WO2022257723A1 - 风险防控的方法、装置及设备 - Google Patents

风险防控的方法、装置及设备 Download PDF

Info

Publication number
WO2022257723A1
WO2022257723A1 PCT/CN2022/093895 CN2022093895W WO2022257723A1 WO 2022257723 A1 WO2022257723 A1 WO 2022257723A1 CN 2022093895 W CN2022093895 W CN 2022093895W WO 2022257723 A1 WO2022257723 A1 WO 2022257723A1
Authority
WO
WIPO (PCT)
Prior art keywords
model
risk prevention
control
terminal device
server
Prior art date
Application number
PCT/CN2022/093895
Other languages
English (en)
French (fr)
Inventor
傅欣艺
Original Assignee
支付宝(杭州)信息技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 支付宝(杭州)信息技术有限公司 filed Critical 支付宝(杭州)信息技术有限公司
Publication of WO2022257723A1 publication Critical patent/WO2022257723A1/zh
Priority to US18/526,555 priority Critical patent/US20240104570A1/en

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q20/00Payment architectures, schemes or protocols
    • G06Q20/38Payment protocols; Details thereof
    • G06Q20/40Authorisation, e.g. identification of payer or payee, verification of customer or shop credentials; Review and approval of payers, e.g. check credit lines or negative lists
    • G06Q20/401Transaction verification
    • G06Q20/4016Transaction verification involving fraud or risk level assessment in transaction processing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/60Protecting data
    • G06F21/62Protecting access to data via a platform, e.g. using keys or access control rules
    • G06F21/6218Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
    • G06F21/6245Protecting personal data, e.g. for financial or medical purposes
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/23Clustering techniques
    • G06F18/232Non-hierarchical techniques
    • G06F18/2321Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
    • G06F18/23213Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods

