A privacy protection method, device and equipment for biometric recognition

Abstract

The embodiment of the specification discloses a privacy protection method, device and equipment for biological identification. The method is applied to a terminal device and includes the following steps: obtaining a biological identification request, wherein the biological identification request carries biological information of a target user; inputting the biological information into a biological attribute model obtained through model training in advance, extracting user attribute information of the target user from the biological information through the biological attribute model, and obtaining user attribute information corresponding to the biological information; sending the user attribute information corresponding to the biological information to a server to trigger the server to determine a privacy protection model for desensitizing the biological information based on the distribution of the user attribute information of the target user; receiving the privacy protection model issued by the server, desensitizing the biological information based on the privacy protection model, obtaining desensitized biological information, and performing biological identification processing corresponding to the biological identification request based on the desensitized biological information.

Description

TECHNICAL FIELD

[0001] The present document relates to the technical field of computer, and particularly relates to a privacy protection method, device and equipment for biometric recognition. BACKGROUND

[0002] With the wide application of biometric recognition technology in recent years, more and more biometric recognition products have entered people's production and life, such as face payment, face access control, fingerprint unlocking of mobile phones, and the like. However, while the biometric recognition technology provides convenience for people, the risk of leakage of private information gradually emerges. Since the biometric recognition system often collects, processes, transmits and stores the biological information of users, each of the above-mentioned links may leak the biological information of users, and the biological information itself is private information, once leaked, it will cause loss to users, and the leakage of biological information will also cause a chain reaction, further endangering the property and information security of users.

[0003] Generally, privacy protection processing can be performed by means of information encryption, specifically, simple linear operations are used to encrypt or perform row and column confusion processing on user biological information, but the above-mentioned method is simple in atomic operation and single in process, and is easy to be cracked by means such as brute force cracking. Based on this, it is necessary to provide a technical scheme of user biometric recognition with higher security, stronger privacy protection capability and the ability to quickly adapt to new scenarios. SUMMARY

[0004] The purpose of the embodiments of the present specification is to provide a technical scheme of user biometric recognition with higher security, stronger privacy protection capability and the ability to quickly adapt to new scenarios.

[0005] In order to achieve the above technical scheme, the embodiments of the present specification are implemented as follows:

[0006] The privacy protection method for biometric recognition provided by the embodiments of the present specification is applied to a terminal device, and the method comprises the following steps: obtaining a biometric recognition request, wherein the biometric recognition request carries biological information of a target user; inputting the biological information into a biological attribute model obtained through model training in advance, extracting user attribute information of the target user from the biological information through the biological attribute model, and obtaining user attribute information corresponding to the biological information; sending the user attribute information corresponding to the biological information to a server, wherein the user attribute information is used to trigger the server to determine a privacy protection model for desensitizing processing of the biological information based on a distribution of the user attribute information of the target user; receiving the privacy protection model issued by the server, and performing desensitizing processing on the biological information based on the privacy protection model to obtain desensitized biological information, and performing biometric recognition processing corresponding to the biometric recognition request based on the desensitized biological information.

[0007] The privacy protection method for biometric recognition provided by the embodiments of the present specification is applied to a server, and the method comprises: receiving user attribute information corresponding to biometric information of a target user sent by a terminal device, the biometric information of the target user being information obtained by the terminal device when obtaining a biometric recognition request of the target user, the user attribute information being information extracted from the biometric information by inputting the biometric information into a pre-trained biometric attribute model. Based on the distribution of the user attribute information of the target user, a privacy protection model for desensitizing the biometric information is determined. The determined privacy protection model is sent to the terminal device, and the privacy protection model is used to trigger the terminal device to desensitize the biometric information to obtain desensitized biometric information, and to perform biometric recognition processing corresponding to the biometric recognition request based on the desensitized biometric information.

[0008] The privacy protection device for biometric recognition provided by the embodiments of the present specification comprises: a request obtaining module that obtains a biometric recognition request, the biometric recognition request carrying biometric information of a target user. An attribute extraction module inputs the biometric information into a pre-trained biometric attribute model, extracts user attribute information of the target user from the biometric information through the biometric attribute model, and obtains user attribute information corresponding to the biometric information. An attribute information sending module sends the user attribute information corresponding to the biometric information to a server, and the user attribute information is used to trigger the server to determine a privacy protection model for desensitizing the biometric information based on the distribution of the user attribute information of the target user. A desensitization module receives the privacy protection model sent by the server, desensitizes the biometric information based on the privacy protection model to obtain desensitized biometric information, and performs biometric recognition processing corresponding to the biometric recognition request based on the desensitized biometric information.

[0009] The embodiment of the present specification provides a privacy protection device for biological identification, the device comprises: an attribute information receiving module, which receives user attribute information corresponding to biological information of a target user sent by a terminal device, the biological information of the target user is information obtained by the terminal device when obtaining a biological identification request of the target user, and the user attribute information is information extracted from the biological information by inputting the biological information into a pre-trained biological attribute model. A model determining module determines a privacy protection model for desensitizing the biological information based on the distribution of the user attribute information of the target user. A model issuing module issues the determined privacy protection model to the terminal device, and the privacy protection model is used to trigger the terminal device to desensitize the biological information to obtain desensitized biological information and perform biological identification processing corresponding to the biological identification request based on the desensitized biological information.

[0010] The embodiment of the present specification provides a privacy protection device for biological identification, the device comprises: a processor; and a memory arranged to store computer executable instructions, which, when executed, cause the processor to: obtain a biological identification request, the biological identification request carrying biological information of a target user. Input the biological information into a biological attribute model obtained by model training in advance, extract user attribute information of the target user from the biological information through the biological attribute model, and obtain user attribute information corresponding to the biological information. Send the user attribute information corresponding to the biological information to a server, and the user attribute information is used to trigger the server to determine a privacy protection model for desensitizing the biological information based on the distribution of the user attribute information of the target user. Receive the privacy protection model issued by the server, and desensitize the biological information based on the privacy protection model to obtain desensitized biological information, and perform biological identification processing corresponding to the biological identification request based on the desensitized biological information.

[0011] An embodiment of the present specification provides a privacy protection device for biometric recognition, comprising: a processor; and a memory arranged to store computer executable instructions, which, when executed, cause the processor to: receive user attribute information corresponding to biometric information of a target user sent by a terminal device, the biometric information of the target user being information obtained by the terminal device when obtaining a biometric recognition request of the target user, the user attribute information being information extracted from the biometric information by inputting the biometric information into a pre-trained biometric attribute model. Determine a privacy protection model for desensitizing the biometric information based on a distribution of the user attribute information of the target user. Issue the determined privacy protection model to the terminal device, and the privacy protection model is used to trigger the terminal device to desensitize the biometric information to obtain desensitized biometric information, and perform biometric recognition processing corresponding to the biometric recognition request based on the desensitized biometric information.

[0012] An embodiment of the present specification also provides a storage medium for storing computer executable instructions, which, when executed by a processor, implement the following processes: obtaining a biometric recognition request, the biometric recognition request carrying biometric information of a target user. Input the biometric information into a biometric attribute model obtained by model training in advance, extract user attribute information of the target user from the biometric information through the biometric attribute model, and obtain user attribute information corresponding to the biometric information. Send the user attribute information corresponding to the biometric information to a server, and the user attribute information is used to trigger the server to determine a privacy protection model for desensitizing the biometric information based on a distribution of the user attribute information of the target user. Receive the privacy protection model issued by the server, and desensitize the biometric information based on the privacy protection model to obtain desensitized biometric information, and perform biometric recognition processing corresponding to the biometric recognition request based on the desensitized biometric information.

[0013] The embodiment of the present specification further provides a storage medium for storing computer executable instructions, which, when executed by a processor, implement the following process: receiving user attribute information corresponding to biological information of a target user sent by a terminal device, the biological information of the target user being information obtained when the terminal device obtains a biological recognition request of the target user, and the user attribute information being information extracted from the biological information by inputting the biological information into a pre-trained biological attribute model. A privacy protection model for desensitizing the biological information is determined based on the distribution of the user attribute information of the target user. The determined privacy protection model is issued to the terminal device, and the privacy protection model is used to trigger the terminal device to desensitize the biological information to obtain desensitized biological information, and perform biological recognition processing corresponding to the biological recognition request based on the desensitized biological information. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present specification or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments described in the present specification, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0015] Figure 1A A biological recognition privacy protection method embodiment of the present specification;

[0016] Figure 1B A biological recognition privacy protection process schematic diagram of the present specification;

[0017] Figure 2 A biological recognition privacy protection system structure schematic diagram of the present specification;

[0018] Figure 3 Another biological recognition privacy protection process schematic diagram of the present specification;

[0019] Figure 4A Another biological recognition privacy protection method embodiment of the present specification;

[0020] Figure 4B Another biological recognition privacy protection process schematic diagram of the present specification;

[0021] Figure 5 Another biological recognition privacy protection process schematic diagram of the present specification;

[0022] Figure 6 A biological recognition privacy protection device embodiment of the present specification;

[0023] Figure 7 Another privacy protection device embodiment of the present specification for biometric recognition;

[0024] Figure 8 A privacy protection device embodiment of the present specification for biometric recognition. DETAILED DESCRIPTION

[0025] The embodiments of the present specification provide a privacy protection method, device and equipment for biometric recognition.

