggering the execution of the business and attribute data of the user, the pseudo sample data cannot identify a real user, and the second sample data and the third sample data are real user data stored in the server; The target device is used to train a preset second risk identification model based on the pseudo sample data to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model.

9. A data processing device, comprising: a data determination module, configured to determine pseudo sample data based on feature information of the first sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data, the first sample data includes real user data consisting of business data generated by the user when triggering the execution of a business and attribute data of the user, and the pseudo sample data cannot identify a real user; a model training module, configured to train a preset first risk identification model based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data to obtain a trained first risk identification model, and determine a risk identification accuracy rate of the trained first risk identification model based on third sample data, wherein the second sample data and the third sample data are real user data stored on the server; A data sending module is used to send the pseudo-sample data to a target device if it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model. The pseudo-sample data is used to trigger the target device to train the second risk identification model based on the pseudo-sample data, and perform risk identification processing on the user based on the trained second risk identification model.

10. A data processing device, comprising: A data receiving module, used for receiving pseudo sample data sent by a server, wherein the pseudo sample data is sample data determined based on feature information of first sample data and sent by the server when the server determines that the pseudo sample data meets preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, wherein the trained first risk identification model is trained based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data, wherein the first sample data includes business data generated by the user when triggering the execution of the business and real user data composed of attribute data of the user, wherein the pseudo sample data cannot identify real users, and the second sample data is real user data stored by the server; The model training module is used to train the preset second risk identification model based on the pseudo sample data to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model.

11. A data processing device, comprising: processor; as well as a memory arranged to store computer executable instructions which, when executed, cause the processor to: Determine pseudo sample data based on feature information of the first sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data, the first sample data includes real user data consisting of business data generated by the user when triggering the execution of the business and attribute data of the user, and the pseudo sample data cannot identify a real user; Training a preset first risk identification model based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data to obtain a trained first risk identification model, and determining a risk identification accuracy rate of the trained first risk identification model based on third sample data, wherein the second sample data and the third sample data are real user data stored on the server; If it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo-sample data is sent to the target device, and the pseudo-sample data is used to trigger the target device to train the second risk identification model based on the pseudo-sample data, and perform risk identification processing on the user based on the trained second risk identification model.

12. A data processing device, comprising: processor; as well as a memory arranged to store computer executable instructions which, when executed, cause the processor to: Receive pseudo sample data sent by a server, the pseudo sample data being sample data determined based on feature information of first sample data and sent by the server when the server determines that the pseudo sample data meets preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the trained first risk identification model being trained based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data, wherein data obtained by restoring the pseudo sample data is different from the first sample data, the first sample data including business data generated by the user when triggering the execution of a business and real user data consisting of attribute data of the user, the pseudo sample data cannot identify a real user, and the second sample data is real user data stored by the server; Based on the pseudo sample data, a preset second risk identification model is trained to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model.

13. A storage medium for storing computer executable instructions, wherein the executable instructions implement the following process when executed: Based on the feature information of the first sample data, pseudo sample data is determined, wherein: The data obtained by restoring the pseudo sample data is different from the first sample data, the first sample data includes the service data generated by the user when triggering the execution of the service and the real user data composed of the attribute data of the user, and the pseudo sample data cannot identify the real user; Training a preset first risk identification model based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data to obtain a trained first risk identification model, and determining a risk identification accuracy rate of the trained first risk identification model based on third sample data, wherein the second sample data and the third sample data are real user data stored on the server; If it is determined that the pseudo-sample data meets the preset risk identification requirements based on the risk identification accuracy of the trained first risk identification model, the pseudo-sample data is sent to the target device, and the pseudo-sample data is used to trigger the target device to train the second risk identification model based on the pseudo-sample data, and perform risk identification processing on the user based on the trained second risk identification model.

14. A storage medium for storing computer executable instructions, wherein the executable instructions implement the following process when executed: Receive pseudo sample data sent by the server, the pseudo sample data being sample data determined based on feature information of the first sample data and sent by the server when the server determines that the pseudo sample data meets a preset risk identification requirement based on the risk identification accuracy of the trained first risk identification model, the trained first risk identification model being trained based on the pseudo sample data and second sample data determined by the data volume of the pseudo sample data, wherein: The data obtained by restoring the pseudo sample data is different from the first sample data. The first sample data includes business data generated by the user when triggering the execution of the business and real user data composed of the attribute data of the user. The pseudo sample data cannot identify the real user. The second sample data is the real user data stored by the server. Based on the pseudo sample data, a preset second risk identification model is trained to obtain a pre-trained second risk identification model, so as to perform risk identification processing on the user based on the pre-trained second risk identification model.

Images