Definitions

  • This manual relates to the field of computer technology, especially to methods, devices and equipment for risk prevention and control.
  • the architecture of the current risk prevention and control system is usually made by the server to make risk decisions, and the terminal device collects the data of the corresponding business and uploads it to the server. Through the powerful computing power of the server, analyze the risk control strategy and risk prevention and control model of the above data, and finally output the risk control decision.
  • the terminal device needs to upload data to the server, which may bring security risks to the user's private data. Moreover, if the terminal device makes a centralized request, the resource consumption of the server will be large. In addition, Because different users use the same risk prevention and control model, the habits of different users and the preferences of each user are different. Therefore, the risk prevention and control model is difficult to meet the needs of different users, and the flexibility of the risk prevention and control model is poor. Based on this, it is necessary to provide a personalized risk prevention and control system based on shared learning between the terminal and the server (or terminal cloud). The risk prevention and control system can not only protect private data, but also reduce the interaction between the terminal device and the server. The server calculates the pressure, and it can also improve the flexibility of the risk prevention and control model.
  • the purpose of the embodiments of this specification is to provide a personalized risk prevention and control system based on shared learning between the terminal and the server (or terminal cloud).
  • the risk prevention and control system can not only protect private data, but also reduce Interaction, reduce server computing pressure, but also improve the flexibility of the risk prevention and control model.
  • a risk prevention and control method provided by the embodiment of this specification is applied to a terminal device, and the method includes: receiving the target service to be trained corresponding to the classification group to which the terminal device belongs issued by the server The initial risk prevention and control model.
  • Model training is performed on the initial risk prevention and control model based on pre-stored training sample data to obtain a risk prevention and control sub-model corresponding to the terminal device, and the training sample data at least includes information related to the user of the terminal device and the Data related to the target business.
  • a risk prevention and control method provided by an embodiment of this specification is applied to a server, and includes: acquiring information of a classification group to which a terminal device belongs. Based on the information of the classification group to which the terminal device belongs, the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs is obtained, and the initial risk prevention and control model is sent to the said initial risk prevention and control model.
  • the terminal device so that the terminal device performs model training on the initial risk prevention and control model based on the pre-stored training sample data, and obtains the risk prevention and control sub-model corresponding to the terminal device, and the training sample data includes at least Data related to the user of the terminal device and the target service.
  • the sub-model performs model fusion processing to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the risk prevention and control sub-model provided by the other terminal device is based on the other terminal device.
  • the stored training sample data is obtained by performing model training on the initial risk prevention and control model.
  • the risk prevention and control model is sent to the terminal device, so that the terminal device performs risk prevention and control processing on the acquired data of the target service based on the risk prevention and control model.
  • a risk prevention and control device includes: an initial model receiving module that receives an initial risk prevention and control model of a target service to be trained corresponding to the classification group to which the device belongs issued by the server.
  • the sub-model training module performs model training on the initial risk prevention and control model based on the pre-stored training sample data to obtain the risk prevention and control sub-model corresponding to the device, and the training sample data includes at least the user of the device Data related to the target business.
  • a sub-model sending module which sends the risk prevention and control sub-model to the server, so that the server performs model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the device belongs,
  • a risk prevention and control model corresponding to the classification group to which the device belongs is obtained.
  • the risk prevention and control module receives the risk prevention and control model corresponding to the classification group to which the device belongs sent by the server, and performs risk prevention and control processing on the acquired data of the target service based on the risk prevention and control model.
  • a risk prevention and control device includes: a group information acquisition module, which acquires information of a classification group to which a terminal device belongs.
  • the initial model delivery module based on the information of the classification group to which the terminal device belongs, obtains the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs, and transfers the initial risk prevention and control model to The control model is sent to the terminal device, so that the terminal device performs model training on the initial risk prevention and control model based on the pre-stored training sample data, and obtains the risk prevention and control sub-model corresponding to the terminal device.
  • the training sample data at least includes data related to the user of the terminal device and the target service.
  • the sub-model receiving module receives the risk prevention and control sub-model sent by the terminal device, and compares the risk prevention and control sub-model sent by the terminal device and other terminal devices in the classification group to which the terminal device belongs.
  • the provided risk prevention and control sub-model performs model fusion processing to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the risk prevention and control sub-model provided by the other terminal device is based on the other terminal device.
  • the training sample data pre-stored in other terminal devices is obtained by performing model training on the initial risk prevention and control model.
  • a risk control model sending module which sends the risk prevention and control model to the terminal device, so that the terminal device performs risk prevention and control processing on the acquired data of the target service based on the risk prevention and control model.
  • a risk prevention and control device provided in an embodiment of this specification includes a processor and a memory arranged to store computer-executable instructions.
  • the processor receives the The initial risk prevention and control model of the target business to be trained corresponding to the classification group to which the device belongs.
  • Model training is performed on the initial risk prevention and control model based on pre-stored training sample data to obtain a risk prevention and control sub-model corresponding to the device, and the training sample data includes at least the user of the device and the target business related data.
  • a risk prevention and control device includes a processor and a memory arranged to store computer-executable instructions, and when the executable instructions are executed, the processor: obtains the classification group to which the terminal device belongs group information. Based on the information of the classification group to which the terminal device belongs, the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs is obtained, and the initial risk prevention and control model is sent to the said initial risk prevention and control model.
  • the terminal device so that the terminal device performs model training on the initial risk prevention and control model based on the pre-stored training sample data, and obtains the risk prevention and control sub-model corresponding to the terminal device, and the training sample data includes at least Data related to the user of the terminal device and the target service.
  • receiving the risk prevention and control sub-model sent by the terminal device and providing risk prevention and control for the risk prevention and control sub-model sent by the terminal device and other terminal devices in the classification group to which the terminal device belongs
  • the sub-model performs model fusion processing to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the risk prevention and control sub-model provided by the other terminal device is based on the other terminal device.
  • the stored training sample data is obtained by performing model training on the initial risk prevention and control model.
  • the risk prevention and control model is sent to the terminal device, so that the terminal device performs risk prevention and control processing on the acquired data of the target service based on the risk prevention and control model.
  • the embodiment of this specification also provides a storage medium, which is used to store computer-executable instructions.
  • the executable instructions When executed, the following process is implemented: receiving the information to be trained corresponding to the classification group to which the terminal device belongs issued by the server; The initial risk prevention and control model of the target business. Model training is performed on the initial risk prevention and control model based on pre-stored training sample data to obtain a risk prevention and control sub-model corresponding to the terminal device, and the training sample data at least includes information related to the user of the terminal device and the Data related to the target business.
  • the embodiment of the present specification also provides a storage medium, which is used for storing computer-executable instructions, and the executable instructions implement the following process when executed: acquire the information of the classification group to which the terminal device belongs. Based on the information of the classification group to which the terminal device belongs, the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs is obtained, and the initial risk prevention and control model is sent to the said initial risk prevention and control model.
  • the terminal device so that the terminal device performs model training on the initial risk prevention and control model based on the pre-stored training sample data, and obtains the risk prevention and control sub-model corresponding to the terminal device, and the training sample data includes at least Data related to the user of the terminal device and the target service.
  • the sub-model performs model fusion processing to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the risk prevention and control sub-model provided by the other terminal device is based on the other terminal device.
  • the stored training sample data is obtained by performing model training on the initial risk prevention and control model.
  • the risk prevention and control model is sent to the terminal device, so that the terminal device performs risk prevention and control processing on the acquired data of the target service based on the risk prevention and control model.
  • Figure 1A is an embodiment of a risk prevention and control method in this specification
  • Figure 1B is a schematic diagram of a risk prevention and control process in this specification
  • Figure 2 is a schematic structural diagram of a risk prevention and control system in this specification
  • FIG. 3 is a schematic diagram of another risk prevention and control process in this specification.
  • Figure 4A is an embodiment of another risk prevention and control method in this manual
  • Fig. 4B is a schematic diagram of another risk prevention and control process in this manual.
  • FIG. 5 is a schematic diagram of another risk prevention and control process in this specification.
  • Figure 6 is an embodiment of a risk prevention and control device in this specification.
  • Figure 7 is another embodiment of a risk prevention and control device in this specification.
  • Fig. 8 is an embodiment of a risk prevention and control device in this specification.
  • the embodiment of this specification provides a risk prevention and control method
  • the execution body of the method may be a terminal device, wherein the terminal device may be a mobile phone, a tablet computer, a personal computer, and the like.
  • the method may include the following steps: In step S102, the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs is received from the server.
  • the server may be a server of a certain business (such as a transaction business or a financial business, etc.), specifically, the server may be a payment business server, or a server related to finance or instant messaging, etc., or , can also be a server for risk prevention and control of a certain business, which can be set according to the actual situation, which is not limited in the embodiment of this specification.
  • a server of a certain business such as a transaction business or a financial business, etc.
  • the server may be a payment business server, or a server related to finance or instant messaging, etc., or , can also be a server for risk prevention and control of a certain business, which can be set according to the actual situation, which is not limited in the embodiment of this specification.
  • the classification group can be a classification group obtained by dividing the terminal devices into groups by means of clustering, etc., and the classification group can be set in a variety of different ways, for example, it can be classified according to different age groups set
  • the division of groups, or the division of classification groups can also be carried out according to the different regions where the users are located, or the division of classification groups can also be carried out according to the length of time the user registers the target business, etc., which can be set according to the actual situation.
  • the embodiment of this specification does not limit this.
  • the target business can be any business that may be risky or requires risk prevention and control, such as payment business or transfer business, which can be set according to the actual situation, and is not limited in the embodiment of this specification.
  • the initial risk prevention and control model can be an untrained model for risk prevention and control of a certain business.
  • the initial risk prevention and control model can only have a model structure, and the model parameters in it are not accurate.
  • it can be model parameters
  • the unknown or preset numerical value or random value can be specifically set according to the actual situation, which is not limited in the embodiment of this specification.
  • the architecture of the current risk prevention and control system is usually made by the server to make risk decisions, and the terminal equipment collects the data of the corresponding business and uploads it to the server.
  • the terminal equipment collects the data of the corresponding business and uploads it to the server.
  • the server Through the powerful computing power of the server, analyze the risk control strategy and risk prevention and control model of the above data, and finally output the risk control decision.
  • the terminal device needs to upload data to the server, which may bring security risks to the user's private data.
  • the terminal device makes a centralized request, the resource consumption of the server will be large.
  • different users use the same risk prevention and control model, the habits of different users and the preferences of each user are different.
  • the risk prevention and control model is difficult to meet the needs of different users, and the flexibility of the risk prevention and control model is poor. Based on this, it is necessary to provide a personalized risk prevention and control system based on shared learning between the terminal and the server (or terminal cloud).
  • the risk prevention and control system can not only protect private data, but also reduce the interaction between the terminal device and the server.
  • the server calculates the pressure, and it can also improve the flexibility of the risk prevention and control model.
  • the embodiment of this specification provides a realizable technical solution, which may specifically include the following content:
  • Figure 2 taking into account the habits and preferences of different users, as well as the strength of risk prevention and control awareness, different users can By means of clustering, classification, or manual division of groups, users with the same attribute can be classified into the same classification group, and then the classification groups to which different users belong can be determined. Correspondingly, it can also be used to determine the The classification group to which the end device belongs.
  • the server can build corresponding model architectures of risk prevention and control models for different classification groups, wherein the model architectures of risk prevention and control models corresponding to different classification groups can be the same or different, and the model parameters in the model architecture
  • the parameter value can be unknown, or set randomly, or set according to expert experience, specifically, it can be set according to the actual situation, which is not limited in the embodiment of this specification.
  • the above risk prevention and control models set by the server for different classification groups can be used as the initial risk prevention and control model of the target business to be trained.
  • the server can acquire different classification groups and information of terminal devices or user information contained in different classification groups, and then, based on the information of terminal devices or information of users, assign the waiting list corresponding to the classification group to which each terminal device belongs.
  • the trained initial risk prevention and control model of the target business is sent to the terminal device, and the terminal device can receive the initial risk prevention and control model corresponding to the classification group to which the terminal device belongs sent by the server.
  • different age groups can be pre-set, such as 0-25 years old, 26-60 years old, and over 60 years old. Construct the model structure of the corresponding risk prevention and control model, in which the parameter values of the model parameters of each risk prevention and control model can be set randomly, so that the initial risk prevention and control models of the target business corresponding to three different age groups can be obtained. Then, the pre-recorded information or the information of each user of the target service registered in the server or the information of the terminal equipment used by the user can be obtained.
  • the classification group to which the user's terminal equipment belongs can be determined, specifically as , a user is 37 years old and belongs to the age group of 26-60 years old, can obtain the initial risk prevention and control model of the target business corresponding to the age group of 26-60 years old, and can send the initial risk prevention and control model to the User's terminal equipment.
  • the processing of the above step S102 can be triggered in many different ways, for example, it can be triggered when the terminal device requests the server to execute the target service, or it can also be that the server can issue a The initial risk prevention and control model corresponding to the classification group to which the terminal device belongs, or, the terminal device may periodically obtain the initial risk prevention and control model corresponding to the classification group to which the terminal device belongs, etc., which may be based on actual conditions It is set, which is not limited in the embodiment of this specification.
  • step S104 model training is performed on the initial risk prevention and control model based on the pre-stored training sample data to obtain a risk prevention and control sub-model corresponding to the terminal device.
  • the training sample data includes at least information related to the user of the terminal device and the target business data.
  • the training sample data may include relevant data generated by the user of the terminal device during the execution of the target service, that is, data related to the user of the terminal device and the target service, and the relevant data may be data generated within a specified time period, or It may be the data generated from the time when the user registers the target service to the current moment, and may be specifically set according to the actual situation, which is not limited in the embodiment of this specification.
  • the training sample data may also include the data related to the target service obtained by the terminal device from other terminal devices, servers or databases through the specified data acquisition method, or may include the data obtained by the terminal device through the specified data acquisition method from Data related to the user of the terminal device and the target service obtained by other terminal devices, servers, or databases.
  • the terminal can The training sample data in the device conducts model training on the initial risk prevention and control model, that is, you can select a sample data from the training sample data, input the sample data into the initial risk prevention and control model, and obtain the expressions about the model parameters, and then , you can select another sample data from the training sample data, input the sample data into the initial risk prevention and control model, and get the expressions about the model parameters again.
  • the model parameters in the initial risk prevention and control model can be updated by using the remaining sample data in the training sample data in the above way until the final result converges, so as to obtain better model parameters, and then obtain the trained risk prevention and control model , and the trained risk prevention and control model can be used as the risk prevention and control sub-model corresponding to the terminal device.
  • step S106 the risk prevention and control sub-model is sent to the server, so that the server performs model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the terminal device belongs, and obtains the classification group to which the terminal device belongs The risk prevention and control model corresponding to the group.
  • the terminal device can send the risk prevention and control sub-model to the server, and the server can receive the risk prevention and control sub-model corresponding to the terminal device.
  • multiple different terminal devices in the classification group to which the terminal device belongs It is also possible to train the risk prevention and control sub-model belonging to each terminal device through the above method, and each terminal device can provide the risk prevention and control sub-model it trains to the server, and the server can use the different sub-models in the classification group to which the terminal device belongs.
  • the risk prevention and control sub-model provided by the terminal device performs model fusion processing, so that through federated learning, the server can obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs.
  • step S108 the risk prevention and control model corresponding to the classification group to which the terminal device belongs is received from the server, and risk prevention and control processing is performed on the acquired data of the target service based on the risk prevention and control model.
  • the terminal device can receive the risk prevention and control model corresponding to the classification group to which the terminal device belongs sent by the server, and when the terminal device obtains the data of the target business, it can input the data of the target business into the risk prevention and control model In the process, the result of whether the data of the target business is at risk is obtained. If the result indicates that the data of the target business is at risk, the terminal device can be refused to execute the target business. If the result indicates that the data of the target business is not at risk, it can be Allow terminal equipment to perform target services.
  • the embodiment of this specification provides a risk prevention and control method.
  • the initial The risk prevention and control model When receiving the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs issued by the server, based on the pre-stored training sample data, the initial The risk prevention and control model performs model training to obtain the risk prevention and control sub-model corresponding to the terminal device, wherein the training sample data includes at least data related to the user of the terminal device and the target business, and then sends the risk prevention and control sub-model to The server performs model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the terminal device belongs to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the terminal device uses the risk prevention and control model provided by the server.
  • the risk prevention and control model corresponding to the classification group to which the terminal device belongs performs risk prevention and control processing on the acquired data of the target business.
  • the risk prevention and control model is determined through federated learning, thereby realizing shared learning based on the terminal and the server.
  • Personalized risk prevention and control, and the above method can not only protect private data, but also reduce the interaction between the terminal device and the server, reduce the computing pressure on the server, and improve the flexibility of the risk prevention and control model.
  • the embodiment of this specification provides a method for risk prevention and control.
  • the execution subject of the method may be a terminal device, where the terminal device may be a mobile phone, a tablet computer, a personal computer, and the like.
  • the method may specifically include the following steps: In step S302, the first model to be trained delivered by the server is received, and the parameter values of the model parameters in the first model are randomly generated parameter values or preset parameter values.
  • the first model can be constructed by a preset neural network model, and the neural network model can include various types, for example, it can include a convolutional neural network model, a recurrent neural network model, a generative confrontation network model, etc., which can be based on actual conditions It is set, which is not limited in the embodiment of this specification.
  • the parameter value of the model parameter in the first model may be a preset parameter value, for example, the parameter value of the model parameter in the first model may be a parameter value set according to expert experience, or, the model in the first model
  • the parameter value of the parameter may be a parameter value set according to relevant historical data, etc., and may be specifically set according to an actual situation, which is not limited in the embodiment of this specification.
  • the relevant data for the target business and users in the terminal device may not be output, but the corresponding training sample data may be retained in the terminal device, and the corresponding training data may be trained in the terminal device. model to get the corresponding model parameters.
  • the server can pre-set the corresponding algorithm, and can use this algorithm to build the first model
  • the model architecture, and the parameter values of the model parameters therein can be set, so that the server can obtain the first model to be trained. In order to be able to accurately classify terminal devices or users of terminal devices, it can be realized based on the training sample data in the terminal device.
  • the server may send the constructed first model to be trained to the terminal device, and the terminal device may receive the first model to be trained delivered by the server.
  • step S304 model training is performed on the first model based on the pre-stored training sample data to obtain the trained first model, and obtain the parameters and corresponding parameter values in the preset network layer in the trained first model .
  • the preset network layer may include one network layer, or may include multiple network layers, specifically, the first model is a convolutional neural network model, and the preset network layer may be multiple convolutional layers, or may be One or more of the convolutional layer, pooling layer and fully connected layer, etc., can be set according to the actual situation. In practical applications, the preset network layer in the first model after training can be the output layer. The previous network layer (such as fully connected layer, etc.).
  • training sample data pre-stored by the terminal device may be obtained, and model training may be performed on the first model by using the training sample data to obtain a trained first model. Then, the parameters and corresponding parameter values in the preset network layer in the first model after training can be obtained, for example, the previous network layer of the output layer in the first model after training (specifically such as fully connected layer, etc. ) parameters and corresponding parameter values, etc.
  • step S306 the parameters in the preset network layer in the trained first model and the corresponding parameter values are sent to the server, so that the server can base on the preset network in the trained first model provided by different terminal devices
  • the parameters in the layer and the corresponding parameter values are used to cluster different terminal devices to obtain the classification groups to which different terminal devices belong.
  • the server receives the parameters in the preset network layer in the trained first model and the corresponding parameter values sent by the terminal device, since the parameters in the preset network layer in the trained first model and The information of the training sample data in the terminal device is retained in the corresponding parameter value, therefore, the parameters and corresponding parameter values in the preset network layer in the trained first model can be used to replace the training sample data in the terminal device,
  • the terminal device or the user of the terminal device is divided into classification groups.
  • other terminal devices can also provide the server with the parameters and corresponding parameter values in the preset network layer in the first model after training in the above-mentioned manner, so that The server may obtain parameters and corresponding parameter values in the preset network layer in the trained first model provided by multiple different terminal devices. Then, the parameters in the preset network layer and the corresponding parameter values in the obtained trained first model can be clustered through a preset clustering algorithm to obtain the classification groups to which different terminal devices belong, so that The classification group to which the terminal device belongs is obtained.
  • step S306 can be various, and an optional processing method is provided below, which can specifically include the following content: use the preset second encryption algorithm to encrypt the preset data in the first model after training