[0026] In order for those skilled in the art to better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be described clearly and completely in conjunction with the drawings in the embodiments of the present specification. Obviously, the described embodiments are only part of the embodiments of the present specification, not all. Based on the embodiments in the present specification, all other embodiments obtained by those skilled in the art without creative labor should be within the scope of protection of the present specification.

[0027] Embodiment one

[0028] As shown in Figure 1A and Figure 1B The embodiments of the present specification provide a privacy protection method for biometric recognition. The execution subject of the method can be a terminal device, which can be a certain terminal device such as a mobile phone, a tablet computer, etc., can also be a computer device such as a notebook computer or a desktop computer, or can also be an IoT device (such as a smart watch, a vehicle-mounted device, etc.), and can also be a machine with certain functions set in advance, etc. The method can specifically include the following steps:

[0029] In step S102, a biometric recognition request is obtained, which carries the biometric information of a target user.

[0030] The target user can be any user, such as the owner of the terminal device described above, and the target user can initiate a biometric recognition request through the terminal device. The user biometric information can include various information such as the user's fingerprint information, palmprint information, face information or iris information, etc. In actual application, the carrier of the above-mentioned user biometric information can include various forms, such as carrying the above-mentioned various user biometric information in the form of image, etc. The specific form can be set according to the actual situation, and the embodiments of the present specification do not limit it.

[0031] In implementation, with the wide application of biometric technology in recent years, more and more biometric products have entered people's production and life, such as face payment, face access control, fingerprint unlocking of mobile phones, and the like. However, while biometric technology provides convenience for people, the risk of leakage of private information gradually emerges. Since the biometric system often collects, processes, transmits, and stores the biological information of the user, each of the above-mentioned links may leak the biological information of the user, and the biological information itself is private information, which will cause loss to the user once leaked, and the leakage of biological information will also cause a chain reaction, further endangering the property and information security of the user.

[0032] The privacy protection capability becomes an important capability of the biometric system. Generally, privacy protection processing can be performed by information encryption. Specifically, simple linear operations are used to encrypt or perform row and column confusion on the biological information of the user, but the above-mentioned method has simple atomic operations, single process, and is easy to be cracked by methods such as brute force cracking. In addition, privacy protection processing can also be performed by deep learning. Specifically, a deep learning model (such as a neural network model) is trained and used to perform privacy protection processing on the biological information of the user to obtain the biological information of the user after privacy protection, but since the deep learning model is a data-driven model, its performance will sharply decay on some unemerged data domains (for example, for face recognition, the age distribution of the face may change greatly), thereby reducing the privacy protection capability, and this method often needs a relatively long preparation period (data collection, model training, etc.) to be deployed to a new scene and business. Therefore, it is necessary to provide a user biometric technology scheme with higher security, stronger privacy protection capability, and the ability to quickly adapt to new scenes. The embodiments of the present specification provide an implementable technical scheme, which can specifically include the following contents:

[0033] When a user (i.e., a target user) needs to perform a specified business (such as a payment business, a login business, etc.), the execution mechanism of the specified business can be triggered, and the identity of the user often needs to be identified before the specified business is executed. At this time, the terminal device can start the corresponding information collection component, such as a fingerprint collection component, a camera component, a palmprint collection component, etc., and can collect the user biological information of the user through the information collection component. Specifically, the fingerprint image, face image, palmprint image, or iris image of the user can be collected. The terminal device can generate a biometric request based on the collected user biological information, and the terminal device can obtain the biometric request.

[0034] In step S104, the biological information is input into a biological attribute model obtained through model training in advance, and the user attribute information of the target user is extracted from the biological information through the biological attribute model, to obtain the user attribute information corresponding to the biological information.

[0035] The biological attribute model can be a model for extracting user attribute information of a user from biological information. The biological attribute model can be constructed by various algorithms, for example, the biological attribute model can be constructed by a neural network algorithm, or the biological attribute model can be constructed by a random forest algorithm, etc. The specific implementation can be determined according to actual conditions, and the present application does not limit the specific implementation. The user attribute information can include various information, such as the gender of the user, the age of the user, the expression of the user, the resolution of the biological information presented by the user, etc. The specific implementation can be determined according to actual conditions, and the present application does not limit the specific implementation.

[0036] In implementation, the initial architecture of the biological attribute model can be constructed by a preset algorithm, then the biological information of different users can be obtained and used as training sample data, then a corresponding loss function can be set, and the biological information obtained and the loss function can be used to train the biological attribute model to obtain a trained biological attribute model, so that the biological attribute model obtained finally can extract the user attribute information of the user from the biological information. The multiple sets of adversarial sample data can be obtained in various ways, for example, multiple original sample data (i.e., biological information of a user) can be obtained, the original sample data can be input into a pre-trained adversarial network model (such as a neural network model, etc.), and corresponding adversarial sample data can be generated by the adversarial network model. One set of adversarial sample data composed of the original sample data and the generated adversarial sample data can be obtained by the above method, and multiple sets of adversarial sample data can be obtained in the same way. Alternatively, multiple original sample data can be obtained, random noise data can be generated for each original sample data, and the random noise data can be added to the corresponding original sample data to obtain adversarial sample data satisfying the adversarial sample property (i.e., the original sample data and the adversarial sample data are the same in visual perception, but the recognition result of the adversarial sample data obtained by the pre-trained model is different from that of the original sample data (or the similarity between the features of the adversarial sample data and the features of the original sample data is less than a preset threshold, i.e., the features of the adversarial sample data and the features of the original sample data are not the same or similar)). One set of adversarial sample data composed of the original sample data and the adversarial sample data can be obtained, and multiple sets of adversarial sample data can be obtained in the same way. The specific implementation can be determined according to actual conditions, and the present application does not limit the specific implementation.

[0037] In addition, the biological attribute model can also be trained in an adversarial feature manner. Specifically, an adversarial feature can be obtained, which can be determined based on a feature corresponding to specified training sample data (i.e., biological information of a user) (for example, adding disturbance data to the feature corresponding to the specified training sample data to obtain an adversarial feature satisfying the adversarial sample property, etc.), or can also be obtained through a pre-trained model, which can be set according to actual conditions. A network layer of the biological attribute model can be selected as a feature space, and then the adversarial feature can be input into the selected network layer, and the biological attribute model can be trained in combination with a preset loss function to obtain a trained biological attribute model, so that the trained biological attribute model has high generalization ability for data.

[0038] It should be noted that the training process of the above model can be performed in the terminal device or by the server. When the server trains the biological attribute model, the trained biological attribute model can be provided to the terminal device, and the terminal device can store the trained biological attribute model in the storage device for subsequent use.

[0039] When the biological recognition request is obtained, the biological information in the biological recognition request can be input into the trained biological attribute model, and the user attribute information of the target user can be extracted from the biological information by the biological attribute model, so as to obtain the user attribute information corresponding to the biological information, such as the age, gender, expression performance degree of the user, and resolution of the biological information of the user.

[0040] In step S106, the user attribute information corresponding to the biological information is sent to the server, and the user attribute information is used to trigger the server to determine a privacy protection model for desensitizing the biological information based on the distribution of the user attribute information of the target user.

[0041] The distribution of the user attribute information can include multiple distributions, for example, when the attribute item is gender, the distribution of the user attribute information can be that the proportion of males is 90%, the proportion of females is 10%, or the proportion of males is 25%, the proportion of females is 75%, etc., which can be set according to actual conditions. The privacy protection model can be a model for privacy protection processing of specified data, and the privacy protection model can be constructed through multiple different algorithms, for example, an encoder can be constructed through a neural network algorithm, or the privacy protection model can be constructed through a random forest algorithm, etc., which can be set according to actual conditions, and the embodiments of the present specification do not limit the same.

[0042] In implementation, the server can pre-construct an initial architecture of the privacy protection model through a preset algorithm, then can obtain the distribution of the user attribute information of different users and take the same as training sample data, then can set a corresponding loss function, can use the obtained distribution of the user attribute information and the loss function to train the privacy protection model, obtain the trained privacy protection model, so that the finally obtained privacy protection model can adapt to different scenes and perform desensitization processing on the biological information of the user in the scene. Among them, a plurality of sets of adversarial sample data can be obtained in a plurality of different ways, for example, a plurality of original sample data (i.e. the distribution of the user attribute information of different users) can be obtained, the original sample data can be input into a pre-trained adversarial network model (such as a neural network model, etc.), and the corresponding adversarial sample data can be generated through the adversarial network model. Through the above-mentioned manner, a set of adversarial sample data composed of original sample data and generated adversarial sample data can be obtained, and a plurality of sets of adversarial sample data can be obtained in the same way. Alternatively, a plurality of original sample data (i.e. the distribution of the user attribute information of different users) can be obtained, and random noise data can be generated for each original sample data. The random noise data is added to the corresponding original sample data to obtain adversarial sample data that meets the adversarial sample property (i.e. the original sample data is the same in visual perception, but the recognition result obtained by identifying the adversarial sample data through the pre-trained model is different from the original sample data (or the similarity between the features of the adversarial sample data and the features of the original sample data is less than a preset threshold, i.e. the features of the adversarial sample data and the features of the original sample data are not the same or similar)). The original sample data and the adversarial sample data can form a set of adversarial sample data, thereby obtaining a plurality of sets of adversarial sample data, etc. The specific setting can be determined according to the actual situation, and the embodiments of the present application are not limited thereto.