  • the parameters in the network layer and corresponding parameter values are encrypted to obtain encrypted parameters and corresponding encrypted parameter values, and the encrypted parameters and corresponding encrypted parameter values are sent to the server.
  • the second encryption algorithm may include multiple types, such as a homomorphic encryption algorithm, a partially homomorphic encryption algorithm, or a fully homomorphic encryption algorithm, etc., which may be specifically set according to actual conditions, which is not limited in this embodiment of the specification.
  • an encryption algorithm (that is, the second encryption algorithm) can be preset, and the second encryption algorithm can be used to compare the parameters and corresponding parameters in the preset network layer in the first model after training.
  • the parameter value is encrypted to obtain the encrypted parameter and the corresponding encrypted parameter value, and then the encrypted parameter and the corresponding encrypted parameter value are sent to the server, so as to ensure the security of the data during the data transmission process .
  • step S308 the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs is received from the server.
  • the initial risk prevention and control model can be constructed by a preset neural network model, and the neural network model can include various types, for example, it can include a convolutional neural network model, a recurrent neural network model, a generative confrontation network model, and the like.
  • the above-mentioned first model may be different from the initial risk prevention and control model.
  • the classification group to which the terminal device belongs is based on the parameters and corresponding parameter values in the preset network layer in the trained first model provided by different terminal devices, and clusters different terminal devices through a preset clustering algorithm.
  • the clustering algorithm can be K-means clustering algorithm, etc.
  • the clustering algorithm can also include a variety of other clustering algorithms, which can be determined according to the actual situation selected, which is not limited in the embodiments of this specification.
  • step S310 model training is performed on the initial risk prevention and control model based on the above training sample data to obtain a risk prevention and control sub-model corresponding to the terminal device.
  • the training sample data includes at least data related to the user of the terminal device and the target business.
  • step S312 the risk prevention and control sub-model is encrypted using the preset first encryption algorithm to obtain the encrypted risk prevention and control sub-model, and the encrypted risk prevention and control sub-model is sent to the server, so that the server Perform model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the terminal device belongs, to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs.
  • the first encryption algorithm may include multiple types, such as homomorphic encryption algorithm, partially homomorphic encryption algorithm or fully homomorphic encryption algorithm, etc.
  • the first encryption algorithm may be the same as the above-mentioned second encryption algorithm, or may be the same as the above-mentioned second encryption algorithm.
  • Algorithms are different, and can be specifically set according to actual conditions, which is not limited in the embodiments of this specification.
  • step S314 the risk prevention and control model corresponding to the classification group to which the terminal device belongs is received from the server.
  • the risk prevention and control model corresponding to the classification group to which the terminal device belongs is obtained through the above processing, and the terminal device can use the risk prevention and control model to perform risk detection and risk prevention and control on the business data of the target business, which may specifically include the following steps S316 to S320 processing.
  • step S316 the service data of the target service to be detected is acquired.
  • step S318 the above-mentioned business data is input into the risk prevention and control model to detect whether the business data has preset risks, and obtain corresponding detection results.
  • step S320 if the detection result indicates that the service data has a preset risk, cancel the service processing of the terminal device to execute the target service.
  • the risk prevention and control model corresponding to the classification group to which the terminal device belongs may also be updated, which specifically may include the following steps S322 to S326.
  • step S322 if it is detected that the training sample data is updated, the updated training sample data is acquired.
  • the training sample data stored in the terminal device may include a data identifier, and the terminal device may periodically or irregularly detect whether the training sample data is updated. Whether the data identification of the training sample data is increased, so as to determine whether the training sample data is updated, or, it can be detected whether the increased data volume in the training sample data exceeds a predetermined threshold, so as to determine whether the training sample data is updated, etc., which can be set according to the actual situation , which is not limited in the embodiment of this specification.
  • step S324 the risk prevention and control model is trained based on the updated training sample data, the trained risk prevention and control model is obtained, and the trained risk prevention and control model is sent to the server, so that the server has The risk prevention and control model corresponding to the classification group is updated to obtain the updated risk prevention and control model.
  • the terminal device can send the trained risk prevention and control model to the server, and the server can receive the trained risk prevention and control model.
  • multiple different terminal devices in the classification group to which the terminal device belongs The risk prevention and control model belonging to each terminal device can also be trained through the above method, and each terminal device can provide the risk prevention and control model trained by it to the server, and the server can The risk prevention and control model provided performs model fusion processing to update the risk prevention and control model corresponding to the classification group to which the terminal device belongs, so that through federated learning, the server can continuously update the risk prevention and control corresponding to the classification group to which the terminal device belongs model, in addition, the server can also update the risk prevention and control model corresponding to the classification group to which the terminal device belongs only through the risk prevention and control model provided by the terminal device, which can be set according to the actual situation. No limit.
  • step S326 the updated risk prevention and control model sent by the server is received, and risk prevention and control processing is performed on the data of the target business based on the updated risk prevention and control model.
  • the above-mentioned process of updating the risk prevention and control model corresponding to the classification group to which the terminal device belongs can be realized in addition to the above-mentioned steps S322 to S326, and can also be realized in the following manner, which specifically includes the following steps A2 to A4 processing.
  • step A2 if it is detected that the training sample data is updated, the updated training sample data is acquired.
  • step A4 the risk prevention and control model is trained based on the updated training sample data to obtain the trained risk prevention and control model, and the target business data is subjected to risk prevention and control processing based on the trained risk prevention and control model.
  • the embodiment of this specification provides a risk prevention and control method.
  • the initial The risk prevention and control model When receiving the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs issued by the server, based on the pre-stored training sample data, the initial The risk prevention and control model performs model training to obtain the risk prevention and control sub-model corresponding to the terminal device, wherein the training sample data includes at least data related to the user of the terminal device and the target business, and then sends the risk prevention and control sub-model to The server performs model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the terminal device belongs to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the terminal device uses the risk prevention and control model provided by the server.
  • the risk prevention and control model corresponding to the classification group to which the terminal device belongs performs risk prevention and control processing on the acquired data of the target business.
  • the risk prevention and control model is determined through federated learning, thereby realizing shared learning based on the terminal and the server.
  • Personalized risk prevention and control, and the above method can not only protect private data, but also reduce the interaction between the terminal device and the server, reduce the computing pressure on the server, and improve the flexibility of the risk prevention and control model.
  • a personalized risk prevention and control model is customized for each category of users, and the risk prevention and control model is trained through clustering algorithms combined with federated learning, which can realize your high-end
  • the data in the device does not need to be transmitted to the server, thereby protecting user privacy.
  • the embodiment of this specification provides a risk prevention and control method
  • the execution subject of the method may be a server, for example, a server of a certain business (such as a transaction business or a financial business, etc.).
  • the server may be a server for payment services, or a server for related services such as finance or instant messaging, or a server that needs to perform risk detection or risk prevention and control on business data.
  • the method may specifically include the following steps: In step S402, information of a classification group to which the terminal device belongs is acquired.
  • the classification group to which the terminal device belongs can be determined in a variety of different ways, for example, the classification group obtained by grouping the terminal devices by means of clustering or the like, or the classification group of each terminal device can be The user divides the classification groups manually to obtain the classification groups to which the terminal device belongs.
  • step S402 can be triggered in many different ways, for example, it can be triggered when the terminal device requests the server to execute the target service, or it can also be that the server can obtain the The information of the classification group, etc., can be specifically set according to the actual situation, which is not limited in the embodiment of this specification.
  • step S404 based on the information of the classification group to which the terminal device belongs, the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs is obtained, and the initial risk prevention and control model is sent to the terminal device , so that the terminal device performs model training on the initial risk prevention and control model based on the pre-stored training sample data, and obtains the risk prevention and control sub-model corresponding to the terminal device.
  • the training sample data includes at least information related to the user of the terminal device and the target business data.
  • step S406 receive the risk prevention and control sub-model sent by the terminal device, and perform model fusion on the risk prevention and control sub-model sent by the terminal device and the risk prevention and control sub-model provided by other terminal devices in the classification group to which the terminal device belongs processing to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs.
  • the risk prevention and control sub-model provided by other terminal devices is based on the pre-stored training sample data in other terminal devices to model the initial risk prevention and control model. Get trained.
  • step S408 the risk prevention and control model is sent to the terminal device, so that the terminal device performs risk prevention and control processing on the acquired data of the target business based on the risk prevention and control model.
  • the embodiment of this specification provides a risk prevention and control method.
  • the initial The risk prevention and control model When receiving the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs issued by the server, based on the pre-stored training sample data, the initial The risk prevention and control model performs model training to obtain the risk prevention and control sub-model corresponding to the terminal device, wherein the training sample data includes at least data related to the user of the terminal device and the target business, and then sends the risk prevention and control sub-model to The server performs model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the terminal device belongs to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the terminal device uses the risk prevention and control model provided by the server.
  • the risk prevention and control model corresponding to the classification group to which the terminal device belongs performs risk prevention and control processing on the acquired data of the target business.
  • the risk prevention and control model is determined through federated learning, thereby realizing shared learning based on the terminal and the server.
  • Personalized risk prevention and control, and the above method can not only protect private data, but also reduce the interaction between the terminal device and the server, reduce the computing pressure on the server, and improve the flexibility of the risk prevention and control model.
  • the embodiment of this specification provides a method for risk prevention and control.
  • the execution subject of the method may be a server, wherein the server may be a server for a certain business (such as a transaction business or a financial business, etc.)
  • the server may be a server for payment services, or a server for related services such as finance or instant messaging, or it may also be a server that needs to perform risk detection or risk prevention and control on business data.
  • the method may specifically include the following steps: In step S502, deliver the first model to be trained to the terminal device, and the parameter value of the model parameter in the first model is a randomly generated parameter value or a preset parameter value, so that The terminal device performs model training on the first model based on the pre-stored training sample data, obtains the trained first model, and sends the parameters and corresponding parameter values in the preset network layer in the trained first model to the server .
  • the first model may be constructed by a preset neural network model.
  • step S504 the parameters in the preset network layer and corresponding parameter values in the trained first model sent by the terminal device are received, and based on the parameters in the preset network layer in the trained first model and corresponding parameter values The parameter value of the terminal device is clustered by a preset clustering algorithm to obtain the classification group to which the terminal device belongs.
  • the clustering algorithm may include K-means clustering algorithm and the like.
  • the parameters and corresponding parameter values in the preset network layer in the trained first model may be encrypted parameters and corresponding encrypted parameter values obtained after the terminal device uses the preset second encryption algorithm for encryption processing, Based on this, after the server receives the parameters and corresponding parameter values in the preset network layer in the trained first model sent by the terminal device, it can use the decryption algorithm corresponding to the second encryption algorithm to encrypt the encrypted parameters and corresponding The encrypted parameter values are decrypted to obtain the parameters and corresponding parameter values in the preset network layer in the trained first model.
  • step S506 the information of the classification group to which the terminal device belongs is acquired.
  • step S508 based on the information of the classification group to which the terminal device belongs, the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs is obtained, and the initial risk prevention and control model is sent to the terminal device , so that the terminal device performs model training on the initial risk prevention and control model based on the pre-stored training sample data, and obtains the risk prevention and control sub-model corresponding to the terminal device.
  • the training sample data includes at least information related to the user of the terminal device and the target business data.
  • the initial risk prevention and control model can be constructed by a preset neural network model.
  • the initial risk prevention and control model may be different from the above-mentioned first model.
  • step S510 receive the risk prevention and control sub-model sent by the terminal device, and perform model fusion on the risk prevention and control sub-model sent by the terminal device and the risk prevention and control sub-model provided by other terminal devices in the classification group to which the terminal device belongs processing to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs.
  • the risk prevention and control sub-model provided by other terminal devices is based on the pre-stored training sample data in other terminal devices to model the initial risk prevention and control model. Get trained.
  • the risk prevention and control sub-model may be an encrypted risk prevention and control sub-model obtained by the terminal device using the preset first encryption algorithm for encryption processing. Based on this, after the server receives the risk prevention and control sub-model sent by the terminal device, , the encrypted risk prevention and control sub-model can be decrypted using the decryption algorithm corresponding to the first encryption algorithm to obtain the risk prevention and control sub-model.
  • step S512 the risk prevention and control model is sent to the terminal device, so that the terminal device performs risk prevention and control processing on the acquired data of the target business based on the risk prevention and control model.
  • step S514 the trained risk prevention and control model sent by the terminal device is received.
  • the trained risk prevention and control model is when the terminal device detects that the training sample data is updated, the risk prevention and control model is updated based on the updated training sample data. Obtained by model training.
  • step S5166 based on the trained risk prevention and control model, the risk prevention and control model corresponding to the classification group to which the terminal device belongs is updated to obtain an updated risk prevention and control model.
  • step S5128 the updated risk prevention and control model is sent to the terminal device, so that the terminal device performs risk prevention and control processing on the data of the target business based on the updated risk prevention and control model.
  • the embodiment of this specification provides a risk prevention and control method.
  • the initial The risk prevention and control model When receiving the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs issued by the server, based on the pre-stored training sample data, the initial The risk prevention and control model performs model training to obtain the risk prevention and control sub-model corresponding to the terminal device, wherein the training sample data includes at least data related to the user of the terminal device and the target business, and then sends the risk prevention and control sub-model to The server performs model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the terminal device belongs to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the terminal device uses the risk prevention and control model provided by the server.
  • the risk prevention and control model corresponding to the classification group to which the terminal device belongs performs risk prevention and control processing on the acquired data of the target business.
  • the risk prevention and control model is determined through federated learning, thereby realizing shared learning based on the terminal and the server.
  • Personalized risk prevention and control, and the above method can not only protect private data, but also reduce the interaction between the terminal device and the server, reduce the computing pressure on the server, and improve the flexibility of the risk prevention and control model.
  • a personalized risk prevention and control model is customized for each category of users, and the risk prevention and control model is trained through clustering algorithms combined with federated learning, which can realize your high-end
  • the data in the device does not need to be transmitted to the server, thereby protecting user privacy.
  • the embodiment of this specification also provides a risk prevention and control device, as shown in FIG. 6 .
  • the risk prevention and control device includes: an initial model receiving module 601, a sub-model training module 602, a sub-model sending module 603, and a risk prevention and control module 604, wherein: the initial model receiving module 601 receives the classification of the device issued by the server The initial risk prevention and control model of the target service to be trained corresponding to the group; the sub-model training module 602 performs model training on the initial risk prevention and control model based on the pre-stored training sample data, and obtains the risk prevention and control corresponding to the device A sub-model, the training sample data at least includes data related to the user of the device and the target service; the sub-model sending module 603 sends the risk prevention and control sub-model to the server, so that the The server performs model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the device belongs to obtain the risk prevention and control model corresponding to the classification group to which the device belongs; the risk prevention and control module 604 receives The server sends the risk prevention and control model corresponding to
  • the initial risk prevention and control model is constructed by a preset neural network model.
  • the sub-model sending module 603 encrypts the risk prevention and control sub-model using a preset first encryption algorithm to obtain an encrypted risk prevention and control sub-model, and sends the encrypted The risk prevention and control sub-model is sent to the server.
  • the device further includes: a first model receiving module, which receives the first model to be trained issued by the server, and the parameter values of the model parameters in the first model are randomly generated parameter values or preset Set parameter values; the first model training module performs model training on the first model based on the training sample data, obtains the first model after training, and obtains the preset network in the first model after training The parameters in the layer and the corresponding parameter values; the parameter sending module sends the parameters and the corresponding parameter values in the preset network layer in the first model after training to the server, so that the server is based on different The parameters and corresponding parameter values in the preset network layer in the trained first model provided by the terminal device are used to cluster different terminal devices to obtain classification groups to which different terminal devices belong.
  • a first model receiving module which receives the first model to be trained issued by the server, and the parameter values of the model parameters in the first model are randomly generated parameter values or preset Set parameter values
  • the first model training module performs model training on the first model based on the training sample data, obtains
  • the first model is constructed by a preset neural network model, and the first model is different from the initial risk prevention and control model.
  • the classification group to which the device belongs is that the server is based on the parameters and corresponding parameter values in the preset network layer in the trained first model provided by different terminal devices, through the preset aggregation
  • the class algorithm is used to cluster different terminal devices.
  • the clustering algorithm is a K-means clustering algorithm.
  • the parameter sending module uses the preset second encryption algorithm to encrypt the parameters in the preset network layer and the corresponding parameter values in the trained first model, and obtain the encrypted parameters and corresponding encrypted parameter values, and send the encrypted parameters and corresponding encrypted parameter values to the server.
  • the risk prevention and control module 604 includes: a data acquisition unit, which acquires the business data of the target business to be detected; a risk prevention and control unit, which inputs the business data into the risk prevention and control model , to detect whether the business data has a preset risk, and obtain a corresponding detection result; if the detection result indicates that the business data has a preset risk, cancel the execution of the target by the device Business business processing.
  • the device further includes: a first sample update detection module, if it is detected that the training sample data is updated, then obtain updated training sample data; a model update module, based on the updated training sample Perform model training on the risk prevention and control model based on the data to obtain a trained risk prevention and control model, and perform risk prevention and control processing on the data of the target business based on the trained risk prevention and control model.
  • the device further includes: a second sample update detection module, if it is detected that the training sample data is updated, then acquire the updated training sample data; update the training module, based on the updated
  • the training sample data is used to perform model training on the risk prevention and control model to obtain the trained risk prevention and control model, and send the trained risk prevention and control model to the server, so that the server can control the risk prevention and control model to which the device belongs.
  • the risk prevention and control model corresponding to the classification group is updated to obtain the updated risk prevention and control model; the update model receiving module receives the updated risk prevention and control model sent by the server, and based on the updated risk prevention and control model
  • the data of the target business is processed for risk prevention and control.
  • the embodiment of this specification provides a risk prevention and control device.
  • the initial The risk prevention and control model When receiving the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs issued by the server, the initial The risk prevention and control model performs model training to obtain the risk prevention and control sub-model corresponding to the terminal device, wherein the training sample data includes at least data related to the user of the terminal device and the target business, and then sends the risk prevention and control sub-model to The server performs model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the terminal device belongs to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the terminal device uses the risk prevention and control model provided by the server.
  • the risk prevention and control model corresponding to the classification group to which the terminal device belongs performs risk prevention and control processing on the acquired data of the target business.
  • the risk prevention and control model is determined through federated learning, thereby realizing shared learning based on the terminal and the server.
  • Personalized risk prevention and control, and the above method can not only protect private data, but also reduce the interaction between the terminal device and the server, reduce the computing pressure on the server, and improve the flexibility of the risk prevention and control model.
  • a personalized risk prevention and control model is customized for each category of users, and the risk prevention and control model is trained through clustering algorithms combined with federated learning, which can realize your high-end
  • the data in the device does not need to be transmitted to the server, thereby protecting user privacy.
  • the embodiment of this specification also provides a risk prevention and control device, as shown in FIG. 7 .
  • the risk prevention and control device includes: a group information acquisition module 701, an initial model delivery module 702, a sub-model receiving module 703, and a risk control model delivery module 704, wherein: the group information acquisition module 701 acquires the classification group to which the terminal device belongs Group information; the initial model delivery module 702, based on the information of the classification group to which the terminal device belongs, obtains the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs, and sends The initial risk prevention and control model is sent to the terminal device, so that the terminal device performs model training on the initial risk prevention and control model based on pre-stored training sample data, and obtains the risk prevention and control corresponding to the terminal device.
  • the sub-model, the training sample data at least includes data related to the user of the terminal device and the target service;
  • the sub-model receiving module 703 receives the risk prevention and control sub-model sent by the terminal device, and performing model fusion processing on the risk prevention and control sub-model sent by the terminal device and the risk prevention and control sub-model provided by other terminal devices in the classification group to which the terminal device belongs, to obtain the classification group to which the terminal device belongs.
  • the corresponding risk prevention and control model, the risk prevention and control sub-model provided by the other terminal device is obtained by the other terminal device performing model training on the initial risk prevention and control model based on the pre-stored training sample data in the other terminal device
  • the risk control model sending module 704 which sends the risk prevention and control model to the terminal device, so that the terminal device performs risk prevention and control processing on the acquired data of the target business based on the risk prevention and control model .
  • the device further includes: a first model sending module, which sends the first model to be trained to the terminal device, and the parameter values of the model parameters in the first model are randomly generated parameters value or a preset parameter value, so that the terminal device performs model training on the first model based on the training sample data to obtain the trained first model, and the The parameters in the preset network layer and the corresponding parameter values are sent to the device; the parameter receiving module receives the parameters in the preset network layer and the corresponding parameters in the trained first model sent by the terminal device value, and based on the parameters in the preset network layer and the corresponding parameter values in the first model after training, the terminal devices are clustered by a preset clustering algorithm to obtain the taxonomic group to which the terminal devices belong Group.
  • a first model sending module which sends the first model to be trained to the terminal device, and the parameter values of the model parameters in the first model are randomly generated parameters value or a preset parameter value, so that the terminal device performs model training on the first model based on the training sample data
  • the device further includes: an update model receiving module, which receives the trained risk prevention and control model sent by the terminal device, and the trained risk prevention and control model is the When the training sample data is updated, it is obtained by performing model training on the risk prevention and control model based on the updated training sample data; the update module, based on the trained risk prevention and control model, updates the terminal device The risk prevention and control model corresponding to the classification group to which it belongs is updated to obtain the updated risk prevention and control model; the update model delivery module sends the updated risk prevention and control model to the terminal device, so that the terminal device Risk prevention and control processing is performed on the data of the target business based on the updated risk prevention and control model.
  • an update model receiving module which receives the trained risk prevention and control model sent by the terminal device, and the trained risk prevention and control model is the When the training sample data is updated, it is obtained by performing model training on the risk prevention and control model based on the updated training sample data
  • the update module based on the trained risk prevention and control model, updates the terminal device The risk prevention and control model
  • the embodiment of this specification provides a risk prevention and control device.
  • the initial The risk prevention and control model When receiving the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs issued by the server, the initial The risk prevention and control model performs model training to obtain the risk prevention and control sub-model corresponding to the terminal device, wherein the training sample data includes at least data related to the user of the terminal device and the target business, and then sends the risk prevention and control sub-model to The server performs model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the terminal device belongs to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the terminal device uses the risk prevention and control model provided by the server.
  • the risk prevention and control model corresponding to the classification group to which the terminal device belongs performs risk prevention and control processing on the acquired data of the target business.
  • the risk prevention and control model is determined through federated learning, thereby realizing shared learning based on the terminal and the server.
  • Personalized risk prevention and control, and the above method can not only protect private data, but also reduce the interaction between the terminal device and the server, reduce the computing pressure on the server, and improve the flexibility of the risk prevention and control model.
  • a personalized risk prevention and control model is customized for each category of users, and the risk prevention and control model is trained through clustering algorithms combined with federated learning, which can realize your high-end
  • the data in the device does not need to be transmitted to the server, thereby protecting user privacy.
  • the embodiment of this specification also provides a risk prevention and control device, as shown in FIG. 8 .
  • the risk prevention and control device may be the server or terminal device provided in the above embodiments.
  • the risk prevention and control equipment may have relatively large differences due to different configurations or performances, and may include one or more processors 801 and memory 802, and one or more storage applications or data may be stored in the memory 802.
  • the storage 802 may be a short-term storage or a persistent storage.
  • the application program stored in the memory 802 may include one or more modules (not shown in the figure), and each module may include a series of computer-executable instructions for risk prevention and control equipment.
  • the processor 801 may be configured to communicate with the memory 802, and execute a series of computer-executable instructions in the memory 802 on the risk prevention and control device.
  • the risk prevention and control equipment may also include one or more power sources 803 , one or more wired or wireless network interfaces 804 , one or more input and output interfaces 805 , and one or more keyboards 806 .
  • the risk prevention and control device includes a memory, and one or more programs, wherein one or more programs are stored in the memory, and one or more programs may include one or more modules, and each A module may include a series of computer-executable instructions in the risk prevention and control equipment, and is configured to be executed by one or more processors.
  • the one or more programs include computer-executable instructions for performing the following: receiving the server The initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the device belongs; the initial risk prevention and control model is trained based on the pre-stored training sample data, and the risk prevention and control model corresponding to the device is obtained.
  • a control sub-model the training sample data at least includes data related to the user of the device and the target business;
  • the risk control sub-model is sent to the server, so that the server can control the device performing model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the device belongs, to obtain the risk prevention and control model corresponding to the classification group to which the device belongs; receiving the classification information sent by the server to which the device belongs A risk prevention and control model corresponding to the group, and based on the risk prevention and control model, perform risk prevention and control processing on the acquired data of the target business.
  • the initial risk prevention and control model is constructed by a preset neural network model.
  • the sending the risk prevention and control sub-model to the server includes: using a preset first encryption algorithm to encrypt the risk prevention and control sub-model to obtain the encrypted risk prevention and control sub-model. control sub-model, and send the encrypted risk control sub-model to the server.
  • it also includes: receiving the first model to be trained issued by the server, the parameter value of the model parameter in the first model is a randomly generated parameter value or a preset parameter value; based on the training The sample data performs model training on the first model, obtains the first model after training, and obtains the parameters and corresponding parameter values in the preset network layer in the first model after training; The parameters in the preset network layer in the first model and the corresponding parameter values are sent to the server, so that the server is based on the parameters in the preset network layer in the trained first model provided by different terminal devices and corresponding parameter values to cluster different terminal devices to obtain classification groups to which different terminal devices belong.
  • the first model is constructed by a preset neural network model, and the first model is different from the initial risk prevention and control model.
  • the classification group to which the device belongs is that the server is based on the parameters and corresponding parameter values in the preset network layer in the trained first model provided by different terminal devices, through the preset aggregation
  • the class algorithm is used to cluster different terminal devices.
  • the clustering algorithm is a K-means clustering algorithm.
  • the sending the parameters and corresponding parameter values in the preset network layer in the trained first model to the server includes: using the preset second encryption algorithm to encrypt the The parameters in the preset network layer and the corresponding parameter values in the first model after training are encrypted to obtain the encrypted parameters and the corresponding encrypted parameter values, and the encrypted parameters and the corresponding encrypted parameters are encrypted.
  • the parameter value is sent to the server.
  • the risk prevention and control process is performed on the acquired data of the target business based on the risk prevention and control model, including: acquiring the business data of the target business to be detected; inputting the business data into the In the above risk prevention and control model, it is used to detect whether the business data has a preset risk and obtain the corresponding detection result; if the detection result indicates that the business data has a preset risk, cancel the device to execute the Business processing of the target business.
  • it also includes: if it is detected that the training sample data has been updated, acquiring the updated training sample data; performing model training on the risk prevention and control model based on the updated training sample data, and obtaining the trained risk prevention and control model, and perform risk prevention and control processing on the data of the target business based on the trained risk prevention and control model.
  • it also includes: if it is detected that the training sample data is updated, acquiring the updated training sample data; performing model training on the risk prevention and control model based on the updated training sample data , obtain the trained risk prevention and control model, and send the trained risk prevention and control model to the server, so that the server can update the risk prevention and control model corresponding to the classification group to which the device belongs, and obtain the updated An updated risk prevention and control model; receiving the updated risk prevention and control model sent by the server, and performing risk prevention and control processing on the data of the target business based on the updated risk prevention and control model.
  • the risk prevention and control device includes a memory, and one or more programs, wherein one or more programs are stored in the memory, and one or more programs may include one or more modules, And each module may include a series of computer-executable instructions in the risk prevention and control equipment, and configured to be executed by one or more processors.
  • the one or more programs include computer-executable instructions for performing the following: obtaining terminal information of the classification group to which the device belongs; based on the information of the classification group to which the terminal device belongs, obtain an initial risk prevention and control model of a target service to be trained corresponding to the classification group to which the terminal device belongs, and convert the The initial risk prevention and control model is delivered to the terminal device, so that the terminal device performs model training on the initial risk prevention and control model based on pre-stored training sample data, and obtains a risk prevention and control sub-model corresponding to the terminal device , the training sample data at least includes data related to the user of the terminal device and the target service; receiving the risk prevention and control sub-model sent by the terminal device, and performing the The risk prevention and control sub-model and the risk prevention and control sub-model provided by other terminal devices in the classification group to which the terminal device belongs perform model fusion processing to obtain a risk prevention and control model corresponding to the classification group to which the terminal device belongs.
  • the risk prevention and control sub-models provided by the other terminal devices are obtained by the other terminal devices performing model training on the initial risk prevention and control model based on the pre-stored training sample data in the other terminal devices; sending to the terminal device, so that the terminal device performs risk prevention and control processing on the acquired data of the target service based on the risk prevention and control model.
  • it also includes: sending the first model to be trained to the terminal device, the parameter value of the model parameter in the first model is a randomly generated parameter value or a preset parameter value, so that The terminal device performs model training on the first model based on the training sample data, obtains the trained first model, and combines the parameters in the preset network layer and the corresponding parameters in the trained first model Sending parameter values to the device; receiving parameters and corresponding parameter values in the preset network layer in the trained first model sent by the terminal device, and based on the preset in the trained first model The parameters in the network layer and the corresponding parameter values are used to cluster the terminal devices through a preset clustering algorithm to obtain the classification group to which the terminal devices belong.
  • it also includes: receiving the trained risk prevention and control model sent by the terminal device, and the trained risk prevention and control model is that the terminal device detects that the training sample data is updated , the risk prevention and control model is obtained by performing model training on the basis of the updated training sample data; based on the trained risk prevention and control model, the risk prevention and control model corresponding to the classification group to which the terminal device belongs performing an update to obtain an updated risk prevention and control model; sending the updated risk prevention and control model to the terminal device, so that the terminal device performs risk management on the data of the target business based on the updated risk prevention and control model Prevention and control treatment.
  • the embodiment of this specification provides a risk prevention and control device.
  • the initial The risk prevention and control model When receiving the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs issued by the server, the initial The risk prevention and control model performs model training to obtain the risk prevention and control sub-model corresponding to the terminal device, wherein the training sample data includes at least data related to the user of the terminal device and the target business, and then sends the risk prevention and control sub-model to The server performs model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the terminal device belongs to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the terminal device uses the risk prevention and control model provided by the server.
  • the risk prevention and control model corresponding to the classification group to which the terminal device belongs performs risk prevention and control processing on the acquired data of the target business.
  • the risk prevention and control model is determined through federated learning, thereby realizing shared learning based on the terminal and the server.
  • Personalized risk prevention and control, and the above method can not only protect private data, but also reduce the interaction between the terminal device and the server, reduce the computing pressure on the server, and improve the flexibility of the risk prevention and control model.
  • a personalized risk prevention and control model is customized for each category of users, and the risk prevention and control model is trained through clustering algorithms combined with federated learning, which can realize your high-end
  • the data in the device does not need to be transmitted to the server, thereby protecting user privacy.
  • one or more embodiments of this specification also provide a storage medium for storing computer-executable instruction information.
  • the storage The medium can be a USB flash drive, a CD, a hard disk, etc.
  • the following process can be realized: receiving the training information corresponding to the classification group to which the terminal equipment belongs issued by the server; An initial risk prevention and control model of the target business; model training is performed on the initial risk prevention and control model based on pre-stored training sample data to obtain a risk prevention and control sub-model corresponding to the terminal device, and the training sample data includes at least the same The data related to the user of the terminal device and the target service; sending the risk prevention and control sub-model to the server, so that the server can provide different terminal devices in the classification group to which the terminal device belongs performing model fusion processing on the risk prevention and control sub-model to obtain a risk prevention and control model corresponding to the classification group to which the terminal device belongs; receiving the risk prevention and control model corresponding to the classification group to which the terminal device belongs from the server, and performing risk prevention and control processing on the acquired data of the target business based on the risk prevention and control model.
  • the initial risk prevention and control model is constructed by a preset neural network model.
  • the sending the risk prevention and control sub-model to the server includes: using a preset first encryption algorithm to encrypt the risk prevention and control sub-model to obtain the encrypted risk prevention and control sub-model. control sub-model, and send the encrypted risk control sub-model to the server.
  • it also includes: receiving the first model to be trained issued by the server, the parameter value of the model parameter in the first model is a randomly generated parameter value or a preset parameter value; based on the training The sample data performs model training on the first model, obtains the first model after training, and obtains the parameters and corresponding parameter values in the preset network layer in the first model after training; The parameters in the preset network layer in the first model and the corresponding parameter values are sent to the server, so that the server is based on the parameters in the preset network layer in the trained first model provided by different terminal devices and corresponding parameter values to cluster different terminal devices to obtain classification groups to which different terminal devices belong.
  • the first model is constructed by a preset neural network model, and the first model is different from the initial risk prevention and control model.
  • the classification group to which the terminal device belongs is based on the parameters and corresponding parameter values in the preset network layer in the trained first model provided by the server based on different terminal devices, through the preset
  • the clustering algorithm is obtained by clustering different terminal devices.
  • the clustering algorithm is a K-means clustering algorithm.
  • the sending the parameters and corresponding parameter values in the preset network layer in the trained first model to the server includes: using the preset second encryption algorithm to encrypt the The parameters in the preset network layer and the corresponding parameter values in the first model after training are encrypted to obtain the encrypted parameters and the corresponding encrypted parameter values, and the encrypted parameters and the corresponding encrypted parameters are encrypted.
  • the parameter value is sent to the server.
  • the risk prevention and control processing of the acquired data of the target business based on the risk prevention and control model includes: acquiring the business data of the target business to be detected; inputting the business data Into the risk prevention and control model to detect whether the business data has a preset risk, and obtain a corresponding detection result; if the detection result indicates that the business data has a preset risk, cancel the terminal device Execute the service processing of the target service.
  • it also includes: if it is detected that the training sample data has been updated, acquiring the updated training sample data; performing model training on the risk prevention and control model based on the updated training sample data, and obtaining the trained risk prevention and control model, and perform risk prevention and control processing on the data of the target business based on the trained risk prevention and control model.
  • it also includes: if it is detected that the training sample data is updated, acquiring the updated training sample data; performing model training on the risk prevention and control model based on the updated training sample data , obtain the trained risk prevention and control model, and send the trained risk prevention and control model to the server, so that the server can update the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and obtain An updated risk prevention and control model; receiving the updated risk prevention and control model sent by the server, and performing risk prevention and control processing on the data of the target business based on the updated risk prevention and control model.
  • the storage medium may be a USB flash drive, an optical disk, a hard disk, etc.
  • the computer-executable instruction information stored in the storage medium can realize the following process when executed by the processor: obtain the category to which the terminal device belongs Group information; based on the information of the classification group to which the terminal device belongs, the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs is obtained, and the initial risk prevention and control model is The model is sent to the terminal device, so that the terminal device performs model training on the initial risk prevention and control model based on the pre-stored training sample data, and obtains the risk prevention and control sub-model corresponding to the terminal device, and the training
  • the sample data includes at least data related to the user of the terminal device and the target service; receiving the risk prevention and control sub-model sent by the terminal device, and analyzing the risk prevention and control sub-model sent by the terminal device Perform model fusion processing on the model and the risk prevention and control sub-models provided by other terminal devices in the classification
  • it also includes: sending the first model to be trained to the terminal device, the parameter value of the model parameter in the first model is a randomly generated parameter value or a preset parameter value, so that The terminal device performs model training on the first model based on the training sample data, obtains the trained first model, and combines the parameters in the preset network layer and the corresponding parameters in the trained first model Sending parameter values to the server; receiving parameters and corresponding parameter values in the preset network layer in the trained first model sent by the terminal device, and based on the preset in the trained first model The parameters in the network layer and the corresponding parameter values are used to cluster the terminal devices through a preset clustering algorithm to obtain the classification group to which the terminal devices belong.
  • it also includes: receiving the trained risk prevention and control model sent by the terminal device, and the trained risk prevention and control model is that the terminal device detects that the training sample data is updated , the risk prevention and control model is obtained by performing model training on the basis of the updated training sample data; based on the trained risk prevention and control model, the risk prevention and control model corresponding to the classification group to which the terminal device belongs performing an update to obtain an updated risk prevention and control model; sending the updated risk prevention and control model to the terminal device, so that the terminal device performs risk management on the data of the target business based on the updated risk prevention and control model Prevention and control treatment.
  • the embodiment of this specification provides a storage medium.
  • the initial risk prevention and control model of the target service to be trained corresponding to the classification group to which the terminal device belongs issued by the server is based on the pre-stored training sample data.
  • the server performs model fusion processing on the risk prevention and control sub-models provided by different terminal devices in the classification group to which the terminal device belongs to obtain the risk prevention and control model corresponding to the classification group to which the terminal device belongs, and the terminal device uses the terminal device provided by the server
  • the risk prevention and control model corresponding to the classification group to which it belongs performs risk prevention and control processing on the acquired data of the target business.
  • the risk prevention and control model is determined through federated learning, thereby realizing the individuality based on shared learning between the terminal and the server.
  • the above method can not only protect private data, but also reduce the interaction between the terminal device and the server, reduce the computing pressure on the server, and improve the flexibility of the risk prevention and control model.
  • a personalized risk prevention and control model is customized for each category of users, and the risk prevention and control model is trained through clustering algorithms combined with federated learning, which can realize your high-end
  • the data in the device does not need to be transmitted to the server, thereby protecting user privacy.
  • a Programmable Logic Device such as a Field Programmable Gate Array (FPGA)
  • FPGA Field Programmable Gate Array
  • HDL Hardware Description Language
  • the controller may be implemented in any suitable way, for example the controller may take the form of a microprocessor or processor and a computer readable medium storing computer readable program code (such as software or firmware) executable by the (micro)processor , logic gates, switches, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable logic controllers and embedded microcontrollers, examples of controllers include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320, the memory controller can also be implemented as part of the control logic of the memory.
  • controller in addition to realizing the controller in a purely computer-readable program code mode, it is entirely possible to make the controller use logic gates, switches, application-specific integrated circuits, programmable logic controllers, and embedded The same function can be realized in the form of a microcontroller or the like. Therefore, such a controller can be regarded as a hardware component, and the devices included in it for realizing various functions can also be regarded as structures within the hardware component. Or even, means for realizing various functions can be regarded as a structure within both a software module realizing a method and a hardware component.
  • a typical implementing device is a computer.
  • 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 Combinations of any of these devices.
  • one or more embodiments of the present description may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects.
  • one or more embodiments of the present description may employ a computer program embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein. The form of the product.
  • Embodiments of the present specification are described with reference to flowcharts and/or block diagrams of methods, apparatus (systems), and computer program products according to the embodiments of the present specification. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable fraudulent serial device to produce a machine such that processing by a computer or other programmable fraudulent serial device The instructions executed by the device generate means for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.
  • These computer program instructions may also be stored in a computer readable memory capable of directing a computer or other programmable fraud case serial parallel device to operate in a specific manner such that the instructions stored in the computer readable memory produce an article of manufacture comprising instruction means , the instruction means realizes the functions specified in one or more procedures of the flow chart and/or one or more blocks of the block diagram.
  • These computer program instructions may also be loaded onto a computer or other programmable fraud case serial-parallel device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, whereby the computer or other programmable device
  • the instructions executed above provide steps for implementing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.
  • a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
  • processors CPUs
  • input/output interfaces network interfaces
  • memory volatile and non-volatile memory
  • Memory may include non-permanent storage in computer-readable media, in the form of random access memory (RAM) and/or nonvolatile memory such as read-only memory (ROM) or flash RAM. Memory is an example of computer readable media.
  • RAM random access memory
  • ROM read-only memory
  • flash RAM flash random access memory
  • Computer-readable media including both permanent and non-permanent, removable and non-removable media, can be implemented by any method or technology for storage of information.
  • Information may be computer readable instructions, data structures, modules of a program, 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 Disc Read-Only Memory (CD-ROM), Digital Versatile Disc (DVD) or other optical storage, Magnetic tape cartridge, tape magnetic disk storage or other magnetic storage device or any other non-transmission medium that can be used to store information that can be accessed by a computing device.
  • computer-readable media excludes transitory computer-readable media, such as modulated data signals and carrier waves.
  • one or more embodiments of the present description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects.
  • one or more embodiments of the present description may employ a computer program embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein. The form of the product.
  • program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types.
  • program modules may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network.
  • program modules may be located in both local and remote computer storage media including storage devices.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Accounting & Taxation (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Computer Security & Cryptography (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Software Systems (AREA)
  • General Business, Economics & Management (AREA)
  • Finance (AREA)
  • Strategic Management (AREA)
  • Mathematical Physics (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Computational Linguistics (AREA)
  • Bioethics (AREA)
  • Molecular Biology (AREA)
  • Probability & Statistics with Applications (AREA)
  • Computer Hardware Design (AREA)
  • Databases & Information Systems (AREA)
  • Medical Informatics (AREA)
  • Evolutionary Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Telephonic Communication Services (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