[0043] In addition, the privacy protection model can also be trained in the manner of adversarial features. Specifically, the adversarial features can be obtained, which can be determined based on the features corresponding to the specified training sample data (i.e. the distribution of the user attribute information of different users) (such as adding disturbance data to the features corresponding to the specified training sample data to obtain adversarial features that meet the adversarial sample property, etc.), or can be obtained through a pre-trained model. The specific setting can be determined according to the actual situation. A network layer can be selected as a feature space in the network layer of the privacy protection model, then the adversarial features can be input into the selected network layer, and the privacy protection model can be trained in combination with a preset loss function to obtain a trained privacy protection model, so that the trained privacy protection model has high generalization ability for data.

[0044] For example, Figure 2As shown, the user attribute information corresponding to the biological information can be sent to the server. After the server receives the user attribute information, the distribution of the user attribute information of the target user can be determined. Based on the distribution of the user attribute information of the target user, the privacy protection model corresponding to the distribution of the user attribute information can be obtained from the correspondence between the stored distribution of the user attribute information and the privacy protection model, so as to obtain the privacy protection model suitable for the current scene. The obtained privacy protection model can be determined as the privacy protection model for desensitizing the biological information.

[0045] In step S108, the privacy protection model issued by the receiving server is received, and the biological information is desensitized based on the privacy protection model to obtain desensitized biological information. The biological recognition processing corresponding to the biological recognition request is performed based on the desensitized biological information.

[0046] In implementation, the server can send the determined privacy protection model to the terminal device. The terminal device can receive the privacy protection model, and then input the biological information of the target user into the privacy protection model. The privacy protection model is used to perform privacy protection processing on the biological information of the target user, so that sensitive information (such as information of fingerprint lines, clear contour information of face, etc.) contained in the biological information of the target user is removed or hidden, thereby desensitizing the biological information to obtain desensitized biological information. The desensitized biological information can be compared with the pre-stored reference biological information to obtain the identification result of the target user. The identification result can be presented to the target user. If the identification result indicates that the target user passes the identity verification, the target user can continue to perform the corresponding business. If the identification result indicates that the target user fails the identity verification, the target user can be sent a corresponding notification message. The notification message can notify the target user that the identity verification of the target user fails. At this time, the terminal device can refuse to perform the corresponding business.

[0047] The embodiment of the present specification provides a privacy protection method of biometric identification. The method comprises the following steps: obtaining a biometric identification request, wherein the biometric identification request carries biometric information of a target user; inputting the biometric information into a biological attribute model obtained through model training in advance; extracting user attribute information of the target user from the biometric information through the biological attribute model to obtain user attribute information corresponding to the biometric information; sending the user attribute information corresponding to the biometric information to a server to trigger the server to determine a privacy protection model for desensitizing the biometric information based on the distribution of the user attribute information of the target user; receiving the privacy protection model issued by the server and desensitizing the biometric information based on the privacy protection model to obtain desensitized biometric information; and performing biometric identification processing corresponding to the biometric identification request based on the desensitized biometric information. In this way, in order to balance the privacy protection deployment efficiency and security capability of biometric identification, a privacy protection cold start method of biometric identification is proposed, so as to decouple each new scene (or data domain) into a combination of some user attribute information. Then, the privacy protection model of the distribution of different user attribute information is trained in advance on the server side to adapt to different new scenes. After deployment to the new scene, the user attribute information is analyzed by collecting data on the client side to obtain the most suitable privacy protection model, which is directly adapted by the server, thereby avoiding long-period preparation work for the new scene and achieving rapid adaptation to the new scene.

[0048] Embodiment two

[0049] As Figure 3 shown, the embodiment of the present specification provides a privacy protection method of biometric identification. The execution subject of the method can be a terminal device, wherein the terminal device can be a certain terminal device such as a mobile phone, a tablet computer, etc., can also be a computer device such as a notebook computer or a desktop computer, or can also be an IoT device (specifically, such as a smart watch, a vehicle-mounted device, etc.), and can also be a machine tool with a certain function set in advance, etc. The method can specifically comprise the following steps:

[0050] In step S302, a biometric identification request is obtained, wherein the biometric identification request carries biometric information of a target user.

[0051] The biometric information can include face information, and the face information can be determined based on a face image.

[0052] In step S304, the above biometric information is input into a biological attribute model obtained through model training in advance, and user attribute information of the target user is extracted from the biometric information through the biological attribute model to obtain user attribute information corresponding to the biometric information.

[0053] The user attribute information can include attribute items and attribute values, and the attribute items include one or more of exaggeration degree of expression, resolution of face, gender, age, blur degree, etc. The specific setting can be based on actual conditions, and the embodiments of the present specification are not limited thereto. The biological attribute model can be a MobileNet as a backbone network, and a plurality of classifiers are arranged after the feature. The number of classifiers can be determined based on the number of attribute items included in the user attribute information.

[0054] In implementation, since the biological attribute model needs to be run on the terminal device side, MobileNetV2x0.25 can be used as a backbone (i.e., a backbone network), and a plurality of classifiers are connected after the last feature. If the number of attribute items included in the user attribute information is 5, i.e., the exaggeration degree of expression, the resolution of face, the gender, the age, and the blur degree, 5 classifiers can be connected after the last feature, corresponding to 5 attribute items, i.e., {exaggeration degree of expression, resolution of face, gender, age, blur degree}.

[0055] In step S306, the user attribute information corresponding to the biological information is encrypted based on a preset encryption key to obtain encrypted user attribute information.

[0056] In step S308, the encrypted user attribute information is sent to the server. The encrypted user attribute information is used to trigger the server to perform decryption processing on the encrypted user attribute information to obtain the user attribute information, and based on the distribution of the user attribute information of the target user, a privacy protection model for desensitizing the biological information is determined.

[0057] The privacy protection model can be constructed by a convolutional neural network model including a preset number of convolutional layers, for example, a convolutional neural network model including 3 convolutional layers, or a convolutional neural network model including 5 convolutional layers, etc. Each convolutional layer can include a 3*3 convolutional kernel, or a 4*4 convolutional kernel, etc. The specific setting can be based on actual conditions, and the embodiments of the present specification are not limited thereto.

[0058] In step S310, the privacy protection model issued by the server is received, and the biological information is desensitized based on the privacy protection model to obtain desensitized biological information, and the biological recognition processing corresponding to the biological recognition request is performed based on the desensitized biological information.

[0059] The embodiment of the present specification provides a privacy protection method of biometric identification. The method comprises the following steps: obtaining a biometric identification request, the biometric identification request carrying biometric information of a target user; inputting the biometric information into a biological attribute model obtained through model training in advance; extracting user attribute information of the target user from the biometric information through the biological attribute model to obtain user attribute information corresponding to the biometric information; sending the user attribute information corresponding to the biometric information to a server to trigger the server to determine a privacy protection model for desensitizing the biometric information based on the distribution of the user attribute information of the target user; receiving the privacy protection model issued by the server and desensitizing the biometric information based on the privacy protection model to obtain desensitized biometric information; and performing biometric identification processing corresponding to the biometric identification request based on the desensitized biometric information. In this way, in order to balance the privacy protection deployment efficiency and security capability of biometric identification, a privacy protection cold start method of biometric identification is proposed, so as to decouple each new scene (or data domain) into a combination of some user attribute information. Then, the privacy protection model of the distribution of different user attribute information is trained in advance on the server side to adapt to different new scenes. After deployment to the new scene, the user attribute information is analyzed by collecting data on the client side to obtain the most suitable privacy protection model, and the server issues the model directly for adaptation, thereby avoiding long-period preparation work for the new scene and achieving rapid adaptation to the new scene.

[0060] Embodiment three

[0061] As shown in Figure 4A and Figure 4B The embodiment of the present specification provides a privacy protection method of biometric identification. The method can be executed by a server. The server can be an independent server or a server cluster composed of multiple servers. The server can be a background server of a financial service or an online shopping service, or a background server of an application program.

[0062] In step S402, the user attribute information corresponding to the biometric information of the target user sent by the terminal device is received. The biometric information of the target user is obtained when the terminal device obtains the biometric identification request of the target user. The user attribute information is obtained by inputting the biometric information into a pre-trained biological attribute model and extracting information from the biometric information through the biological attribute model.

[0063] In step S404, a privacy protection model for desensitizing the biometric information is determined based on the distribution of the user attribute information of the target user.

[0064] In step S406, the determined privacy protection model is issued to the terminal device, and the privacy protection model is used to trigger the terminal device to perform desensitization processing on the biological information to obtain desensitized biological information, and perform biological recognition processing corresponding to the biological recognition request based on the desensitized biological information.