一种风险防控的方法、装置及设备。该方法应用于终端设备,包括:接收服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型(S102);基于预先存储的训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,训练样本数据中至少包括与终端设备的用户和目标业务相关的数据(S104);将风险防控子模型发送给服务器,以使服务器对终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型(S106);接收服务器发送的终端设备所属的分类群组对应的风险防控模型,并基于风险防控模型对目标业务的数据进行风险防控处理(S108)。

Description

风险防控的方法、装置及设备 技术领域
本说明书涉及计算机技术领域,尤其涉及风险防控的方法、装置及设备。
背景技术
当前风险防控系统的架构通常是由服务端进行风险决策,终端设备采集相应业务的数据,并将其上传给服务器。通过服务器强大的运算能力,对上述数据进行风控策略和风险防控模型的分析,最终输出风控决策。
上述服务端进行风险决策的方式中,需要终端设备将数据上传至服务器,这样就可能会给用户的隐私数据带来安全隐患,而且,如果终端设备进行集中请求,服务器的资源消耗大,此外,由于不同用户使用同一个风险防控模型,不同用户的习惯和每个用户的喜好存在差别,因此,该风险防控模型很难满足不同用户的需求,风险防控模型的灵活性差。基于此,需要提供一种基于终端与服务器(或端云)共享学习的个性化风险防控系统,该风险防控系统不仅可以保护隐私数据,还可以减少终端设备和服务器之间的交互,减轻服务器计算压力,而且还可以提高风险防控模型的灵活性。
发明内容
本说明书实施例的目的是提供一种基于终端与服务器(或端云)共享学习的个性化风险防控系统,该风险防控系统不仅可以保护隐私数据,还可以减少终端设备和服务器之间的交互,减轻服务器计算压力,而且还可以提高风险防控模型的灵活性。
为了实现上述技术方案,本说明书实施例提供的一种风险防控方法,应用于终端设备,所述方法包括:接收服务器下发的所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型。基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据。将所述风险防控子模型发送给所述服务器,以使所述服务器对所述终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型。接收所述服务器发送的所述终端设备所属的分类群组对应的风险防控模型,并基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例提供的一种风险防控方法,应用于服务器,包括:获取终端设备所属的分类群组的信息。基于所述终端设备所属的分类群组的信息,获取所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将所述初始风险防控模型下发到所述终端设备,以使所述终端设备基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据。接收所述终端设备发送的所述风险防控子模型,并对所述终端设备发送的所述风险防控子模型和所述终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型,所述其它终端设备提供的风险防控子模型是所述其它终端设备基于所述其它终端设备中预先存储的训练样本数据对所述初始风险防控模型进行模型训练得到。将所述风险防控模型发送给所述终端设备,以使所述终端设备基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例提供的一种风险防控装置,包括:初始模型接收模块,接收服务器下发的所述装置所属的分类群组对应的待训练的目标业务的初始风险防控模型。子模型训练模块,基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述装置对应的风险防控子模型,所述训练样本数据中至少包括与所述装置的用户和所述目标业务相关的数据。子模型发送模块,将所述风险防控子模型发送给所述服务器,以使所述服务器对所述装置所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到所述装置所属的分类群组对应的风险防控模型。风险防控模块, 接收所述服务器发送的所述装置所属的分类群组对应的风险防控模型,并基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例提供的一种风险防控装置,包括:群组信息获取模块,获取终端设备所属的分类群组的信息。初始模型下发模块,基于所述终端设备所属的分类群组的信息,获取所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将所述初始风险防控模型下发到所述终端设备,以使所述终端设备基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据。子模型接收模块,接收所述终端设备发送的所述风险防控子模型,并对所述终端设备发送的所述风险防控子模型和所述终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型,所述其它终端设备提供的风险防控子模型是所述其它终端设备基于所述其它终端设备中预先存储的训练样本数据对所述初始风险防控模型进行模型训练得到。风控模型下发模块,将所述风险防控模型发送给所述终端设备,以使所述终端设备基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例提供的一种风险防控设备,包括处理器以及被安排成存储计算机可执行指令的存储器,所述可执行指令在被执行时使所述处理器:接收服务器下发的所述设备所属的分类群组对应的待训练的目标业务的初始风险防控模型。基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述设备对应的风险防控子模型,所述训练样本数据中至少包括与所述设备的用户和所述目标业务相关的数据。将所述风险防控子模型发送给所述服务器,以使所述服务器对所述设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到所述设备所属的分类群组对应的风险防控模型。接收所述服务器发送的所述设备所属的分类群组对应的风险防控模型,并基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例提供的一种风险防控设备,包括处理器以及被安排成存储计算机可执行指令的存储器,所述可执行指令在被执行时使所述处理器:获取终端设备所属的分类群组的信息。基于所述终端设备所属的分类群组的信息,获取所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将所述初始风险防控模型下发到所述终端设备,以使所述终端设备基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据。接收所述终端设备发送的所述风险防控子模型,并对所述终端设备发送的所述风险防控子模型和所述终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型,所述其它终端设备提供的风险防控子模型是所述其它终端设备基于所述其它终端设备中预先存储的训练样本数据对所述初始风险防控模型进行模型训练得到。将所述风险防控模型发送给所述终端设备,以使所述终端设备基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例还提供了一种存储介质,其用于存储计算机可执行指令,所述可执行指令在被执行时实现以下流程:接收服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型。基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据。将所述风险防控子模型发送给所述服务器,以使所述服务器对所述终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型。接收所述服务器发送的所述终端设备所属的分类群组对应的风险防控模型,并基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例还提供了一种存储介质,其用于存储计算机可执行指令,所述可执行指令在被执行时实现以下流程:获取终端设备所属的分类群组的信息。基于所述终端设备所属的分类群组的信息,获取所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将所述初始风险防控模型下发到所述终端设备,以使所述终端设备基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据。接收所述终端设备发送的所述风险防控子模型,并对所述终端设备发送的所述风险防控子模型和所述终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型,所述其它终端设备提供的风险防控子模型是所述其它终端设备基于所述其它终端设备中预先存储的训练样本数据对所述初始风险防控模型进行模型训练得到。将所述风险防控模型发送给所述终端设备,以使所述终端设备基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
附图说明
图1A为本说明书一种风险防控方法实施例;
图1B为本说明书一种风险防控的处理过程示意图;
图2为本说明书一种风险防控系统的结构示意图;
图3为本说明书另一种风险防控的处理过程示意图;
图4A为本说明书另一种风险防控方法实施例;
图4B为本说明书又一种风险防控的处理过程示意图;
图5为本说明书又一种风险防控的处理过程示意图;
图6为本说明书一种风险防控装置实施例;
图7为本说明书另一种风险防控装置实施例;
图8为本说明书一种风险防控设备实施例。
具体实施方式
为了使本技术领域的人员更好地理解本说明书中的技术方案,下面将结合本说明书实施例中的附图,对本说明书实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例仅仅是本说明书一部分实施例,而不是全部的实施例。基于本说明书中的实施例,本领域普通技术人员在没有作出创造性劳动前提下所获得的所有其他实施例,都应当属于本说明书保护的范围。
实施例一
如图1A和图1B所示,本说明书实施例提供一种风险防控方法,该方法的执行主体可以为终端设备,其中,该终端设备可以如手机、平板电脑、个人计算机等。该方法具体可以包括以下步骤:在步骤S102中,接收服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型。
其中,服务器可以是某项业务(如进行交易的业务或金融业务等)的服务器,具体如,该服务器可以是支付业务的服务器,也可以是与金融或即时通讯等相关业务的服务器等,或者,也可以是对某项业务进行风险防控的服务器,具体可以根据实际情况设定,本说明书实施例对此不做限定。分类群组可以是对终端设备通过聚类等方式进行群组划分后得到的分类群组,该分类群组可以通过多种不同的方式设定,例如,可以通过设定的不同年龄段进行分类群组的划分,或者,也可以通过用户所处区域的不同进行分类群组的划分,或者,也可以根据用户注册目标业务的时长进行分类群组的划分等,具体可以根据实际情况设定,本说明书实施例对此不做限定。目标业务可以是任意的可能存在风险或需要进行风险防控的业务,例如支付业务或转账业务等,具体可以根据实际情况设定,本说明书实施例对此不做限定。初始风险防控模型可以是还未经过训练的用于对某业务进行风险防控的模型,该初始风险防控模型可以仅具有模型架构,其中的模型参 数还不准确,具体如可以是模型参数还未知或预先设定的数值或随机值等,具体可以根据实际情况设定,本说明书实施例对此不做限定。
在实施中,当前风险防控系统的架构通常是由服务端进行风险决策,终端设备采集相应业务的数据,并将其上传给服务器。通过服务器强大的运算能力,对上述数据进行风控策略和风险防控模型的分析,最终输出风控决策。上述服务端进行风险决策的方式中,需要终端设备将数据上传至服务器,这样就可能会给用户的隐私数据带来安全隐患,而且,如果终端设备进行集中请求,服务器的资源消耗大,此外,由于不同用户使用同一个风险防控模型,不同用户的习惯和每个用户的喜好存在差别,因此,该风险防控模型很难满足不同用户的需求,风险防控模型的灵活性差。基于此,需要提供一种基于终端与服务器(或端云)共享学习的个性化风险防控系统,该风险防控系统不仅可以保护隐私数据,还可以减少终端设备和服务器之间的交互,减轻服务器计算压力,而且还可以提高风险防控模型的灵活性。本说明书实施例提供一种可实现的技术方案,具体可以包括以下内容:如图2所示,考虑到不同的用户的习惯和喜好,以及对风险防控意识的强弱,可以对不同的用户进行聚类、分类或人工划分群组等方式,将具有某相同属性的用户划分到同一个分类群组,进而可以确定不同用户所属的分类群组,相应的,也可以以此确定用户使用的终端设备所属的分类群组。服务器可以针对不同的分类群组分别构建相应的风险防控模型的模型架构,其中,不同分类群组对应的风险防控模型的模型架构可相同,也可不同,该模型架构中的模型参数的参数值可以是未知的,也可是随机设定的,还可是根据专家经验而设定的,具体可以根据实际情况设定,本说明书实施例对此不做限定。服务器为不同的分类群组设定的上述风险防控模型可以作为待训练的目标业务的初始风险防控模型。服务器可以获取不同分类群组和不同分类群组中包含的终端设备的信息或用户的信息,然后,可基于终端设备的信息或用户的信息,将每个终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型发送给终端设备,终端设备可以接收服务器发送的该终端设备所属的分类群组对应的初始风险防控模型。
例如,针对目标业务,可以预先设定不同的年龄段,如0-25岁、26-60岁和大于60岁等三个年龄段,可以使用相应的算法,分别为上述三个不同的年龄段构建相应的风险防控模型的模型架构,其中,各个风险防控模型的模型参数的参数值可以随机设定,从而可以得到三个不同年龄段对应的目标业务的初始风险防控模型。然后,可以获取预先记录的或在服务器中注册目标业务的每个用户的信息或用户使用的终端设备的信息,基于用户的年龄信息,可以确定该用户的终端设备所属的分类群组,具体如,某用户的年龄为37岁,属于26-60岁的年龄段,可以获取26-60岁的年龄段对应的目标业务的初始风险防控模型,并可以将该初始风险防控模型发送给该用户的终端设备。
上述步骤S102的处理可以通过多种不同的方式触发,例如,可以是在终端设备向服务器请求执行目标业务时触发,或者,也可以是每当到达预设周期时,服务器可以向终端设备下发该终端设备所属的分类群组对应的初始风险防控模型,或者,也可以是终端设备周期性的向服务器获取终端设备所属的分类群组对应的初始风险防控模型等,具体可以根据实际情况设定,本说明书实施例对此不做限定。
在步骤S104中,基于预先存储的训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,该训练样本数据中至少包括与终端设备的用户和目标业务相关的数据。
其中,训练样本数据可以包括终端设备的用户在执行目标业务的过程中产生的相关数据,即与终端设备的用户和目标业务相关的数据,该相关数据可以是指定时间段内产生的数据,也可以是从用户注册目标业务开始到当前时刻产生的数据等,具体可以根据实际情况设定,本说明书实施例对此不做限定。在实际应用中,训练样本数据也可以包括终端设备通过指定的数据获取方式从其它终端设备、服务器或数据库等获取的与目标业务相关的数据,或者,可以包括终端设备通过指定的数据获取方式从其它终端设备、 服务器或数据库等获取的与终端设备的用户和目标业务相关的数据等。
在实施中,由于初始风险防控模型中的模型参数是不准确的,因此,每个终端设备中可以维护一定量的训练样本数据,在通过上述方式得到初始风险防控模型后,可以通过终端设备中的训练样本数据对初始风险防控模型进行模型训练,即可以从训练样本数据中选择一个样本数据,将该样本数据输入到初始风险防控模型中,得到关于模型参数的表达式,之后,可以从训练样本数据中再选择一个样本数据,将该样本数据输入到初始风险防控模型中,再次得到关于模型参数的表达式,通过上述方式可以得到多个关于模型参数的表达式,可以将多个关于模型参数的表达式组成方程组,求解方程组,可以得到各个模型参数的参数值。可以通过上述方式使用训练样本数据中剩余的样本数据对初始风险防控模型中的模型参数进行更新,直到最终的结果收敛为止,从而得到较优的模型参数,进而得到训练后的风险防控模型,并可以将训练后的风险防控模型作为终端设备对应的风险防控子模型。
在步骤S106中,将风险防控子模型发送给服务器,以使服务器对终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型。
在实施中,终端设备可以将风险防控子模型发送给服务器,服务器可以接收该终端设备对应的风险防控子模型,同样的,该终端设备所属的分类群组中的多个不同的终端设备也可以通过上述方式训练属于每个终端设备的风险防控子模型,每个终端设备可以将其训练的风险防控子模型提供给服务器,服务器可以对该终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,这样通过联邦学习的方式,服务器可以得到终端设备所属的分类群组对应的风险防控模型。
在步骤S108中,接收服务器发送的终端设备所属的分类群组对应的风险防控模型,并基于该风险防控模型对获取的目标业务的数据进行风险防控处理。
在实施中,终端设备可以接收到服务器发送的终端设备所属的分类群组对应的风险防控模型,当终端设备获取到目标业务的数据时,可以将目标业务的数据输入到该风险防控模型中,得到该目标业务的数据是否存在风险的结果,如果该结果指示该目标业务的数据存在风险,则可以拒绝终端设备执行目标业务,如果该结果指示该目标业务的数据不存在风险,则可以允许终端设备执行目标业务。
本说明书实施例提供一种风险防控方法,在接收到服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型时,基于预先存储的训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,其中,该训练样本数据中至少包括与终端设备的用户和目标业务相关的数据,然后,将风险防控子模型发送给服务器,服务器对终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型,终端设备使用服务器提供的该终端设备所属的分类群组对应的风险防控模型对获取的所述目标业务的数据进行风险防控处理,这样,通过联邦学习的方式确定风险防控模型,从而实现了基于终端与服务器共享学习的个性化风险防控,而且上述方式不仅可以保护隐私数据,还可以减少终端设备和服务器之间的交互,减轻服务器计算压力,并可以提高风险防控模型的灵活性。
实施例二
如图3所示,本说明书实施例提供一种风险防控方法,该方法的执行主体可以为终端设备,其中,该终端设备可以如手机、平板电脑、个人计算机等。该方法具体可以包括以下步骤:在步骤S302中,接收服务器下发的待训练的第一模型,第一模型中的模型参数的参数值为随机产生的参数值或预设的参数值。
其中,第一模型可以由预设的神经网络模型构建,该神经网络模型可以包括多种,例如可以包括卷积神经网络模型、循环神经网络模型、生成式对抗网络模型等,具体可 以根据实际情况设定,本说明书实施例对此不做限定。第一模型中的模型参数的参数值可以为预设的参数值,例如,第一模型中的模型参数的参数值可以是根据专家经验而设定的参数值,或者,第一模型中的模型参数的参数值可以是根据相关的历史数据而设定的参数值等,具体可以根据实际情况设定,本说明书实施例对此不做限定。
在实施中,为了提高用户数据的安全性,终端设备中的针对目标业务和用户的相关数据可以不进行输出,而是在终端设备中保留相应的训练样本数据,并可以在终端设备中训练相应模型,以得到相应的模型参数。在实际应用中,模型训练的过程中模型的某一个或多个网络层中保留的训练样本数据的信息较多,因此,服务器可以预先设定相应的算法,并可以通过该算法构建第一模型的模型架构,并可以设定其中的模型参数的参数值,从而服务器可以得到待训练的第一模型。为了能够准确对终端设备或终端设备的用户进行分类群组的划分,可以基于终端设备中的训练样本数据实现,但是由于训练样本数据无法由终端设备输出,因此,可以通过模型训练的方式获取终端设备中的训练样本数据的相关信息,具体地,服务器可以将构建的待训练的第一模型发送给终端设备,终端设备可以接收服务器下发的待训练的第一模型。
在步骤S304中,基于预先存储的训练样本数据对第一模型进行模型训练,得到训练后的第一模型,并获取训练后的第一模型中的预设网络层中的参数和相应的参数值。
其中,预设网络层可以包括一个网络层,也可以包括多个网络层,具体如,第一模型为卷积神经网络模型,预设网络层可以是其中的多个卷积层,也可以是卷积层、池化层和全连接层中的一种或多种等,具体可以根据实际情况设定,在实际应用中,训练后的第一模型中的预设网络层可以是输出层的前一网络层(如全连接层等)。
在实施中,模型训练的过程中模型的某一个或多个网络层中保留的训练样本数据的信息较多,在综合考虑通信量和有效性后,可选择仅向服务器上传模型训练的过程中模型的某一个或多个网络层的信息。具体地,可获取终端设备预先存储的训练样本数据,可通过该训练样本数据对第一模型进行模型训练,得到训练后的第一模型。然后,可获取训练后的第一模型中的预设网络层中的参数和相应的参数值,例如,可获取训练后的第一模型中输出层的前一网络层(具体如全连接层等)的参数和相应的参数值等。
在步骤S306中,将训练后的第一模型中的预设网络层中的参数和相应的参数值发送给服务器,以使服务器基于不同终端设备提供的训练后的第一模型中的预设网络层中的参数和相应的参数值,对不同的终端设备进行聚类,得到不同终端设备所属的分类群组。
在实施中,服务器接收到终端设备发送的训练后的第一模型中的预设网络层中的参数和相应的参数值后,由于训练后的第一模型中的预设网络层中的参数和相应的参数值中保留有终端设备中的训练样本数据的信息,因此,可以使用训练后的第一模型中的预设网络层中的参数和相应的参数值代替终端设备中的训练样本数据,对终端设备或终端设备的用户进行分类群组的划分,此外,其它终端设备也可以通过上述方式向服务器提供训练后的第一模型中的预设网络层中的参数和相应的参数值,从而服务器可以得到多个不同终端设备提供的训练后的第一模型中的预设网络层中的参数和相应的参数值。然后,可以通过预设的聚类算法对获取的训练后的第一模型中的预设网络层中的参数和相应的参数值进行聚类处理,得到不同终端设备所属的分类群组,从而可以得到上述终端设备所属的分类群组。
在实际应用中,上述步骤S306的处理可以多种多样,以下提供一种可选的处理方式,具体可以包括以下内容:使用预设的第二加密算法对训练后的第一模型中的预设网络层中的参数和相应的参数值进行加密处理,得到加密后的参数和相应的加密后的参数值,并将加密后的参数和相应的加密后的参数值发送给服务器。
其中,第二加密算法可以包括多种,例如同态加密算法、部分同态加密算法或全同态加密算法等,具体可以根据实际情况设定,本说明书实施例对此不做限定。
在实施中,为了保证传输数据的安全性,可以预先设定加密算法(即第二加密算法), 可以使用第二加密算法对训练后的第一模型中的预设网络层中的参数和相应的参数值进行加密处理,得到加密后的参数和相应的加密后的参数值,然后,将加密后的参数和相应的加密后的参数值发送给服务器,从而保证数据传输过程中数据的安全性。
在步骤S308中,接收服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型。
其中,初始风险防控模型可以由预设的神经网络模型构建,该神经网络模型可以包括多种,例如可以包括卷积神经网络模型、循环神经网络模型、生成式对抗网络模型等。其中,上述第一模型与初始风险防控模型可以不同。终端设备所属的分类群组是服务器基于不同终端设备提供的训练后的第一模型中的预设网络层中的参数和相应的参数值,通过预设的聚类算法对不同的终端设备进行聚类得到,具体可以参见上述相关内容,其中的聚类算法可以为K-means聚类算法等,在实际应用中,聚类算法也可以包括其它多种不同的聚类算法,具体可以根据实际情况选定,本说明书实施例对此不做限定。
在步骤S310中,基于上述训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,训练样本数据中至少包括与终端设备的用户和目标业务相关的数据。
在步骤S312中,使用预设的第一加密算法对风险防控子模型进行加密处理,得到加密后的风险防控子模型,并将加密后的风险防控子模型发送给服务器,以使服务器对终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型。
其中,第一加密算法可以包括多种,例如同态加密算法、部分同态加密算法或全同态加密算法等,第一加密算法可以与上述第二加密算法相同,也可以与上述第二加密算法不同,具体可以根据实际情况设定,本说明书实施例对此不做限定。
在步骤S314中,接收服务器发送的终端设备所属的分类群组对应的风险防控模型。
通过上述处理得到终端设备所属的分类群组对应的风险防控模型,终端设备可以利用该风险防控模型对目标业务的业务数据进行风险检测和风险防控,具体可以包括以下步骤S316~步骤S320的处理。
在步骤S316中,获取待检测的目标业务的业务数据。
在步骤S318中,将上述业务数据输入到风险防控模型中,以对业务数据是否存在预设风险进行检测,得到相应的检测结果。
在步骤S320中,如果检测结果指示业务数据存在预设风险,则取消终端设备执行目标业务的业务处理。
此外,还可以对终端设备所属的分类群组对应的风险防控模型进行更新,具体可以包括以下步骤S322~步骤S326的处理。
在步骤S322中,如果检测到训练样本数据被更新,则获取更新后的训练样本数据。
在实施中,检测终端设备中存储的训练样本数据是否被更新的方式可以包括多种,例如,终端设备存储的训练样本数据可以包括数据标识,终端设备可以周期性或不定期检测训练样本数据中的数据标识是否增加,从而确定训练样本数据是否被更新,或者,可以检测训练样本数据中增加的数据量是否超过预定数量阈值,从而确定训练样本数据是否被更新等,具体可以根据实际情况设定,本说明书实施例对此不做限定。
在步骤S324中,基于更新后的训练样本数据对风险防控模型进行模型训练,得到训练后的风险防控模型,并将训练后的风险防控模型发送给服务器,以使服务器对终端设备所属的分类群组对应的风险防控模型进行更新,得到更新后的风险防控模型。
在实施中,终端设备可以将训练后的风险防控模型发送给服务器,服务器可以接收该训练后的风险防控模型,同样的,该终端设备所属的分类群组中的多个不同的终端设备也可以通过上述方式训练属于每个终端设备的风险防控模型,每个终端设备可以将其训练的风险防控模型提供给服务器,服务器可以对该终端设备所属的分类群组中的不同 终端设备提供的风险防控模型进行模型融合处理,以更新终端设备所属的分类群组对应的风险防控模型,这样通过联邦学习的方式,服务器可以不断更新终端设备所属的分类群组对应的风险防控模型,此外,服务器也可以只通过该终端设备提供的风险防控模型对该终端设备所属的分类群组对应的风险防控模型进行更新,具体可以根据实际情况设定,本说明书实施例对此不做限定。
在步骤S326中,接收服务器发送的更新后的风险防控模型,并基于更新后的风险防控模型对目标业务的数据进行风险防控处理。
另外,上述对终端设备所属的分类群组对应的风险防控模型进行更新的处理除了可以通过上述步骤S322~步骤S326实现外,还可以通过下述方式实现,具体可以包括以下步骤A2~步骤A4的处理。
在步骤A2中,如果检测到训练样本数据被更新,则获取更新后的训练样本数据。
在步骤A4中,基于更新后训练样本数据对风险防控模型进行模型训练,得到训练后的风险防控模型,并基于训练后的风险防控模型对目标业务的数据进行风险防控处理。
本说明书实施例提供一种风险防控方法,在接收到服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型时,基于预先存储的训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,其中,该训练样本数据中至少包括与终端设备的用户和目标业务相关的数据,然后,将风险防控子模型发送给服务器,服务器对终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型,终端设备使用服务器提供的该终端设备所属的分类群组对应的风险防控模型对获取的所述目标业务的数据进行风险防控处理,这样,通过联邦学习的方式确定风险防控模型,从而实现了基于终端与服务器共享学习的个性化风险防控,而且上述方式不仅可以保护隐私数据,还可以减少终端设备和服务器之间的交互,减轻服务器计算压力,并可以提高风险防控模型的灵活性。
此外,基于用户的相关数据进行聚类,为每个类别群体的用户定制个性化的风险防控模型,而且,通过聚类算法结合联邦学习的方式对风险防控模型进行训练,可以实现你高端设备中的数据不需传输给服务器,从而保护了用户隐私。
实施例三
如图4A和图4B所示,本说明书实施例提供一种风险防控方法,该方法的执行主体可为服务器,例如为某项业务(如进行交易的业务或金融业务等)的服务器。具体地,该服务器可是支付业务的服务器,也可是与金融或即时通讯等相关业务的服务器等,或者,也可是需要对业务数据进行风险检测或风险防控的服务器等。该方法具体可以包括以下步骤:在步骤S402中,获取终端设备所属的分类群组的信息。
在实施中,终端设备所属的分类群组可以通过多种不同的方式确定,例如可以对终端设备通过聚类等方式进行群组划分后得到的分类群组,或者,可以对每个终端设备的用户通过人工的方式进行分类群组的划分,从而得到终端设备所属的分类群组。
上述步骤S402的处理可以通过多种不同的方式触发,例如,可以是在终端设备向服务器请求执行目标业务时触发,或者,也可以是每当到达预设周期时,服务器可以获取终端设备所属的分类群组的信息等,具体可以根据实际情况设定,本说明书实施例对此不做限定。
在步骤S404中,基于终端设备所属的分类群组的信息,获取终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将初始风险防控模型下发到终端设备,以使终端设备基于预先存储的训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,该训练样本数据中至少包括与终端设备的用户和目标业务相关的数据。
在步骤S406中,接收终端设备发送的风险防控子模型,并对终端设备发送的风险防 控子模型和终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型,其它终端设备提供的风险防控子模型是其它终端设备基于其它终端设备中预先存储的训练样本数据对初始风险防控模型进行模型训练得到。
在步骤S408中,将风险防控模型发送给终端设备,以使终端设备基于风险防控模型对获取的目标业务的数据进行风险防控处理。
本说明书实施例提供一种风险防控方法,在接收到服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型时,基于预先存储的训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,其中,该训练样本数据中至少包括与终端设备的用户和目标业务相关的数据,然后,将风险防控子模型发送给服务器,服务器对终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型,终端设备使用服务器提供的该终端设备所属的分类群组对应的风险防控模型对获取的所述目标业务的数据进行风险防控处理,这样,通过联邦学习的方式确定风险防控模型,从而实现了基于终端与服务器共享学习的个性化风险防控,而且上述方式不仅可以保护隐私数据,还可以减少终端设备和服务器之间的交互,减轻服务器计算压力,并可以提高风险防控模型的灵活性。
实施例四
如图5所示,本说明书实施例提供一种风险防控方法,该方法的执行主体可以为服务器,其中,该服务器可以是为某项业务(如进行交易的业务或金融业务等)的服务器,具体如,该服务器可以是支付业务的服务器,也可以是与金融或即时通讯等相关业务的服务器等,或者,也可以是需要对业务数据进行风险检测或风险防控的服务器等。该方法具体可以包括以下步骤:在步骤S502中,向终端设备下发待训练的第一模型,第一模型中的模型参数的参数值为随机产生的参数值或预设的参数值,以使终端设备基于预先存储的训练样本数据对第一模型进行模型训练,得到训练后的第一模型,并将训练后的第一模型中的预设网络层中的参数和相应的参数值发送给服务器。
其中,第一模型可以由预设的神经网络模型构建。
在步骤S504中,接收终端设备发送的训练后的第一模型中的预设网络层中的参数和相应的参数值,并基于训练后的第一模型中的预设网络层中的参数和相应的参数值,通过预设的聚类算法对终端设备进行聚类,得到终端设备所属的分类群组。
其中,聚类算法可以包括K-means聚类算法等。
训练后的第一模型中的预设网络层中的参数和相应的参数值可是终端设备使用预设的第二加密算法进行加密处理后得到的加密后的参数和相应的加密后的参数值,基于此,服务器接收到终端设备发送的训练后的第一模型中的预设网络层中的参数和相应的参数值后,可使用与第二加密算法对应的解密算法对加密后的参数和相应的加密后的参数值进行解密处理,得到训练后的第一模型中的预设网络层中的参数和相应的参数值。
在步骤S506中,获取终端设备所属的分类群组的信息。
在步骤S508中,基于终端设备所属的分类群组的信息,获取终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将初始风险防控模型下发到终端设备,以使终端设备基于预先存储的训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,该训练样本数据中至少包括与终端设备的用户和目标业务相关的数据。
其中,初始风险防控模型可以由预设的神经网络模型构建。初始风险防控模型可以与上述第一模型不同。
在步骤S510中,接收终端设备发送的风险防控子模型,并对终端设备发送的风险防控子模型和终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模 型融合处理,得到终端设备所属的分类群组对应的风险防控模型,其它终端设备提供的风险防控子模型是其它终端设备基于其它终端设备中预先存储的训练样本数据对初始风险防控模型进行模型训练得到。
其中,风险防控子模型可以是终端设备使用预设的第一加密算法进行加密处理后得到的加密后的风险防控子模型,基于此,服务器接收到终端设备发送的风险防控子模型后,可以使用与第一加密算法对应的解密算法对加密后的风险防控子模型进行解密处理,得到风险防控子模型。
在步骤S512中,将风险防控模型发送给终端设备,以使终端设备基于风险防控模型对获取的目标业务的数据进行风险防控处理。
在步骤S514中,接收终端设备发送的训练后的风险防控模型,训练后的风险防控模型是终端设备在检测到训练样本数据被更新时,基于更新后的训练样本数据对风险防控模型进行模型训练得到。
在步骤S516中,基于训练后的风险防控模型,对终端设备所属的分类群组对应的风险防控模型进行更新,得到更新后的风险防控模型。
在步骤S518中,将更新后的风险防控模型发送给终端设备,以使终端设备基于更新后的风险防控模型对目标业务的数据进行风险防控处理。
本说明书实施例提供一种风险防控方法,在接收到服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型时,基于预先存储的训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,其中,该训练样本数据中至少包括与终端设备的用户和目标业务相关的数据,然后,将风险防控子模型发送给服务器,服务器对终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型,终端设备使用服务器提供的该终端设备所属的分类群组对应的风险防控模型对获取的所述目标业务的数据进行风险防控处理,这样,通过联邦学习的方式确定风险防控模型,从而实现了基于终端与服务器共享学习的个性化风险防控,而且上述方式不仅可以保护隐私数据,还可以减少终端设备和服务器之间的交互,减轻服务器计算压力,并可以提高风险防控模型的灵活性。
此外,基于用户的相关数据进行聚类,为每个类别群体的用户定制个性化的风险防控模型,而且,通过聚类算法结合联邦学习的方式对风险防控模型进行训练,可以实现你高端设备中的数据不需传输给服务器,从而保护了用户隐私。
实施例五
以上为本说明书实施例提供的风险防控方法,基于同样的思路,本说明书实施例还提供一种风险防控装置,如图6所示。
该风险防控装置包括:初始模型接收模块601、子模型训练模块602、子模型发送模块603和风险防控模块604,其中:初始模型接收模块601,接收服务器下发的所述装置所属的分类群组对应的待训练的目标业务的初始风险防控模型;子模型训练模块602,基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述装置对应的风险防控子模型,所述训练样本数据中至少包括与所述装置的用户和所述目标业务相关的数据;子模型发送模块603,将所述风险防控子模型发送给所述服务器,以使所述服务器对所述装置所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到所述装置所属的分类群组对应的风险防控模型;风险防控模块604,接收所述服务器发送的所述装置所属的分类群组对应的风险防控模型,并基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例中,所述初始风险防控模型由预设的神经网络模型构建。
本说明书实施例中,所述子模型发送模块603,使用预设的第一加密算法对所述风险防控子模型进行加密处理,得到加密后的风险防控子模型,并将所述加密后的风险防 控子模型发送给所述服务器。
本说明书实施例中,所述装置还包括:第一模型接收模块,接收服务器下发的待训练的第一模型,所述第一模型中的模型参数的参数值为随机产生的参数值或预设的参数值;第一模型训练模块,基于所述训练样本数据对所述第一模型进行模型训练,得到训练后的第一模型,并获取所述训练后的第一模型中的预设网络层中的参数和相应的参数值;参数发送模块,将所述训练后的第一模型中的预设网络层中的参数和相应的参数值发送给所述服务器,以使所述服务器基于不同终端设备提供的训练后的第一模型中的预设网络层中的参数和相应的参数值,对不同的终端设备进行聚类,得到不同终端设备所属的分类群组。
本说明书实施例中,所述第一模型由预设的神经网络模型构建,所述第一模型与所述初始风险防控模型不同。
本说明书实施例中,所述装置所属的分类群组是所述服务器基于不同终端设备提供的训练后的第一模型中的预设网络层中的参数和相应的参数值,通过预设的聚类算法对不同的终端设备进行聚类得到。
本说明书实施例中,所述聚类算法为K-means聚类算法。
本说明书实施例中,所述参数发送模块,使用预设的第二加密算法对所述训练后的第一模型中的预设网络层中的参数和相应的参数值进行加密处理,得到加密后的参数和相应的加密后的参数值,并将加密后的参数和相应的加密后的参数值发送给所述服务器。
本说明书实施例中,所述风险防控模块604,包括:数据获取单元,获取待检测的所述目标业务的业务数据;风险防控单元,将所述业务数据输入到所述风险防控模型中,以对所述业务数据是否存在预设风险进行检测,得到相应的检测结果;业务处理单元,如果所述检测结果指示所述业务数据存在预设风险,则取消所述装置执行所述目标业务的业务处理。
本说明书实施例中,所述装置还包括:第一样本更新检测模块,如果检测到所述训练样本数据被更新,则获取更新后的训练样本数据;模型更新模块,基于更新后的训练样本数据对所述风险防控模型进行模型训练,得到训练后的风险防控模型,并基于训练后的风险防控模型对所述目标业务的数据进行风险防控处理。
本说明书实施例中,所述装置还包括:第二样本更新检测模块,如果检测到所述训练样本数据被更新,则获取更新后的所述训练样本数据;更新训练模块,基于更新后的所述训练样本数据对所述风险防控模型进行模型训练,得到训练后的风险防控模型,并将训练后的风险防控模型发送给所述服务器,以使所述服务器对所述装置所属的分类群组对应的风险防控模型进行更新,得到更新后的风险防控模型;更新模型接收模块,接收所述服务器发送的更新后的风险防控模型,并基于更新后的风险防控模型对所述目标业务的数据进行风险防控处理。
本说明书实施例提供一种风险防控装置,在接收到服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型时,基于预先存储的训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,其中,该训练样本数据中至少包括与终端设备的用户和目标业务相关的数据,然后,将风险防控子模型发送给服务器,服务器对终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型,终端设备使用服务器提供的该终端设备所属的分类群组对应的风险防控模型对获取的所述目标业务的数据进行风险防控处理,这样,通过联邦学习的方式确定风险防控模型,从而实现了基于终端与服务器共享学习的个性化风险防控,而且上述方式不仅可以保护隐私数据,还可以减少终端设备和服务器之间的交互,减轻服务器计算压力,并可以提高风险防控模型的灵活性。
此外,基于用户的相关数据进行聚类,为每个类别群体的用户定制个性化的风险防 控模型,而且,通过聚类算法结合联邦学习的方式对风险防控模型进行训练,可以实现你高端设备中的数据不需传输给服务器,从而保护了用户隐私。
实施例六
基于同样的思路,本说明书实施例还提供一种风险防控装置,如图7所示。
该风险防控装置包括:群组信息获取模块701、初始模型下发模块702、子模型接收模块703和风控模型下发模块704,其中:群组信息获取模块701,获取终端设备所属的分类群组的信息;初始模型下发模块702,基于所述终端设备所属的分类群组的信息,获取所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将所述初始风险防控模型下发到所述终端设备,以使所述终端设备基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据;子模型接收模块703,接收所述终端设备发送的所述风险防控子模型,并对所述终端设备发送的所述风险防控子模型和所述终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型,所述其它终端设备提供的风险防控子模型是所述其它终端设备基于所述其它终端设备中预先存储的训练样本数据对所述初始风险防控模型进行模型训练得到;风控模型下发模块704,将所述风险防控模型发送给所述终端设备,以使所述终端设备基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例中,所述装置还包括:第一模型下发模块,向所述终端设备下发待训练的第一模型,所述第一模型中的模型参数的参数值为随机产生的参数值或预设的参数值,以使所述终端设备基于所述训练样本数据对所述第一模型进行模型训练,得到训练后的第一模型,并将所述训练后的第一模型中的预设网络层中的参数和相应的参数值发送给所述装置;参数接收模块,接收所述终端设备发送的所述训练后的第一模型中的预设网络层中的参数和相应的参数值,并基于训练后的第一模型中的预设网络层中的参数和相应的参数值,通过预设的聚类算法对所述终端设备进行聚类,得到所述终端设备所属的分类群组。
本说明书实施例中,所述装置还包括:更新模型接收模块,接收所述终端设备发送的训练后的所述风险防控模型,训练后的所述风险防控模型是所述终端设备在检测到所述训练样本数据被更新时,基于更新后的所述训练样本数据对所述风险防控模型进行模型训练得到;更新模块,基于训练后的所述风险防控模型,对所述终端设备所属的分类群组对应的风险防控模型进行更新,得到更新后的风险防控模型;更新模型下发模块,将更新后的风险防控模型发送给所述终端设备,以使所述终端设备基于更新后的风险防控模型对所述目标业务的数据进行风险防控处理。
本说明书实施例提供一种风险防控装置,在接收到服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型时,基于预先存储的训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,其中,该训练样本数据中至少包括与终端设备的用户和目标业务相关的数据,然后,将风险防控子模型发送给服务器,服务器对终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型,终端设备使用服务器提供的该终端设备所属的分类群组对应的风险防控模型对获取的所述目标业务的数据进行风险防控处理,这样,通过联邦学习的方式确定风险防控模型,从而实现了基于终端与服务器共享学习的个性化风险防控,而且上述方式不仅可以保护隐私数据,还可以减少终端设备和服务器之间的交互,减轻服务器计算压力,并可以提高风险防控模型的灵活性。
此外,基于用户的相关数据进行聚类,为每个类别群体的用户定制个性化的风险防控模型,而且,通过聚类算法结合联邦学习的方式对风险防控模型进行训练,可以实现 你高端设备中的数据不需传输给服务器,从而保护了用户隐私。
实施例七
以上为本说明书实施例提供的风险防控装置,基于同样的思路,本说明书实施例还提供一种风险防控设备,如图8所示。
所述风险防控设备可以为上述实施例提供的服务器或终端设备等。
风险防控设备可因配置或性能不同而产生比较大的差异,可以包括一个或一个以上的处理器801和存储器802,存储器802中可以存储有一个或一个以上存储应用程序或数据。其中,存储器802可以是短暂存储或持久存储。存储在存储器802的应用程序可以包括一个或一个以上模块(图示未示出),每个模块可以包括对风险防控设备中的一系列计算机可执行指令。更进一步地,处理器801可以设置为与存储器802通信,在风险防控设备上执行存储器802中的一系列计算机可执行指令。风险防控设备还可以包括一个或一个以上电源803,一个或一个以上有线或无线网络接口804,一个或一个以上输入输出接口805,一个或一个以上键盘806。