[0065] The embodiment of the present specification provides a privacy protection method for biological recognition. The biological recognition request is obtained, and the biological information of the target user is carried in the biological recognition request. The biological information is input into the biological attribute model obtained through model training in advance. The user attribute information of the target user is extracted from the biological information by the biological attribute model, and the user attribute information corresponding to the biological information is obtained. The user attribute information corresponding to the biological information is sent to the server to trigger the server to determine the privacy protection model for desensitization processing of the biological information based on the distribution of the user attribute information of the target user. The privacy protection model issued by the server is received, and the biological information is desensitized based on the privacy protection model. The biological recognition processing corresponding to the biological recognition request is performed based on the desensitized biological information. In this way, in order to balance the privacy protection deployment efficiency and security capability of biological recognition, a privacy protection cold start method for biological recognition is proposed, so as to decouple each new scene (or data domain) into a combination of some user attribute information. Then, the privacy protection model of the distribution of different user attribute information is trained in advance on the server side to adapt to different new scenes. After deployment to the new scene, the user attribute information is analyzed by collecting data on the client side to obtain the most suitable privacy protection model, and the server is directly adapted by the server to issue, thereby avoiding long-period preparation work for the new scene, and achieving rapid adaptation to the new scene.

[0066] Embodiment four

[0067] As shown in Figure 5 , the embodiment of the present application provides a privacy protection method for biological recognition. The execution subject of the method can be a server. The server can be an independent server, or a server cluster composed of multiple servers, etc. The server can be a background server of a financial service or a network shopping service, etc., or a background server of an application program, etc. The system architecture can be as shown in Figure 2 , and the method can specifically include the following steps:

[0068] In step S502, biological information samples of multiple different users are obtained.

[0069] The biological information sample can be a sample data composed of facial information.

[0070] In implementation, factors affecting privacy protection effect can be abstracted, taking face recognition as an example, mainly including the following multiple attribute items: {exaggeration degree of expression, resolution of face, gender, age, blur degree}, and the following model training method also takes face recognition as an example, and other biometric recognition systems can be analogized.

[0071] In step S504, the biological attribute model is trained based on the biological information sample and the preset first loss function, and a trained biological attribute model is obtained. The biological attribute model is used to extract sample attribute information of the user from the biological information sample, and the sample attribute information includes attribute items and attribute values.

[0072] In implementation, for the model structure of the biological attribute model: since it needs to run on the terminal device side, MobileNetV2x0.25 can be used as the backbone, and four classifiers are connected after the last feature, corresponding to four attribute items, i.e., exaggeration degree of expression, gender, age, and blur degree. The resolution of the face can be directly calculated and does not need to be predicted using the biological attribute model.

[0073] The input information of the biological attribute model can be a face image carrying face information, and the output information is the attribute values of the four attribute items of the face information. The first loss function can be determined according to the input information and the output information of the biological attribute model, and then the biological information sample and the preset first loss function can be used to train the biological attribute model, and a trained biological attribute model can be obtained. The trained biological attribute model can be used to extract sample attribute information of the user from the biological information sample.

[0074] In actual application, the first loss function can include multiple types, in addition to the above-mentioned construction, it can also include multiple construction methods, and the following provides an optional construction method, which can include the following contents: the first loss function can be determined based on the loss function corresponding to the preset classification algorithm, and the classification algorithm includes a binary classification algorithm and / or a multi-classification algorithm.

[0075] In implementation, for the attribute items {exaggeration degree of expression, gender, blur degree}, a loss function based on a binary classification algorithm can be used, for the above-mentioned attribute items, the loss function based on the binary classification algorithm can be used as the first loss function, and for the attribute item age, a loss function based on a multi-classification algorithm can be used, and in actual application, a class can be set every 5 years, etc.

[0076] In step S506, the trained biological attribute model is distributed to the terminal device.

[0077] In step S508, a plurality of user attribute information of different users is obtained to form an attribute sample set.

[0078] In step S510, based on the attribute sample set, an attribute value corresponding to each attribute item is determined to obtain a distribution of user attribute information corresponding to each attribute item.

[0079] In implementation, for the above-mentioned five different attribute items, a plurality of common distributions and proportions are selected from the attribute sample set of the plurality of user attribute information of different users obtained, for example, for gender, it is assumed that the five common distributions include {male accounts for 50%, female accounts for 50%; male accounts for 10%, female accounts for 90%; male accounts for 90%, female accounts for 10%; male accounts for 25%, female accounts for 75%; male accounts for 75%, female accounts for 25%}, through traversal matching, five distributions of user attribute information can be obtained, i.e. 3125 distributions are included, sample data sampling is performed on the above-mentioned 3125 distributions, the corresponding attribute distribution is fitted as much as possible, and 3125 training data sets are obtained. 5

[0080] In step S512, each attribute item and the distribution of user attribute information corresponding to each attribute item are taken as training samples, and the privacy protection model is trained through the training samples and a preset second loss function to obtain a privacy protection model corresponding to the distribution of user attribute information corresponding to each attribute item.

[0081] In implementation, for the model structure of the privacy protection model: the privacy protection model includes a convolutional network model of three network layers, each network layer includes a 3*3 convolution kernel. The input data of the privacy protection model can be biological information, and the output data of the privacy protection model is desensitized biological information. The second loss function can be determined according to the input information and the output information of the privacy protection model, for any attribute item, the attribute item and the distribution of user attribute information corresponding to the attribute item are taken as training samples, the privacy protection model is trained through the training samples and the above-mentioned second loss function to obtain a privacy protection model corresponding to the distribution of user attribute information corresponding to the attribute item, and the distribution of user attribute information and the obtained privacy protection model corresponding to the attribute item can be stored. The same method can be used to determine a privacy protection model corresponding to the distribution of user attribute information corresponding to another attribute item, and so on, a privacy protection model corresponding to the distribution of user attribute information corresponding to each attribute item can be obtained, and for the above-mentioned 3125 training data sets, the above-mentioned operation can be repeated to obtain 3125 privacy protection models.

[0082] ​The second loss function can include multiple types, in addition to being constructed in the above manner, and can include multiple construction methods. An alternative construction method is provided below, which can include the following: the second loss function is determined based on an inverse Euclidean distance function.

[0083] In implementation, the second loss function Loss

[0084]

[0085] Wherein, f represents the input information of the privacy protection model, and f' represents the output information of the privacy protection model.

[0086] In step S514, the user attribute information corresponding to the biological information of the target user sent by the terminal device is received. The biological information of the target user is the information obtained by the terminal device when the target user requests biological recognition. The user attribute information is the information extracted from the biological information by inputting the biological information into the pre-trained biological attribute model.

[0087] In implementation, for new scenarios, the terminal device such as a machine tool can be deployed in the corresponding scene for trial operation. During operation, the terminal device can send the user attribute information corresponding to the biological information of the target user.

[0088] Wherein, the user attribute information can be encrypted information, based on which the following steps can be executed.

[0089] In step S516, when the preset period is reached, the user attribute information of the target user received in the period is obtained, and the user attribute information is decrypted to obtain decrypted user attribute information.

[0090] Wherein, the preset period can be 24 hours or 48 hours, etc. The specific period can be set according to actual conditions, and the embodiments of the present application do not limit the period.

[0091] In implementation, in the preset period, privacy protection can be achieved by encryption. Since the preset period is relatively short, it can meet the security requirements. In the preset period, for each face recognition, the attribute values of multiple attribute items can be calculated. When the preset period is reached, the user attribute information of the target user received in the period is obtained, and the user attribute information is decrypted to obtain decrypted user attribute information.

[0092] In step S518, based on the decrypted user attribute information in the above period, the distribution of the user attribute information of the target user is determined.

[0093] In implementation, based on the decrypted user attribute information in the above period, the distribution statistics of all attribute items and corresponding attribute values in the trial operation stage can be performed to obtain the distribution of the user attribute information of the corresponding target user.

[0094] In step S520, based on the distribution of the user attribute information of the target user, a privacy protection model for desensitizing the above biological information is determined.

[0095] In implementation, based on a terminal device such as a machine tool in a new scene, the distribution of the user attribute information corresponding to multiple attribute items can be calculated in the above manner, and compared with the training data distribution of the above 3125 privacy protection models to find the closest distribution of the user attribute information, and the privacy protection model corresponding to the closest distribution of the user attribute information is the model that can adapt to the new scene.

[0096] In step S522, the determined privacy protection model is issued to the terminal device, and the privacy protection model is used to trigger the terminal device to desensitize the above biological information to obtain desensitized biological information, and perform the biological recognition processing corresponding to the above biological recognition request based on the desensitized biological information.

[0097] The embodiment of the present specification provides a privacy protection method for biological recognition, by obtaining a biological recognition request, the biological recognition request carrying biological information of a target user, inputting the biological information into a biological attribute model obtained by model training in advance, extracting user attribute information of the target user from the biological information through the biological attribute model, obtaining user attribute information corresponding to the biological information, sending the user attribute information corresponding to the biological information to a server, triggering the server to determine a privacy protection model for desensitizing the biological information based on the distribution of the user attribute information of the target user, receiving the privacy protection model issued by the server, and desensitizing the biological information based on the privacy protection model to obtain desensitized biological information, and performing biological recognition processing corresponding to the biological recognition request based on the desensitized biological information. In this way, in order to balance the privacy protection deployment efficiency and security capability of biological recognition, a privacy protection cold start method for biological recognition is proposed, so that each new scene (or data domain) is decoupled into a combination of some user attribute information, and then the privacy protection model of the distribution of different user attribute information is trained in advance on the server side to adapt to different new scenes. After deployment to the new scene, the most suitable privacy protection model is obtained by collecting data on the client side for user attribute information analysis, and is directly adapted by the server, thereby avoiding long-period preparation work for the new scene, and achieving rapid adaptation to the new scene.

[0098] Embodiment five

[0099] Based on the same idea, the embodiment of the present specification also provides a privacy protection device for biometric identification, as shown in Figure 6

[0100] The privacy protection device for biometric identification comprises a request acquisition module 601, an attribute extraction module 602, an attribute information sending module 603 and a desensitization module 604, wherein:

[0101] The request acquisition module 601 acquires a biometric identification request, wherein the biometric identification request carries biometric information of a target user;

[0102] The attribute extraction module 602 inputs the biometric information into a biometric attribute model obtained through model training in advance, extracts user attribute information of the target user from the biometric information through the biometric attribute model, and obtains user attribute information corresponding to the biometric information;

[0103] The attribute information sending module 603 sends the user attribute information corresponding to the biometric information to a server, and the user attribute information is used to trigger the server to determine a privacy protection model for desensitizing the biometric information based on a distribution of the user attribute information of the target user;

[0104] The desensitization module 604 receives the privacy protection model issued by the server, and desensitizes the biometric information based on the privacy protection model to obtain desensitized biometric information, and performs biometric identification processing corresponding to the biometric identification request based on the desensitized biometric information.

[0105] In the embodiment of the present specification, the biometric information comprises face information, the face information is determined based on a face image, the user attribute information comprises an attribute item and an attribute value, and the attribute item comprises one or more of the following: exaggeration degree of expression, resolution of face, gender, age, and blur degree.

[0106] In the embodiment of the present specification, the biometric attribute model uses MobileNet as a backbone network, and a plurality of classifiers are arranged after feature, and the number of the classifiers is determined based on the number of attribute items contained in the user attribute information.

[0107] In the embodiment of the present specification, the privacy protection model is constructed by a convolutional neural network model comprising a preset number of convolutional layers.

[0108] In the embodiment of the present specification, the preset number of convolutional layers is three convolutional layers, and each convolutional layer comprises a 3*3 convolutional kernel.

[0109] In the embodiment of the present specification, the attribute information sending module 603 comprises: ​

[0110] The encryption unit performs encryption processing on the user attribute information corresponding to the biological information based on a preset encryption key, to obtain encrypted user attribute information.

[0111] The attribute information sending unit sends the encrypted user attribute information to the server, and the encrypted user attribute information is used to trigger the server to perform decryption processing on the encrypted user attribute information to obtain the user attribute information, and determine a privacy protection model for desensitizing processing of the biological information based on the distribution of the user attribute information of the target user.

[0112] The embodiments of the present specification provide a privacy protection device for biological identification. The biological identification request is obtained, and the biological information of the target user is carried in the biological identification request. The biological attribute model obtained through model training in advance is input into the biological attribute model. The user attribute information of the target user is extracted from the biological information by the biological attribute model, and the user attribute information corresponding to the biological information is obtained. The user attribute information corresponding to the biological information is sent to the server to trigger the server to determine the privacy protection model for desensitizing processing of the biological information based on the distribution of the user attribute information of the target user. The privacy protection model issued by the server is received, and the biological information is desensitized based on the privacy protection model to obtain desensitized biological information. The biological identification processing corresponding to the biological identification request is performed based on the desensitized biological information. In this way, in order to balance the privacy protection deployment efficiency and security capability of biological identification, a privacy protection cold start mode for biological identification is proposed, so as to decouple each new scene (or data domain) into a combination of some user attribute information. Then, the privacy protection model of the distribution of different user attribute information is trained in advance on the server side to adapt to different new scenes. After deployment to the new scene, the user attribute information is analyzed by collecting data on the client side to obtain the most suitable privacy protection model, and the server issues a direct adaptation, thereby avoiding long-period preparation work for the new scene, and achieving rapid adaptation to the new scene.

[0113] Embodiment six

[0114] Based on the same idea, the embodiments of the present specification also provide a privacy protection device for biological identification, as shown in Figure 7 .

[0115] The privacy protection device for biological identification includes an attribute information receiving module 701, a model determining module 702, and a model issuing module 703, wherein:

[0116] The attribute information receiving module 701 receives user attribute information corresponding to biological information of a target user sent by a terminal device, the biological information of the target user being information obtained by the terminal device when obtaining a biological recognition request of the target user, and the user attribute information being information extracted from the biological information by inputting the biological information into a pre-trained biological attribute model;

[0117] The model determining module 702 determines a privacy protection model for desensitizing the biological information based on a distribution of the user attribute information of the target user.

[0118] The model issuing module 703 issues the determined privacy protection model to the terminal device, and the privacy protection model is used to trigger the terminal device to desensitize the biological information to obtain desensitized biological information, and perform biological recognition processing corresponding to the biological recognition request based on the desensitized biological information.

[0119] In the embodiments of the present specification, the apparatus further comprises:

[0120] The first sample obtaining module obtains biological information samples of a plurality of different users.

[0121] The first training module performs model training on a biological attribute model based on the biological information samples and a preset first loss function to obtain a trained biological attribute model, and the biological attribute model is used to extract sample attribute information of a user from the biological information samples, and the sample attribute information includes an attribute item and an attribute value.

[0122] The first model issuing module issues the trained biological attribute model to the terminal device.

[0123] In the embodiments of the present specification, the first loss function is determined based on a loss function corresponding to a preset classification algorithm, and the classification algorithm includes a binary classification algorithm and / or a multi-classification algorithm.

[0124] In the embodiments of the present specification, the apparatus further comprises:

[0125] The second sample obtaining module obtains an attribute sample set composed of user attribute information of a plurality of different users.

[0126] The information distribution determining module determines an attribute value corresponding to each attribute item based on the attribute sample set to obtain a distribution of user attribute information corresponding to each attribute item.

[0127] The second training module takes each attribute item and the distribution of the user attribute information corresponding to each attribute item as a training sample, and respectively trains the privacy protection model based on the training sample and a preset second loss function, to obtain a privacy protection model corresponding to the distribution of the user attribute information corresponding to each attribute item.

[0128] In an embodiment of the present specification, the second loss function is determined based on an inverse Euclidean distance function.

[0129] In an embodiment of the present specification, the user attribute information is information processed by encryption, and the model determination module 702 comprises:

[0130] The decryption unit obtains the user attribute information of the target user received in a preset period when the period is reached, and decrypts the user attribute information to obtain decrypted user attribute information;

[0131] The distribution determination unit determines the distribution of the user attribute information of the target user based on the decrypted user attribute information in the period;

[0132] The model determination unit determines a privacy protection model for desensitizing the biological information based on the distribution of the user attribute information of the target user.

[0133] An embodiment of the present specification provides a privacy protection device for biological identification. By obtaining a biological identification request, the biological identification request carrying biological information of a target user, inputting the biological information into a biological attribute model obtained by pre-model training, extracting user attribute information of the target user from the biological information through the biological attribute model, obtaining user attribute information corresponding to the biological information, sending the user attribute information corresponding to the biological information to a server to trigger the server to determine a privacy protection model for desensitizing the biological information based on the distribution of the user attribute information of the target user, receiving the privacy protection model issued by the server, and desensitizing the biological information based on the privacy protection model to obtain desensitized biological information, and performing biological identification processing corresponding to the biological identification request based on the desensitized biological information. In this way, in order to balance the privacy protection deployment efficiency and security capability of biological identification, a privacy protection cold start method for biological identification is proposed, so that each new scene (or data domain) is decoupled into a combination of some user attribute information. Then, the privacy protection model of the distribution of different user attribute information is trained in advance on the server side to adapt to different new scenes. After deployment to a new scene, the most suitable privacy protection model is obtained by analyzing the user attribute information through data collection on the client side, and is directly adapted by the server, thereby avoiding long-period preparation work for new scenes and achieving rapid adaptation to new scenes.

[0134] Embodiment Seven

[0135] The privacy protection device for biometric recognition provided in the above embodiments of the present specification is based on the same idea. The present specification also provides a privacy protection device for biometric recognition, as shown in Figure 8

[0136] The privacy protection device for biometric recognition can be a terminal device or a server, etc. provided in the above embodiments.

[0137] The privacy protection device for biometric recognition can vary greatly due to different configurations or performances. It can include one or more processors 801 and memories 802. The memories 802 can store one or more applications or data. The memories 802 can be temporary or persistent memories. The applications stored in the memories 802 can include one or more modules (not shown in the figure). Each module can include a series of computer-executable instructions for the privacy protection device for biometric recognition. Further, the processor 801 can be configured to communicate with the memories 802 and execute the series of computer-executable instructions in the memories 802 on the privacy protection device for biometric recognition. The privacy protection device for biometric recognition can also include one or more power supplies 803, one or more wired or wireless network interfaces 804, one or more input / output interfaces 805, and one or more keyboards 806.