具体在本实施例中,风险防控设备包括有存储器,以及一个或一个以上的程序,其中一个或者一个以上程序存储于存储器中,且一个或者一个以上程序可以包括一个或一个以上模块,且每个模块可以包括对风险防控设备中的一系列计算机可执行指令,且经配置以由一个或者一个以上处理器执行该一个或者一个以上程序包含用于进行以下计算机可执行指令:接收服务器下发的所述设备所属的分类群组对应的待训练的目标业务的初始风险防控模型;基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述设备对应的风险防控子模型,所述训练样本数据中至少包括与所述设备的用户和所述目标业务相关的数据;将所述风险防控子模型发送给所述服务器,以使所述服务器对所述设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到所述设备所属的分类群组对应的风险防控模型;接收所述服务器发送的所述设备所属的分类群组对应的风险防控模型,并基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例中,所述初始风险防控模型由预设的神经网络模型构建。
本说明书实施例中,所述将所述风险防控子模型发送给所述服务器,包括:使用预设的第一加密算法对所述风险防控子模型进行加密处理,得到加密后的风险防控子模型,并将所述加密后的风险防控子模型发送给所述服务器。
本说明书实施例中,还包括:接收服务器下发的待训练的第一模型,所述第一模型中的模型参数的参数值为随机产生的参数值或预设的参数值;基于所述训练样本数据对所述第一模型进行模型训练,得到训练后的第一模型,并获取所述训练后的第一模型中的预设网络层中的参数和相应的参数值;将所述训练后的第一模型中的预设网络层中的参数和相应的参数值发送给所述服务器,以使所述服务器基于不同终端设备提供的训练后的第一模型中的预设网络层中的参数和相应的参数值,对不同的终端设备进行聚类,得到不同终端设备所属的分类群组。
本说明书实施例中,所述第一模型由预设的神经网络模型构建,所述第一模型与所述初始风险防控模型不同。
本说明书实施例中,所述设备所属的分类群组是所述服务器基于不同终端设备提供的训练后的第一模型中的预设网络层中的参数和相应的参数值,通过预设的聚类算法对不同的终端设备进行聚类得到。
本说明书实施例中,所述聚类算法为K-means聚类算法。
本说明书实施例中,所述将所述训练后的第一模型中的预设网络层中的参数和相应的参数值发送给所述服务器,包括:使用预设的第二加密算法对所述训练后的第一模型中的预设网络层中的参数和相应的参数值进行加密处理,得到加密后的参数和相应的加密后的参数值,并将加密后的参数和相应的加密后的参数值发送给所述服务器。
本说明书实施例中,基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理,包括:获取待检测的所述目标业务的业务数据;将所述业务数据输入到所述风险防控模型中,以对所述业务数据是否存在预设风险进行检测,得到相应的检测结果;如果所述检测结果指示所述业务数据存在预设风险,则取消所述设备执行所述目标业务的业务处理。
本说明书实施例中,还包括:如果检测到所述训练样本数据被更新,则获取更新后的训练样本数据;基于更新后的训练样本数据对所述风险防控模型进行模型训练,得到训练后的风险防控模型,并基于训练后的风险防控模型对所述目标业务的数据进行风险防控处理。
本说明书实施例中,还包括:如果检测到所述训练样本数据被更新,则获取更新后的所述训练样本数据;基于更新后的所述训练样本数据对所述风险防控模型进行模型训练,得到训练后的风险防控模型,并将训练后的风险防控模型发送给所述服务器,以使所述服务器对所述设备所属的分类群组对应的风险防控模型进行更新,得到更新后的风险防控模型;接收所述服务器发送的更新后的风险防控模型,并基于更新后的风险防控模型对所述目标业务的数据进行风险防控处理。
此外,具体在本实施例中,风险防控设备包括有存储器,以及一个或一个以上的程序,其中一个或者一个以上程序存储于存储器中,且一个或者一个以上程序可以包括一个或一个以上模块,且每个模块可以包括对风险防控设备中的一系列计算机可执行指令,且经配置以由一个或者一个以上处理器执行该一个或者一个以上程序包含用于进行以下计算机可执行指令:获取终端设备所属的分类群组的信息;基于所述终端设备所属的分类群组的信息,获取所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将所述初始风险防控模型下发到所述终端设备,以使所述终端设备基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据;接收所述终端设备发送的所述风险防控子模型,并对所述终端设备发送的所述风险防控子模型和所述终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型,所述其它终端设备提供的风险防控子模型是所述其它终端设备基于所述其它终端设备中预先存储的训练样本数据对所述初始风险防控模型进行模型训练得到;将所述风险防控模型发送给所述终端设备,以使所述终端设备基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例中,还包括:向所述终端设备下发待训练的第一模型,所述第一模型中的模型参数的参数值为随机产生的参数值或预设的参数值,以使所述终端设备基于所述训练样本数据对所述第一模型进行模型训练,得到训练后的第一模型,并将所述训练后的第一模型中的预设网络层中的参数和相应的参数值发送给所述设备;接收所述终端设备发送的所述训练后的第一模型中的预设网络层中的参数和相应的参数值,并基于训练后的第一模型中的预设网络层中的参数和相应的参数值,通过预设的聚类算法对所述终端设备进行聚类,得到所述终端设备所属的分类群组。
本说明书实施例中,还包括:接收所述终端设备发送的训练后的所述风险防控模型,训练后的所述风险防控模型是所述终端设备在检测到所述训练样本数据被更新时,基于更新后的所述训练样本数据对所述风险防控模型进行模型训练得到;基于训练后的所述风险防控模型,对所述终端设备所属的分类群组对应的风险防控模型进行更新,得到更新后的风险防控模型;将更新后的风险防控模型发送给所述终端设备,以使所述终端设备基于更新后的风险防控模型对所述目标业务的数据进行风险防控处理。
本说明书实施例提供一种风险防控设备,在接收到服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型时,基于预先存储的训练样本数据 对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,其中,该训练样本数据中至少包括与终端设备的用户和目标业务相关的数据,然后,将风险防控子模型发送给服务器,服务器对终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型,终端设备使用服务器提供的该终端设备所属的分类群组对应的风险防控模型对获取的所述目标业务的数据进行风险防控处理,这样,通过联邦学习的方式确定风险防控模型,从而实现了基于终端与服务器共享学习的个性化风险防控,而且上述方式不仅可以保护隐私数据,还可以减少终端设备和服务器之间的交互,减轻服务器计算压力,并可以提高风险防控模型的灵活性。
此外,基于用户的相关数据进行聚类,为每个类别群体的用户定制个性化的风险防控模型,而且,通过聚类算法结合联邦学习的方式对风险防控模型进行训练,可以实现你高端设备中的数据不需传输给服务器,从而保护了用户隐私。
实施例八
进一步地,基于上述图1A和图5所示的方法,本说明书一个或多个实施例还提供了一种存储介质,用于存储计算机可执行指令信息,一种具体的实施例中,该存储介质可以为U盘、光盘、硬盘等,该存储介质存储的计算机可执行指令信息在被处理器执行时,能实现以下流程:接收服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型;基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据;将所述风险防控子模型发送给所述服务器,以使所述服务器对所述终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型;接收所述服务器发送的所述终端设备所属的分类群组对应的风险防控模型,并基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例中,所述初始风险防控模型由预设的神经网络模型构建。
本说明书实施例中,所述将所述风险防控子模型发送给所述服务器,包括:使用预设的第一加密算法对所述风险防控子模型进行加密处理,得到加密后的风险防控子模型,并将所述加密后的风险防控子模型发送给所述服务器。
本说明书实施例中,还包括:接收服务器下发的待训练的第一模型,所述第一模型中的模型参数的参数值为随机产生的参数值或预设的参数值;基于所述训练样本数据对所述第一模型进行模型训练,得到训练后的第一模型,并获取所述训练后的第一模型中的预设网络层中的参数和相应的参数值;将所述训练后的第一模型中的预设网络层中的参数和相应的参数值发送给所述服务器,以使所述服务器基于不同终端设备提供的训练后的第一模型中的预设网络层中的参数和相应的参数值,对不同的终端设备进行聚类,得到不同终端设备所属的分类群组。
本说明书实施例中,所述第一模型由预设的神经网络模型构建,所述第一模型与所述初始风险防控模型不同。
本说明书实施例中,所述终端设备所属的分类群组是所述服务器基于不同终端设备提供的训练后的第一模型中的预设网络层中的参数和相应的参数值,通过预设的聚类算法对不同的终端设备进行聚类得到。
本说明书实施例中,所述聚类算法为K-means聚类算法。
本说明书实施例中,所述将所述训练后的第一模型中的预设网络层中的参数和相应的参数值发送给所述服务器,包括:使用预设的第二加密算法对所述训练后的第一模型中的预设网络层中的参数和相应的参数值进行加密处理,得到加密后的参数和相应的加密后的参数值,并将加密后的参数和相应的加密后的参数值发送给所述服务器。
本说明书实施例中,所述基于所述风险防控模型对获取的所述目标业务的数据进行 风险防控处理,包括:获取待检测的所述目标业务的业务数据;将所述业务数据输入到所述风险防控模型中,以对所述业务数据是否存在预设风险进行检测,得到相应的检测结果;如果所述检测结果指示所述业务数据存在预设风险,则取消所述终端设备执行所述目标业务的业务处理。
本说明书实施例中,还包括:如果检测到所述训练样本数据被更新,则获取更新后的训练样本数据;基于更新后的训练样本数据对所述风险防控模型进行模型训练,得到训练后的风险防控模型,并基于训练后的风险防控模型对所述目标业务的数据进行风险防控处理。
本说明书实施例中,还包括:如果检测到所述训练样本数据被更新,则获取更新后的所述训练样本数据;基于更新后的所述训练样本数据对所述风险防控模型进行模型训练,得到训练后的风险防控模型,并将训练后的风险防控模型发送给所述服务器,以使所述服务器对所述终端设备所属的分类群组对应的风险防控模型进行更新,得到更新后的风险防控模型;接收所述服务器发送的更新后的风险防控模型,并基于更新后的风险防控模型对所述目标业务的数据进行风险防控处理。
另一种具体的实施例中,该存储介质可以为U盘、光盘、硬盘等,该存储介质存储的计算机可执行指令信息在被处理器执行时,能实现以下流程:获取终端设备所属的分类群组的信息;基于所述终端设备所属的分类群组的信息,获取所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将所述初始风险防控模型下发到所述终端设备,以使所述终端设备基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据;接收所述终端设备发送的所述风险防控子模型,并对所述终端设备发送的所述风险防控子模型和所述终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型,所述其它终端设备提供的风险防控子模型是所述其它终端设备基于所述其它终端设备中预先存储的训练样本数据对所述初始风险防控模型进行模型训练得到;将所述风险防控模型发送给所述终端设备,以使所述终端设备基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
本说明书实施例中,还包括:向所述终端设备下发待训练的第一模型,所述第一模型中的模型参数的参数值为随机产生的参数值或预设的参数值,以使所述终端设备基于所述训练样本数据对所述第一模型进行模型训练,得到训练后的第一模型,并将所述训练后的第一模型中的预设网络层中的参数和相应的参数值发送给所述服务器;接收所述终端设备发送的所述训练后的第一模型中的预设网络层中的参数和相应的参数值,并基于训练后的第一模型中的预设网络层中的参数和相应的参数值,通过预设的聚类算法对所述终端设备进行聚类,得到所述终端设备所属的分类群组。
本说明书实施例中,还包括:接收所述终端设备发送的训练后的所述风险防控模型,训练后的所述风险防控模型是所述终端设备在检测到所述训练样本数据被更新时,基于更新后的所述训练样本数据对所述风险防控模型进行模型训练得到;基于训练后的所述风险防控模型,对所述终端设备所属的分类群组对应的风险防控模型进行更新,得到更新后的风险防控模型;将更新后的风险防控模型发送给所述终端设备,以使所述终端设备基于更新后的风险防控模型对所述目标业务的数据进行风险防控处理。
本说明书实施例提供一种存储介质,在接收到服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型时,基于预先存储的训练样本数据对初始风险防控模型进行模型训练,得到终端设备对应的风险防控子模型,其中,该训练样本数据中至少包括与终端设备的用户和目标业务相关的数据,然后,将风险防控子模型发送给服务器,服务器对终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到终端设备所属的分类群组对应的风险防控模型,终端设备 使用服务器提供的该终端设备所属的分类群组对应的风险防控模型对获取的所述目标业务的数据进行风险防控处理,这样,通过联邦学习的方式确定风险防控模型,从而实现了基于终端与服务器共享学习的个性化风险防控,而且上述方式不仅可以保护隐私数据,还可以减少终端设备和服务器之间的交互,减轻服务器计算压力,并可以提高风险防控模型的灵活性。
此外,基于用户的相关数据进行聚类,为每个类别群体的用户定制个性化的风险防控模型,而且,通过聚类算法结合联邦学习的方式对风险防控模型进行训练,可以实现你高端设备中的数据不需传输给服务器,从而保护了用户隐私。
上述对本说明书特定实施例进行了描述。其它实施例在所附权利要求书的范围内。在一些情况下,在权利要求书中记载的动作或步骤可以按照不同于实施例中的顺序来执行并且仍然可以实现期望的结果。另外,在附图中描绘的过程不一定要求示出的特定顺序或者连续顺序才能实现期望的结果。在某些实施方式中,多任务处理和并行处理也是可以的或者可能是有利的。
在20世纪90年代,对于一个技术的改进可以很明显地区分是硬件上的改进(例如,对二极管、晶体管、开关等电路结构的改进)还是软件上的改进(对于方法流程的改进)。然而,随着技术的发展,当今的很多方法流程的改进已经可以视为硬件电路结构的直接改进。设计人员几乎都通过将改进的方法流程编程到硬件电路中来得到相应的硬件电路结构。因此,不能说一个方法流程的改进就不能用硬件实体模块来实现。例如,可编程逻辑器件(Programmable Logic Device,PLD)(例如现场可编程门阵列(Field Programmable Gate Array,FPGA))就是这样一种集成电路,其逻辑功能由用户对器件编程来确定。由设计人员自行编程来把一个数字系统“集成”在一片PLD上,而不需要请芯片制造厂商来设计和制作专用的集成电路芯片。而且,如今,取代手工地制作集成电路芯片,这种编程也多半改用“逻辑编译器(logic compiler)”软件来实现,它与程序开发撰写时所用的软件编译器相类似,而要编译之前的原始代码也得用特定的编程语言来撰写,此称之为硬件描述语言(Hardware Description Language,HDL),而HDL也并非仅有一种,而是有许多种,如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)等,目前最普遍使用的是VHDL(Very-High-Speed Integrated Circuit Hardware Description Language)与Verilog。本领域技术人员也应该清楚,只需要将方法流程用上述几种硬件描述语言稍作逻辑编程并编程到集成电路中,就可以很容易得到实现该逻辑方法流程的硬件电路。
控制器可以按任何适当的方式实现,例如,控制器可以采取例如微处理器或处理器以及存储可由该(微)处理器执行的计算机可读程序代码(例如软件或固件)的计算机可读介质、逻辑门、开关、专用集成电路(Application Specific Integrated Circuit,ASIC)、可编程逻辑控制器和嵌入微控制器的形式,控制器的例子包括但不限于以下微控制器:ARC 625D、Atmel AT91SAM、Microchip PIC18F26K20以及Silicone Labs C8051F320,存储器控制器还可以被实现为存储器的控制逻辑的一部分。本领域技术人员也知道,除了以纯计算机可读程序代码方式实现控制器以外,完全可以通过将方法步骤进行逻辑编程来使得控制器以逻辑门、开关、专用集成电路、可编程逻辑控制器和嵌入微控制器等的形式来实现相同功能。因此这种控制器可以被认为是一种硬件部件,而对其内包括的用于实现各种功能的装置也可以视为硬件部件内的结构。或者甚至,可以将用于实现各种功能的装置视为既可以是实现方法的软件模块又可以是硬件部件内的结构。
上述实施例阐明的系统、装置、模块或单元,具体可以由计算机芯片或实体实现,或者由具有某种功能的产品来实现。一种典型的实现设备为计算机。具体的,计算机例如可以为个人计算机、膝上型计算机、蜂窝电话、相机电话、智能电话、个人数字助理、 媒体播放器、导航设备、电子邮件设备、游戏控制台、平板计算机、可穿戴设备或者这些设备中的任何设备的组合。
为了描述的方便,描述以上装置时以功能分为各种单元分别描述。当然,在实施本说明书一个或多个实施例时可以把各单元的功能在同一个或多个软件和/或硬件中实现。
本领域内的技术人员应明白,本说明书的实施例可提供为方法、系统、或计算机程序产品。因此,本说明书一个或多个实施例可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且,本说明书一个或多个实施例可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本说明书的实施例是参照根据本说明书实施例的方法、设备(系统)、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方框的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程欺诈案例的串并设备的处理器以产生一个机器,使得通过计算机或其他可编程欺诈案例的串并设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程欺诈案例的串并设备以特定方式工作的计算机可读存储器中,使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品,该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程欺诈案例的串并设备上,使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理,从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
在一个典型的配置中,计算设备包括一个或多个处理器(CPU)、输入/输出接口、网络接口和内存。
内存可能包括计算机可读介质中的非永久性存储器,随机存取存储器(RAM)和/或非易失性内存等形式,如只读存储器(ROM)或闪存(flash RAM)。内存是计算机可读介质的示例。
计算机可读介质包括永久性和非永久性、可移动和非可移动媒体可以由任何方法或技术来实现信息存储。信息可以是计算机可读指令、数据结构、程序的模块或其他数据。计算机的存储介质的例子包括,但不限于相变内存(PRAM)、静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、其他类型的随机存取存储器(RAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、快闪记忆体或其他内存技术、只读光盘只读存储器(CD-ROM)、数字多功能光盘(DVD)或其他光学存储、磁盒式磁带,磁带磁磁盘存储或其他磁性存储设备或任何其他非传输介质,可用于存储可以被计算设备访问的信息。按照本文中的界定,计算机可读介质不包括暂存电脑可读媒体(transitory media),如调制的数据信号和载波。
术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含,从而使得包括一系列要素的过程、方法、商品或者设备不仅包括那些要素,而且还包括没有明确列出的其他要素,或者是还包括为这种过程、方法、商品或者设备所固有的要素。在没有更多限制的情况下,由语句“包括一个……”限定的要素,并不排除在包括所述要素的过程、方法、商品或者设备中还存在另外的相同要素。
本领域技术人员应明白,本说明书的实施例可提供为方法、系统或计算机程序产品。因此,本说明书一个或多个实施例可采用完全硬件实施例、完全软件实施例或结合软件和硬件方面的实施例的形式。而且,本说明书一个或多个实施例可采用在一个或多个其 中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本说明书一个或多个实施例可以在由计算机执行的计算机可执行指令的一般上下文中描述,例如程序模块。一般地,程序模块包括执行特定任务或实现特定抽象数据类型的例程、程序、对象、组件、数据结构等等。也可以在分布式计算环境中实践本说明书一个或多个实施例,在这些分布式计算环境中,由通过通信网络而被连接的远程处理设备来执行任务。在分布式计算环境中,程序模块可以位于包括存储设备在内的本地和远程计算机存储介质中。
本说明书中的各个实施例均采用递进的方式描述,各个实施例之间相同相似的部分互相参见即可,每个实施例重点说明的都是与其他实施例的不同之处。尤其,对于系统实施例而言,由于其基本相似于方法实施例,所以描述的比较简单,相关之处参见方法实施例的部分说明即可。
以上所述仅为本说明书的实施例而已,并不用于限制本说明书。对于本领域技术人员来说,本说明书可以有各种更改和变化。凡在本说明书的精神和原理之内所作的任何修改、等同替换、改进等,均应包含在本说明书的权利要求范围之内。