[0138] In particular, in the present embodiment, the privacy protection device for biometric recognition includes a memory and one or more programs. One or more programs are stored in the memory, and the one or more programs can include one or more modules. Each module can include a series of computer-executable instructions for the privacy protection device for biometric recognition. The one or more processors are configured to execute the one or more programs, which include the following computer-executable instructions:

[0139] Obtaining a biometric recognition request carrying biometric information of a target user;

[0140] Inputting the biometric information into a biometric attribute model obtained through pre-model training, extracting user attribute information of the target user from the biometric information through the biometric attribute model, and obtaining user attribute information corresponding to the biometric information;

[0141] Sending the user attribute information corresponding to the biometric information to a server. The user attribute information is used to trigger the server to determine a privacy protection model for desensitizing the biometric information based on a distribution of the user attribute information of the target user;

[0142] ​receive the privacy protection model issued by the server, and perform desensitization processing on the biological information based on the privacy protection model to obtain desensitized biological information, and perform biological recognition processing corresponding to the biological recognition request based on the desensitized biological information.

[0143] In an embodiment of the present specification, the biological information includes face information, the face information is determined based on a face image, the user attribute information includes an attribute item and an attribute value, and the attribute item includes one or more of the following: exaggeration degree of expression, resolution of face, gender, age, and blur degree.

[0144] In an embodiment of the present specification, the biological attribute model is constructed by taking MobileNet as a backbone network and setting a plurality of classifiers after feature.

[0145] In an embodiment of the present specification, the privacy protection model is constructed by a convolutional neural network model including a preset number of convolutional layers.

[0146] In an embodiment of the present specification, the preset number of convolutional layers is three convolutional layers, and each convolutional layer includes a 3*3 convolutional kernel.

[0147] In an embodiment of the present specification, the sending of the user attribute information corresponding to the biological information to the server includes:

[0148] performing encryption processing on the user attribute information corresponding to the biological information based on a preset encryption key to obtain encrypted user attribute information;

[0149] sending the encrypted user attribute information to the server, the encrypted user attribute information being used to trigger the server to perform decryption processing on the encrypted user attribute information to obtain the user attribute information, and determine a privacy protection model for desensitization processing on the biological information based on the distribution of the user attribute information of the target user.

[0150] In particular, in the present embodiment, the privacy protection device for biological recognition includes a memory and one or more programs, wherein one or more programs are stored in the memory, and the one or more programs can include one or more modules, and each module can include a series of computer executable instructions in the privacy protection device for biological recognition, and is configured to be executed by one or more processors. The one or more programs include computer executable instructions for:

[0151] receive user attribute information corresponding to biological information of a target user sent by a terminal device, the biological information of the target user being information acquired by the terminal device when acquiring a biological recognition request of the target user, the user attribute information being information extracted from the biological information by inputting the biological information into a pre-trained biological attribute model;

[0152] determine a privacy protection model for desensitizing the biological information based on a distribution of the user attribute information of the target user;

[0153] issue the determined privacy protection model to the terminal device, the privacy protection model being used to trigger the terminal device to desensitize the biological information to obtain desensitized biological information and perform biological recognition processing corresponding to the biological recognition request based on the desensitized biological information.

[0154] In the embodiments of the present specification, the following are further included:

[0155] acquire biological information samples of a plurality of different users;

[0156] perform model training on a biological attribute model based on the biological information samples and a preset first loss function to obtain a trained biological attribute model, the biological attribute model being used to extract sample attribute information of a user from the biological information samples, the sample attribute information including an attribute item and an attribute value;

[0157] issue the trained biological attribute model to the terminal device.

[0158] In the embodiments of the present specification, the first loss function is determined based on a loss function corresponding to a preset classification algorithm, and the classification algorithm includes a binary classification algorithm and / or a multi-classification algorithm.

[0159] In the embodiments of the present specification, the following are further included:

[0160] acquire an attribute sample set composed of user attribute information of a plurality of different users;

[0161] determine an attribute value corresponding to each attribute item based on the attribute sample set to obtain a distribution of user attribute information corresponding to each attribute item;

[0162] use each attribute item and the distribution of user attribute information corresponding to each attribute item as a training sample, and perform model training on the privacy protection model respectively through the training sample and a preset second loss function to obtain a privacy protection model corresponding to the distribution of user attribute information corresponding to each attribute item.

[0163] In the embodiments of the present specification, the second loss function is determined based on an inverse Euclidean distance function.

[0164] In the embodiments of the present specification, the user attribute information is information processed by encryption, and the distribution of the user attribute information of the target user is used to determine a privacy protection model for desensitizing the biological information, which comprises:

[0165] When a preset period is reached, the user attribute information of the target user received in the period is obtained, and the user attribute information is decrypted to obtain decrypted user attribute information;

[0166] Based on the decrypted user attribute information in the period, the distribution of the user attribute information of the target user is determined;

[0167] Based on the distribution of the user attribute information of the target user, a privacy protection model for desensitizing the biological information is determined.

[0168] The embodiments of the present specification provide a privacy protection device for biological identification. By obtaining a biological identification request, the biological identification request carrying biological information of a target user, inputting the biological information into a biological attribute model obtained by model training in advance, extracting user attribute information of the target user from the biological information through the biological attribute model, obtaining user attribute information corresponding to the biological information, sending the user attribute information corresponding to the biological information to a server to trigger the server to determine a privacy protection model for desensitizing the biological information based on the distribution of the user attribute information of the target user, receiving the privacy protection model issued by the server, and desensitizing the biological information based on the privacy protection model to obtain desensitized biological information, and performing biological identification processing corresponding to the biological identification request based on the desensitized biological information. In this way, in order to balance the privacy protection deployment efficiency and security capability of biological identification, a privacy protection cold start method for biological identification is proposed, so as to decouple each new scene (or data domain) into a combination of some user attribute information, and then train the privacy protection model of the distribution of different user attribute information in advance on the server side to adapt to different new scenes. After deployment to a new scene, the most suitable privacy protection model is obtained by analyzing the user attribute information through the client-side data collection, and is directly adapted by the server, thereby avoiding long-period preparation work for new scenes, and achieving rapid adaptation to new scenes.

[0169] Embodiment Eight

[0170] Further, based on the above Figures 1A to 5The method shown, one or more embodiments of the present specification also provide a storage medium for storing computer executable instruction information, in a specific embodiment, the storage medium can be a U disk, a CD, a hard disk, etc., the computer executable instruction information stored in the storage medium can realize the following process when executed by the processor:

[0171] Obtaining a biometric identification request, the biometric identification request carrying biometric information of a target user;

[0172] Input the biometric information into a biological attribute model obtained by model training in advance, extract the user attribute information of the target user from the biometric information through the biological attribute model, and obtain the user attribute information corresponding to the biometric information;

[0173] Send the user attribute information corresponding to the biometric information to the server, and the user attribute information is used to trigger the server to determine a privacy protection model for desensitizing the biometric information based on the distribution of the user attribute information of the target user;

[0174] Receive the privacy protection model issued by the server, and perform desensitization processing on the biometric information based on the privacy protection model to obtain desensitized biometric information, and perform biometric identification processing corresponding to the biometric identification request based on the desensitized biometric information.

[0175] In an embodiment of the present specification, the biometric information includes face information, the face information is determined based on a face image, the user attribute information includes an attribute item and an attribute value, and the attribute item includes one or more of the following: exaggeration degree of expression, resolution of face, gender, age, and blur degree.

[0176] In an embodiment of the present specification, the biological attribute model uses MobileNet as the backbone network, and sets multiple classifiers after feature.

[0177] In an embodiment of the present specification, the privacy protection model is constructed by a convolutional neural network model including a preset number of convolutional layers.

[0178] In an embodiment of the present specification, the preset number of convolutional layers is 3 convolutional layers, and each convolutional layer includes a 3*3 convolutional kernel.

[0179] In an embodiment of the present specification, the sending of the user attribute information corresponding to the biometric information to the server comprises:

[0180] Encrypt the user attribute information corresponding to the biometric information based on a preset encryption key to obtain encrypted user attribute information.

[0181] send the encrypted user attribute information to the server, the encrypted user attribute information being used to trigger the server to perform decryption processing on the encrypted user attribute information to obtain the user attribute information, and determine a privacy protection model for desensitizing the biological information based on a distribution of the user attribute information of the target user.

[0182] In addition, in another specific embodiment, the storage medium can be a U disk, an optical disk, a hard disk, etc., and the computer executable instruction information stored in the storage medium can realize the following flow when executed by the processor:

[0183] receive user attribute information corresponding to biological information of a target user sent by a terminal device, the biological information of the target user being information obtained by the terminal device when obtaining a biological recognition request of the target user, and the user attribute information being information extracted from the biological information by a biological attribute model by inputting the biological information into the biological attribute model in advance;

[0184] determine a privacy protection model for desensitizing the biological information based on a distribution of the user attribute information of the target user;

[0185] distribute the determined privacy protection model to the terminal device, the privacy protection model being used to trigger the terminal device to perform desensitization processing on the biological information to obtain desensitized biological information, and perform biological recognition processing corresponding to the biological recognition request based on the desensitized biological information.

[0186] In the embodiments of the present specification, the following are further included:

[0187] obtain biological information samples of a plurality of different users;

[0188] perform model training on a biological attribute model based on the biological information samples and a preset first loss function to obtain a trained biological attribute model, the biological attribute model being used to extract sample attribute information of a user from the biological information samples, and the sample attribute information including an attribute item and an attribute value;

[0189] distribute the trained biological attribute model to the terminal device.

[0190] In the embodiments of the present specification, the first loss function is determined based on a loss function corresponding to a preset classification algorithm, and the classification algorithm includes a binary classification algorithm and / or a multi-classification algorithm.

[0191] In the embodiments of the present specification, the following are further included:

[0192] obtain an attribute sample set composed of user attribute information of a plurality of different users;

[0193] determine an attribute value corresponding to each attribute item based on the attribute sample set, to obtain a distribution of user attribute information corresponding to each attribute item;

[0194] take each attribute item and the distribution of user attribute information corresponding to each attribute item as a training sample, respectively train the privacy protection model based on the training sample and a preset second loss function, and obtain a privacy protection model corresponding to the distribution of user attribute information corresponding to each attribute item.

[0195] In the embodiments of the present specification, the second loss function is determined based on an inverse Euclidean distance function.

[0196] In the embodiments of the present specification, the user attribute information is information processed by encryption, and determining a privacy protection model for desensitizing the biological information based on the distribution of user attribute information of the target user comprises:

[0197] When a preset period is reached, obtaining user attribute information of the target user received in the period, and performing decryption processing on the user attribute information to obtain decrypted user attribute information;

[0198] determining a distribution of user attribute information of the target user based on the decrypted user attribute information in the period;

[0199] determining a privacy protection model for desensitizing the biological information based on the distribution of user attribute information of the target user.

[0200] The embodiment of the specification provides a storage medium, by acquiring a biological recognition request, the biological recognition request carrying biological information of a target user, inputting the biological information into a biological attribute model obtained through model training in advance, extracting user attribute information of the target user from the biological information through the biological attribute model, obtaining user attribute information corresponding to the biological information, sending the user attribute information corresponding to the biological information to a server, triggering the server to determine a privacy protection model for desensitizing processing of the biological information based on the distribution of the user attribute information of the target user, receiving the privacy protection model issued by the server, and performing desensitizing processing on the biological information based on the privacy protection model, obtaining desensitized biological information, and performing biological recognition processing corresponding to the biological recognition request based on the desensitized biological information. In this way, in order to balance the privacy protection deployment efficiency and security capability of biological recognition, a privacy protection cold start mode of biological recognition is proposed, so as to decouple each new scene (or data domain) into a combination of some user attribute information, then the privacy protection model of the distribution of different user attribute information is trained in advance on the server side to adapt to different new scenes, after deployment to the new scene, the user attribute information is analyzed through the client to collect data to obtain the most suitable privacy protection model, and the server is directly issued to adapt, thereby avoiding the long period of preparation work for the new scene, and achieving rapid adaptation to the new scene.

[0201] The above describes specific embodiments of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different than the order in the embodiments and still achieve the desired result. In addition, the processes depicted in the figures do not necessarily require the particular order shown or sequential order to achieve the desired results. In certain implementations, multitasking and parallel processing can be advantageous or possible.

[0202] In the 1990s, it was possible to distinguish whether an improvement in a technology was a hardware improvement (e.g., an improvement in the circuit structure of a diode, transistor, switch, etc.) or a software improvement (an improvement in a method flow). However, as technology has advanced, many improvements in method flows today can be considered as direct improvements in hardware circuit structures. Designers almost always obtain a corresponding hardware circuit structure by programming the improved method flow into a hardware circuit. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using a hardware entity module. For example, a programmable logic device (PLD) (e.g., a field programmable gate array (FPGA)) is an integrated circuit whose logic function is determined by user programming of the device. A digital system is "integrated" on a PLD by the designer programming it, rather than by asking a chip manufacturer to design and fabricate a custom integrated circuit chip. Moreover, instead of manually fabricating an integrated circuit chip, this programming is now mostly implemented using "logic compiler" software, which is similar to software compilers used in program development, and the original code to be compiled is written in a specific programming language, which is called a hardware description language (HDL), and there are many such languages, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc., and the most commonly used are VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should be aware that, as long as the method flow is logically programmed using the above-mentioned hardware description languages and programmed into an integrated circuit, a hardware circuit implementing the logical method flow can be easily obtained.

[0203] The controller can be implemented in any suitable way, e.g. the controller can take the form of a microprocessor or processor and a computer readable medium storing computer readable program code, e.g. software or firmware, executable by the (micro)processor, logic gates, switches, an application specific integrated circuit (ASIC), a programmable logic controller and an embedded microcontroller, examples of controllers include but are not limited to the following microcontrollers: ARC 625D, Atmel AT91 SAM, Microchip PIC18F26K20 and Silicone Labs C8051F320, the memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art will also know that in addition to being implemented in pure computer readable program code form, the controller can perfectly well be implemented by means of logic programmed into logic gates, switches, application specific integrated circuits, programmable logic controllers and embedded microcontrollers, etc. to perform the same functions. The controller can thus be considered as a hardware component, and the means comprised therein for performing various functions can be considered as structures within the hardware component. Alternatively, or even, the means for performing various functions can be considered as both a software module implementing a method and a structure within a hardware component.

[0204] The systems, apparatuses, modules or units illustrated by the above embodiments can be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer can be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0205] For the sake of description, the above apparatuses are described in functional division and are described respectively. Of course, the functions of each unit can be implemented in the same or more software and / or hardware when implementing one or more embodiments of the present specification.

[0206] Those skilled in the art will understand that the embodiments of the present specification can be provided as a method, a system, or a computer program product. Therefore, one or more embodiments of the present specification can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of the present specification can take the form of a computer program product implemented on one or more computer usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer usable program code.

[0207] The embodiments of the present specification are described with reference to flowcharts and / or block diagrams of the method, device (system) and computer program product according to embodiments of the present specification. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as a combination of flows and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable electronic devices to produce a machine, so that the instructions executed by the processor of the computer or other programmable electronic devices generate a device that implements the functions specified in one or more flows in the flowcharts and / or one or more blocks in the block diagrams.

[0208] These computer program instructions can also be stored in a computer readable memory that can direct the computer or other programmable electronic devices to work in a specific manner, so that the instructions stored in the computer readable memory produce a product including instruction devices that implement the functions specified in one or more flows in the flowcharts and / or one or more blocks in the block diagrams.

[0209] These computer program instructions can also be loaded into a computer or other programmable electronic devices, so that a series of operation steps are performed on the computer or other programmable electronic devices to produce a computer implemented process, so that the instructions executed on the computer or other programmable electronic devices provide steps for implementing the functions specified in one or more flows in the flowcharts and / or one or more blocks in the block diagrams.

[0210] In a typical configuration, the computing device includes one or more processors (CPUs), input / output interfaces, network interfaces, and memory.

[0211] The memory can include non-persistent memory in the computer readable medium, random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM). The memory is an example of computer readable media.

[0212] Computer-readable media includes permanent and non-permanent, movable and non-movable media that can be implemented by any method or technology to store information. The information can be computer-readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassette, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information accessible to a computing device. According to the definition herein, computer-readable media does not include transitory media such as modulated data signals and carriers.

[0213] It should also be noted that the terms "comprising", "containing", or any other variant thereof are intended to cover non-exclusive inclusion, such that a process, method, article or apparatus that comprises a list of elements does not only include those elements, but can also include other elements not expressly listed or inherent to such process, method, article or apparatus. Without more limitations, the element defined by the statement "comprising a" does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0214] Those skilled in the art will appreciate that embodiments of the present specification can be provided as methods, systems or computer program products. Therefore, one or more embodiments of the present specification can take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Moreover, one or more embodiments of the present specification can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0215] One or more embodiments of the present specification can be described in the general context of computer-executable instructions being executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. One or more embodiments of the present specification can also be practiced in a distributed computing environment, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in both local and remote computer storage media, including storage devices.

[0216] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each of the embodiments focuses on the difference from other embodiments. In particular, for the system embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the part of the method embodiments.

[0217] The above only describes the embodiments of the specification and is not used to limit the application. The specification can have various changes and modifications for those skilled in the art. Any modification, equivalent replacement, improvement, etc. within the spirit and principle of the specification shall be included in the scope of claims of the specification.

Claims

1. A biometric privacy protection method, applied to a terminal device, the method comprising: Obtain a biometric request, wherein the biometric request carries a type of biometric information of the target user; The biological information is input into a biological attribute model that has been pre-trained, and the user attribute information of the target user is extracted from the biological information through the biological attribute model to obtain the user attribute information corresponding to the biological information. The user attribute information corresponding to the biometric information is sent to the server. The user attribute information is used to trigger the server to obtain the privacy protection model corresponding to the distribution of user attribute information of the target user from the stored correspondence between the distribution of user attribute information and the privacy protection model, based on the distribution of user attribute information corresponding to each attribute item of the target user. The obtained privacy protection model is determined as the privacy protection model for desensitizing the biometric information. The system receives the privacy protection model sent by the server, performs desensitization processing on the biometric information based on the privacy protection model to obtain desensitized biometric information, and performs biometric processing corresponding to the biometric request based on the desensitized biometric information.

2. The method according to claim 1, wherein the biometric information includes facial information, the facial information is determined based on a facial image, and the user attribute information includes attribute items and attribute values, wherein the attribute items include one or more of the following: degree of exaggeration of expression, facial resolution, gender, age, and degree of blur.

3. The method according to claim 2, wherein the biological attribute model uses MobileNet as the backbone network and sets multiple classifiers after the feature, the number of classifiers being determined based on the number of attribute items contained in the user attribute information.