Claims (20)

  1. 一种风险防控方法,应用于终端设备,所述方法包括:
    接收服务器下发的所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型;
    基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据;
    将所述风险防控子模型发送给所述服务器,以使所述服务器对所述终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型;
    接收所述服务器发送的所述终端设备所属的分类群组对应的风险防控模型,并基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
  2. 根据权利要求1所述方法,所述初始风险防控模型由预设的神经网络模型构建。
  3. 根据权利要求1所述方法,将所述风险防控子模型发送给所述服务器,包括:
    使用预设的第一加密算法对所述风险防控子模型进行加密处理,得到加密后的风险防控子模型,并将所述加密后的风险防控子模型发送给所述服务器。
  4. 根据权利要求1所述的方法,所述方法还包括:
    接收服务器下发的待训练的第一模型,所述第一模型中的模型参数的参数值为随机产生的参数值或预设的参数值;
    基于所述训练样本数据对所述第一模型进行模型训练,得到训练后的第一模型,并获取所述训练后的第一模型中的预设网络层中的参数和相应的参数值;
    将所述训练后的第一模型中的预设网络层中的参数和相应的参数值发送给所述服务器,以使所述服务器基于不同终端设备提供的训练后的第一模型中的预设网络层中的参数和相应的参数值,对不同的终端设备进行聚类,得到不同终端设备所属的分类群组。
  5. 根据权利要求4所述的方法,所述第一模型由预设的神经网络模型构建,所述第一模型与所述初始风险防控模型不同。
  6. 根据权利要求4所述的方法,所述终端设备所属的分类群组是所述服务器基于不同终端设备提供的训练后的第一模型中的预设网络层中的参数和相应的参数值,通过预设的聚类算法对不同的终端设备进行聚类得到。
  7. 根据权利要求6所述的方法,所述聚类算法为K-means聚类算法。
  8. 根据权利要求4所述的方法,所述将所述训练后的第一模型中的预设网络层中的参数和相应的参数值发送给所述服务器,包括:
    使用预设的第二加密算法对所述训练后的第一模型中的预设网络层中的参数和相应的参数值进行加密处理,得到加密后的参数和相应的加密后的参数值,并将加密后的参数和相应的加密后的参数值发送给所述服务器。
  9. 根据权利要求1所述的方法,所述基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理,包括:
    获取待检测的所述目标业务的业务数据;
    将所述业务数据输入到所述风险防控模型中,以对所述业务数据是否存在预设风险进行检测,得到相应的检测结果;
    如果所述检测结果指示所述业务数据存在预设风险,则取消所述终端设备执行所述目标业务的业务处理。
  10. 根据权利要求1所述的方法,所述方法还包括:
    如果检测到所述训练样本数据被更新,则获取更新后的训练样本数据;
    基于更新后的训练样本数据对所述风险防控模型进行模型训练,得到训练后的风险防控模型,并基于训练后的风险防控模型对所述目标业务的数据进行风险防控处理。
  11. 根据权利要求1所述的方法,所述方法还包括:
    如果检测到所述训练样本数据被更新,则获取更新后的所述训练样本数据;
    基于更新后的所述训练样本数据对所述风险防控模型进行模型训练,得到训练后的风险防控模型,并将训练后的风险防控模型发送给所述服务器,以使所述服务器对所述终端设备所属的分类群组对应的风险防控模型进行更新,得到更新后的风险防控模型;
    接收所述服务器发送的更新后的风险防控模型,并基于更新后的风险防控模型对所述目标业务的数据进行风险防控处理。
  12. 一种风险防控方法,应用于服务器,所述方法包括:
    获取终端设备所属的分类群组的信息;
    基于所述终端设备所属的分类群组的信息,获取所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将所述初始风险防控模型下发到所述终端设备,以使所述终端设备基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据;
    接收所述终端设备发送的所述风险防控子模型,并对所述终端设备发送的所述风险防控子模型和所述终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型,所述其它终端设备提供的风险防控子模型是所述其它终端设备基于所述其它终端设备中预先存储的训练样本数据对所述初始风险防控模型进行模型训练得到;
    将所述风险防控模型发送给所述终端设备,以使所述终端设备基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
  13. 根据权利要求12所述的方法,所述方法还包括:
    向所述终端设备下发待训练的第一模型,所述第一模型中的模型参数的参数值为随机产生的参数值或预设的参数值,以使所述终端设备基于所述训练样本数据对所述第一模型进行模型训练,得到训练后的第一模型,并将所述训练后的第一模型中的预设网络层中的参数和相应的参数值发送给所述服务器;
    接收所述终端设备发送的所述训练后的第一模型中的预设网络层中的参数和相应的参数值,并基于训练后的第一模型中的预设网络层中的参数和相应的参数值,通过预设的聚类算法对所述终端设备进行聚类,得到所述终端设备所属的分类群组。
  14. 根据权利要求12所述的方法,所述方法还包括:
    接收所述终端设备发送的训练后的所述风险防控模型,训练后的所述风险防控模型是所述终端设备在检测到所述训练样本数据被更新时,基于更新后的所述训练样本数据对所述风险防控模型进行模型训练得到;
    基于训练后的所述风险防控模型,对所述终端设备所属的分类群组对应的风险防控模型进行更新,得到更新后的风险防控模型;
    将更新后的风险防控模型发送给所述终端设备,以使所述终端设备基于更新后的风险防控模型对所述目标业务的数据进行风险防控处理。
  15. 一种风险防控装置,所述装置包括:
    初始模型接收模块,接收服务器下发的所述装置所属的分类群组对应的待训练的目标业务的初始风险防控模型;
    子模型训练模块,基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述装置对应的风险防控子模型,所述训练样本数据中至少包括与所述装置的用户和所述目标业务相关的数据;
    子模型发送模块,将所述风险防控子模型发送给所述服务器,以使所述服务器对所述装置所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到所述装置所属的分类群组对应的风险防控模型;
    风险防控模块,接收所述服务器发送的所述装置所属的分类群组对应的风险防控模型,并基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
  16. 一种风险防控装置,所述装置包括:
    群组信息获取模块,获取终端设备所属的分类群组的信息;
    初始模型下发模块,基于所述终端设备所属的分类群组的信息,获取所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将所述初始风险防控模型下发到所述终端设备,以使所述终端设备基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据;
    子模型接收模块,接收所述终端设备发送的所述风险防控子模型,并对所述终端设备发送的所述风险防控子模型和所述终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型,所述其它终端设备提供的风险防控子模型是所述其它终端设备基于所述其它终端设备中预先存储的训练样本数据对所述初始风险防控模型进行模型训练得到;
    风控模型下发模块,将所述风险防控模型发送给所述终端设备,以使所述终端设备基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
  17. 一种风险防控设备,包括:
    处理器;以及
    被安排成存储计算机可执行指令的存储器,所述可执行指令在被执行时使所述处理器:
    接收服务器下发的所述设备所属的分类群组对应的待训练的目标业务的初始风险防控模型;
    基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述设备对应的风险防控子模型,所述训练样本数据中至少包括与所述设备的用户和所述目标业务相关的数据;
    将所述风险防控子模型发送给所述服务器,以使所述服务器对所述设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到所述设备所属的分类群组对应的风险防控模型;
    接收所述服务器发送的所述设备所属的分类群组对应的风险防控模型,并基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
  18. 一种风险防控设备,包括:
    处理器;以及
    被安排成存储计算机可执行指令的存储器,所述可执行指令在被执行时使所述处理器:
    获取终端设备所属的分类群组的信息;
    基于所述终端设备所属的分类群组的信息,获取所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将所述初始风险防控模型下发到所述终端设备,以使所述终端设备基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据;
    接收所述终端设备发送的所述风险防控子模型,并对所述终端设备发送的所述风险防控子模型和所述终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型,所述其它终端设备提供的风险防控子模型是所述其它终端设备基于所述其它终端设备中预先存储的训练样本数据对所述初始风险防控模型进行模型训练得到;
    将所述风险防控模型发送给所述终端设备,以使所述终端设备基于所述风险防控模 型对获取的所述目标业务的数据进行风险防控处理。
  19. 一种存储介质,用于存储计算机可执行指令,所述可执行指令在被执行时实现以下流程:
    接收服务器下发的终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型;
    基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据;
    将所述风险防控子模型发送给所述服务器,以使所述服务器对所述终端设备所属的分类群组中的不同终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型;
    接收所述服务器发送的所述终端设备所属的分类群组对应的风险防控模型,并基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
  20. 一种存储介质,所述存储介质用于存储计算机可执行指令,所述可执行指令在被执行时实现以下流程:
    获取终端设备所属的分类群组的信息;
    基于所述终端设备所属的分类群组的信息,获取所述终端设备所属的分类群组对应的待训练的目标业务的初始风险防控模型,并将所述初始风险防控模型下发到所述终端设备,以使所述终端设备基于预先存储的训练样本数据对所述初始风险防控模型进行模型训练,得到所述终端设备对应的风险防控子模型,所述训练样本数据中至少包括与所述终端设备的用户和所述目标业务相关的数据;
    接收所述终端设备发送的所述风险防控子模型,并对所述终端设备发送的所述风险防控子模型和所述终端设备所属的分类群组中的其它终端设备提供的风险防控子模型进行模型融合处理,得到所述终端设备所属的分类群组对应的风险防控模型,所述其它终端设备提供的风险防控子模型是所述其它终端设备基于所述其它终端设备中预先存储的训练样本数据对所述初始风险防控模型进行模型训练得到;
    将所述风险防控模型发送给所述终端设备,以使所述终端设备基于所述风险防控模型对获取的所述目标业务的数据进行风险防控处理。
PCT/CN2022/093895 2021-06-07 2022-05-19 风险防控的方法、装置及设备 WO2022257723A1 (zh)

Priority Applications (1)

Application Number Priority Date Filing Date Title
US18/526,555 US20240104570A1 (en) 2021-06-07 2023-12-01 Risk prevention and control methods, apparatuses, and devices

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110631595.0A CN113312667B (zh) 2021-06-07 2021-06-07 一种风险防控方法、装置及设备
CN202110631595.0 2021-06-07

Related Child Applications (1)

Application Number Title Priority Date Filing Date
US18/526,555 Continuation US20240104570A1 (en) 2021-06-07 2023-12-01 Risk prevention and control methods, apparatuses, and devices

Publications (1)

Publication Number Publication Date
WO2022257723A1 true WO2022257723A1 (zh) 2022-12-15

Family

ID=77377926

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/093895 WO2022257723A1 (zh) 2021-06-07 2022-05-19 风险防控的方法、装置及设备

Country Status (3)

Country Link
US (1) US20240104570A1 (zh)
CN (1) CN113312667B (zh)
WO (1) WO2022257723A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115952859A (zh) * 2023-03-01 2023-04-11 支付宝(杭州)信息技术有限公司 数据处理方法、装置及设备
CN116070916A (zh) * 2023-03-06 2023-05-05 支付宝(杭州)信息技术有限公司 数据处理方法、装置及设备
CN116151627A (zh) * 2023-04-04 2023-05-23 支付宝(杭州)信息技术有限公司 一种业务风控的方法、装置、存储介质及电子设备
CN116797829A (zh) * 2023-06-13 2023-09-22 北京百度网讯科技有限公司 一种模型生成方法、图像分类方法、装置、设备及介质

Families Citing this family (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113312667B (zh) * 2021-06-07 2022-09-02 支付宝(杭州)信息技术有限公司 一种风险防控方法、装置及设备
US20220414528A1 (en) * 2021-06-24 2022-12-29 Paypal, Inc. Edge Device Machine Learning
CN114090243B (zh) * 2021-11-10 2024-06-18 支付宝(杭州)信息技术有限公司 模型计算方法和装置
CN113992429B (zh) * 2021-12-22 2022-04-29 支付宝(杭州)信息技术有限公司 一种事件的处理方法、装置及设备
CN114819614B (zh) * 2022-04-22 2024-10-15 支付宝(杭州)信息技术有限公司 数据处理方法、装置、系统及设备
WO2024060227A1 (zh) * 2022-09-23 2024-03-28 Oppo广东移动通信有限公司 模型生成方法、信息处理方法和设备
CN115545720B (zh) * 2022-11-29 2023-03-10 支付宝(杭州)信息技术有限公司 一种模型训练的方法、业务风控的方法及装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110222880A (zh) * 2019-05-20 2019-09-10 阿里巴巴集团控股有限公司 业务风险的确定方法、模型训练方法和数据处理方法
US20200202268A1 (en) * 2018-12-20 2020-06-25 Accenture Global Solutions Limited Utilizing artificial intelligence to predict risk and compliance actionable insights, predict remediation incidents, and accelerate a remediation process
CN111738628A (zh) * 2020-08-14 2020-10-02 支付宝(杭州)信息技术有限公司 一种风险群组识别方法及装置
CN112906903A (zh) * 2021-01-11 2021-06-04 北京源堡科技有限公司 网络安全风险预测方法、装置、存储介质及计算机设备
CN113312667A (zh) * 2021-06-07 2021-08-27 支付宝(杭州)信息技术有限公司 一种风险防控方法、装置及设备

Family Cites Families (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN105490841B (zh) * 2015-11-26 2019-03-01 广州华多网络科技有限公司 一种终端日志抓取方法、装置及系统
US11977960B2 (en) * 2018-08-09 2024-05-07 Autodesk, Inc. Techniques for generating designs that reflect stylistic preferences
EP3970074A1 (en) * 2019-05-16 2022-03-23 FRAUNHOFER-GESELLSCHAFT zur Förderung der angewandten Forschung e.V. Concepts for federated learning, client classification and training data similarity measurement
CN111008709A (zh) * 2020-03-10 2020-04-14 支付宝(杭州)信息技术有限公司 联邦学习、资料风险评估方法、装置和系统
CN111831523A (zh) * 2020-06-24 2020-10-27 上海识装信息科技有限公司 一种客户端无感知排查线上问题的方法及系统
CN111866869B (zh) * 2020-07-07 2023-06-23 兰州交通大学 面向边缘计算的联邦学习室内定位隐私保护方法
CN112256874B (zh) * 2020-10-21 2023-08-08 平安科技(深圳)有限公司 模型训练方法、文本分类方法、装置、计算机设备和介质
CN112181971B (zh) * 2020-10-27 2022-11-01 华侨大学 一种基于边缘的联邦学习模型清洗和设备聚类方法、系统
CN112270597A (zh) * 2020-11-10 2021-01-26 恒安嘉新(北京)科技股份公司 业务处理、信用评价模型训练方法、装置、设备及介质
CN112465626B (zh) * 2020-11-24 2023-08-29 平安科技(深圳)有限公司 基于客户端分类聚合的联合风险评估方法及相关设备
CN112488322B (zh) * 2020-12-15 2024-02-13 杭州电子科技大学 一种基于数据特征感知聚合的联邦学习模型训练方法
CN112819156A (zh) * 2021-01-26 2021-05-18 支付宝(杭州)信息技术有限公司 一种数据处理方法、装置及设备
CN112819180B (zh) * 2021-01-26 2021-10-15 华中科技大学 一种基于联邦生成模型的多业务数据生成方法和装置
CN112884164B (zh) * 2021-03-18 2023-06-23 中国地质大学(北京) 面向智能移动终端实现的联邦机器学习迁移方法与系统
CN112712182B (zh) * 2021-03-29 2021-06-01 腾讯科技(深圳)有限公司 一种基于联邦学习的模型训练方法、装置及存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20200202268A1 (en) * 2018-12-20 2020-06-25 Accenture Global Solutions Limited Utilizing artificial intelligence to predict risk and compliance actionable insights, predict remediation incidents, and accelerate a remediation process
CN110222880A (zh) * 2019-05-20 2019-09-10 阿里巴巴集团控股有限公司 业务风险的确定方法、模型训练方法和数据处理方法
CN111738628A (zh) * 2020-08-14 2020-10-02 支付宝(杭州)信息技术有限公司 一种风险群组识别方法及装置
CN112906903A (zh) * 2021-01-11 2021-06-04 北京源堡科技有限公司 网络安全风险预测方法、装置、存储介质及计算机设备
CN113312667A (zh) * 2021-06-07 2021-08-27 支付宝(杭州)信息技术有限公司 一种风险防控方法、装置及设备

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115952859A (zh) * 2023-03-01 2023-04-11 支付宝(杭州)信息技术有限公司 数据处理方法、装置及设备
CN116070916A (zh) * 2023-03-06 2023-05-05 支付宝(杭州)信息技术有限公司 数据处理方法、装置及设备
CN116151627A (zh) * 2023-04-04 2023-05-23 支付宝(杭州)信息技术有限公司 一种业务风控的方法、装置、存储介质及电子设备
CN116151627B (zh) * 2023-04-04 2023-09-01 支付宝(杭州)信息技术有限公司 一种业务风控的方法、装置、存储介质及电子设备
CN116797829A (zh) * 2023-06-13 2023-09-22 北京百度网讯科技有限公司 一种模型生成方法、图像分类方法、装置、设备及介质

Also Published As

Publication number Publication date
CN113312667A (zh) 2021-08-27
CN113312667B (zh) 2022-09-02
US20240104570A1 (en) 2024-03-28

Similar Documents

Publication Publication Date Title
WO2022257723A1 (zh) 风险防控的方法、装置及设备
KR102219627B1 (ko) 생성적 적대 신경망을 이용한 비공개화된 머신 러닝
US11710035B2 (en) Distributed labeling for supervised learning
US20210297267A1 (en) Blockchain consensus node selection
US11989634B2 (en) Private federated learning with protection against reconstruction
WO2019014577A1 (en) METHOD AND DEVICE FOR PROCESSING CONNECTION INFORMATION
WO2023138419A1 (zh) 模型的训练
US10972258B2 (en) Contextual key management for data encryption
GB2527605A (en) System and method for dynamically generating contextual and personalised digital content
CN111985921B (zh) 基于区块链离线支付的验证处理方法及数字金融服务平台
WO2020155173A1 (en) Data processing method, device and system for machine learning model
JP2023014431A (ja) 安全な照会処理のための方法、コンピュータプログラム、およびコンピュータシステム(グラフストアにおける安全な照会処理)
CN114553516B (zh) 一种数据的处理方法、装置及设备
CN114638998A (zh) 模型更新方法、装置、系统及设备
US20210081295A1 (en) Device-usage processing for generating inferred user states
CN112967044A (zh) 一种支付业务的处理方法及装置
Chan Privacy-preserving human action recognition with joint edge-cloud computing
US20230089413A1 (en) Navigation assistance device
CN117689006A (zh) 联邦迁移学习方法及装置
Liao et al. Technologies for Service Capability Collaboration Interconnection
CN117095206A (zh) 生物检测模型训练方法及装置
CN116521360A (zh) 一种数据处理系统、方法及装置
US20190037035A1 (en) Distributed Automated Learning of User Personalization
CN115186278A (zh) 数据的处理方法、装置及设备
CN114662706A (zh) 一种模型训练方法、装置及设备

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 22819321

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 22819321

Country of ref document: EP

Kind code of ref document: A1