4. The method according to claim 1, wherein the privacy protection model is constructed from a convolutional neural network model including a preset number of convolutional layers.

5. The method according to claim 4, wherein the preset number of convolutional layers is 3 convolutional layers, and each convolutional layer includes a 3*3 convolutional kernel.

6. The method according to claim 1, wherein sending the user attribute information corresponding to the biological information to the server comprises: The user attribute information corresponding to the biometric information is encrypted based on a preset encryption key to obtain encrypted user attribute information. The encrypted user attribute information is sent to the server, which triggers the server to decrypt the encrypted user attribute information to obtain the user attribute information. Based on the distribution of user attribute information corresponding to each attribute item of the target user, a privacy protection model for desensitizing the biometric information is determined.

7. A biometric privacy protection method applied to a server, the method comprising: The terminal device receives user attribute information corresponding to a type of biometric information of a target user. The biometric information of the target user is the information obtained by the terminal device when it receives the biometric recognition request of the target user. The user attribute information is the information extracted from the biometric information by inputting the biometric information into a pre-trained biometric attribute model. Based on the distribution of user attribute information corresponding to each attribute item of the target user, the privacy protection model corresponding to the distribution of user attribute information of the target user is obtained from the correspondence between the stored distribution of user attribute information and the privacy protection model, and the obtained privacy protection model is determined as the privacy protection model for desensitizing the biometric information. The determined privacy protection model is sent to the terminal device. The privacy protection model is used to trigger the terminal device to de-identify the biometric information to obtain the de-identified biometric information, and to perform biometric processing corresponding to the biometric recognition request based on the de-identified biometric information.

8. The method according to claim 7, further comprising: Obtain biometric samples from multiple different users; Based on the biological information samples and a preset first loss function, the biological attribute model is trained to obtain the trained biological attribute model. The biological attribute model is used to extract the user's sample attribute information from the biological information samples. The sample attribute information includes attribute items and attribute values. The trained biological attribute model is then distributed to the terminal device.

9. The method according to claim 8, wherein the first loss function is determined based on the loss function corresponding to a preset classification algorithm, and the classification algorithm includes a binary classification algorithm and / or a multi-classification algorithm.

10. The method according to claim 7, further comprising: Obtain an attribute sample set consisting of user attribute information from multiple different users; Based on the attribute sample set, the attribute value corresponding to each attribute item is determined, and the distribution of user attribute information corresponding to each attribute item is obtained. Each attribute item and the distribution of user attribute information corresponding to each attribute item are used as training samples. The privacy protection model is trained using the training samples and a preset second loss function to obtain a privacy protection model corresponding to the distribution of user attribute information corresponding to each attribute item.

11. The method according to claim 10, wherein the second loss function is determined based on the inverse Euclidean distance function.

12. The method according to claim 7, wherein the user attribute information is encrypted information, and the step of obtaining the privacy protection model corresponding to the distribution of user attribute information of the target user from the stored correspondence between the distribution of user attribute information and the privacy protection model based on the distribution of user attribute information corresponding to each attribute item of the target user, and determining the obtained privacy protection model as the privacy protection model for desensitizing the biometric information, includes: When the preset period is reached, the user attribute information of the target user received within the period is obtained, and the user attribute information is decrypted to obtain the decrypted user attribute information. Based on the decrypted user attribute information within the period, determine the distribution of user attribute information corresponding to each attribute item of the target user; Based on the distribution of user attribute information corresponding to each attribute item of the target user, the privacy protection model corresponding to the distribution of user attribute information of the target user is obtained from the correspondence between the stored distribution of user attribute information and the privacy protection model, and the obtained privacy protection model is determined as the privacy protection model for desensitizing the biometric information.

13. A biometric privacy protection device, the device comprising: The request acquisition module acquires a biometric request, wherein the biometric request carries a type of biometric information of the target user; The attribute extraction module inputs the biological information into a biological attribute model that has been pre-trained, and extracts the user attribute information of the target user from the biological information through the biological attribute model to obtain the user attribute information corresponding to the biological information. The attribute information sending module sends the user attribute information corresponding to the biometric information to the server. The user attribute information is used to trigger the server to obtain the privacy protection model corresponding to the distribution of user attribute information of the target user from the stored correspondence between the distribution of user attribute information and the privacy protection model, based on the distribution of user attribute information corresponding to each attribute item of the target user. The obtained privacy protection model is determined as the privacy protection model for desensitizing the biometric information. The desensitization module receives the privacy protection model sent by the server, and performs desensitization processing on the biometric information based on the privacy protection model to obtain desensitized biometric information. Based on the desensitized biometric information, it performs biometric processing corresponding to the biometric recognition request.

14. A biometric privacy protection device, the device comprising: The attribute information receiving module receives user attribute information corresponding to a type of biometric information of a target user sent by a terminal device. The biometric information of the target user is the information obtained by the terminal device when it receives the biometric recognition request of the target user. The user attribute information is the information extracted from the biometric information by inputting the biometric information into a pre-trained biometric attribute model. The model determination module, based on the distribution of user attribute information corresponding to each attribute item of the target user, obtains the privacy protection model corresponding to the distribution of user attribute information of the target user from the correspondence between the stored distribution of user attribute information and the privacy protection model, and determines the obtained privacy protection model as the privacy protection model for desensitizing the biometric information; The model distribution module distributes the determined privacy protection model to the terminal device. The privacy protection model is used to trigger the terminal device to perform desensitization processing on the biometric information to obtain desensitized biometric information, and to perform biometric processing corresponding to the biometric recognition request based on the desensitized biometric information.

15. A biometric privacy protection device, said biometric privacy protection device comprising: processor; as well as A memory configured to store computer-executable instructions, which, when executed, cause the processor to: Obtain a biometric request, wherein the biometric request carries a type of biometric information of the target user; The biological information is input into a biological attribute model that has been pre-trained, and the user attribute information of the target user is extracted from the biological information through the biological attribute model to obtain the user attribute information corresponding to the biological information. The user attribute information corresponding to the biometric information is sent to the server. The user attribute information is used to trigger the server to obtain the privacy protection model corresponding to the distribution of user attribute information of the target user from the stored correspondence between the distribution of user attribute information and the privacy protection model, based on the distribution of user attribute information corresponding to each attribute item of the target user. The obtained privacy protection model is determined as the privacy protection model for desensitizing the biometric information. The system receives the privacy protection model sent by the server, performs desensitization processing on the biometric information based on the privacy protection model to obtain desensitized biometric information, and performs biometric processing corresponding to the biometric request based on the desensitized biometric information.

16. A biometric privacy protection device, said biometric privacy protection device comprising: processor; as well as A memory configured to store computer-executable instructions, which, when executed, cause the processor to: The terminal device receives user attribute information corresponding to a type of biometric information of a target user. The biometric information of the target user is the information obtained by the terminal device when it receives the biometric recognition request of the target user. The user attribute information is the information extracted from the biometric information by inputting the biometric information into a pre-trained biometric attribute model. Based on the distribution of user attribute information corresponding to each attribute item of the target user, the privacy protection model corresponding to the distribution of user attribute information of the target user is obtained from the correspondence between the stored distribution of user attribute information and the privacy protection model, and the obtained privacy protection model is determined as the privacy protection model for desensitizing the biometric information. The determined privacy protection model is sent to the terminal device. The privacy protection model is used to trigger the terminal device to de-identify the biometric information to obtain the de-identified biometric information, and to perform biometric processing corresponding to the biometric recognition request based on the de-identified biometric information.

17. A storage medium for storing computer-executable instructions, which, when executed by a processor, perform the following process: Obtain a biometric request, wherein the biometric request carries a type of biometric information of the target user; The biological information is input into a biological attribute model that has been pre-trained, and the user attribute information of the target user is extracted from the biological information through the biological attribute model to obtain the user attribute information corresponding to the biological information. The user attribute information corresponding to the biometric information is sent to the server. The user attribute information is used to trigger the server to obtain the privacy protection model corresponding to the distribution of user attribute information of the target user from the stored correspondence between the distribution of user attribute information and the privacy protection model, based on the distribution of user attribute information corresponding to each attribute item of the target user. The obtained privacy protection model is determined as the privacy protection model for desensitizing the biometric information. The system receives the privacy protection model sent by the server, performs desensitization processing on the biometric information based on the privacy protection model to obtain desensitized biometric information, and performs biometric processing corresponding to the biometric request based on the desensitized biometric information.

18. A storage medium for storing computer-executable instructions, which, when executed by a processor, perform the following process: The terminal device receives user attribute information corresponding to a type of biometric information of a target user. The biometric information of the target user is the information obtained by the terminal device when it receives the biometric recognition request of the target user. The user attribute information is the information extracted from the biometric information by inputting the biometric information into a pre-trained biometric attribute model. Based on the distribution of user attribute information corresponding to each attribute item of the target user, the privacy protection model corresponding to the distribution of user attribute information of the target user is obtained from the correspondence between the stored distribution of user attribute information and the privacy protection model, and the obtained privacy protection model is determined as the privacy protection model for desensitizing the biometric information. The determined privacy protection model is sent to the terminal device. The privacy protection model is used to trigger the terminal device to de-identify the biometric information to obtain the de-identified biometric information, and to perform biometric processing corresponding to the biometric recognition request based on the de-identified biometric information.

Images