Method and apparatus for processing information

Abstract

The embodiments of the present disclosure disclose a method and device for processing information. A specific implementation of the method includes: the second end, in response to receiving a processing request for an object to be processed sent by the first end, obtains weight information and direction information from the first end, wherein the weight information is used to indicate the weight of the leaf node of the tree model, and the direction information is used to indicate the split direction of the object to be processed at the node managed by the first end; based on the pre-stored feature information and direction information of the object to be processed, the leaf node reached by the object to be processed in the tree model is determined; based on the weight information corresponding to the determined leaf node, the target score of the object to be processed in the tree model is generated; and the target score is sent to the first end. The second end of this implementation can independently complete the prediction process of the model, and the entire process does not need to disclose the split direction of the object to be processed on the managed node to the first end, which helps to improve the security of the data on the second end.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technology, and more particularly, to a method and apparatus for processing information. Background Art

[0002] In machine learning, data determines the upper limit of the model effect. In order to further improve the accuracy of the model, the business party can aggregate the data party's data for federated learning of the model. The business party can be the party that provides both feature data and feature data labels during the model training process, and the data party can be the party that only provides feature data.

[0003] In practice, the trained models can be deployed on the business side and the data side respectively, but the business side and the data side only manage some features and feature thresholds of the models.

[0004] Since the business party only manages some features and feature thresholds of the model, when using the model for prediction, the business party also needs to collaborate with the data party to complete the prediction process. Summary of the invention

[0005] Embodiments of the present disclosure provide methods and devices for processing information.

[0006] In a first aspect, an embodiment of the present disclosure provides a method for processing information, which is executed by a second end among predetermined participating ends, the participating ends include a first end and a second end, the first end and the second end have the same tree model, the first end and the second end respectively manage part of the nodes in the tree model, the first end has the weight of the leaf node of the tree model, the method includes: in response to receiving a processing request for an object to be processed sent by the first end, obtaining weight information and direction information from the first end, wherein the weight information is used to indicate the weight of the leaf node of the tree model, and the direction information is used to indicate the splitting direction of the object to be processed at the node managed by the first end; based on pre-stored feature information and direction information of the object to be processed, determining the leaf node reached by the object to be processed in the tree model; based on the weight information corresponding to the determined leaf node, generating a target score for the object to be processed in the tree model; and sending the target score to the first end.

[0007] In some embodiments, the tree model includes at least two trees; and pre-stored feature information and direction information of the object to be processed, determining the leaf node reached by the object to be processed in the tree model includes: based on the pre-stored feature information and direction information of the object to be processed, determining at least two leaf nodes reached by the object to be processed in the tree model; and based on the weight information corresponding to the determined leaf nodes, generating a target score for the object to be processed in the tree model includes: generating a target score for the object to be processed in the tree model based on at least two weight information corresponding to the at least two determined leaf nodes.

[0008] In some embodiments, the weight information includes an encrypted weight obtained by encrypting the weight of a leaf node of a tree model by a first end; and based on at least two weight information corresponding to at least two determined leaf nodes, generating a target score for the object to be processed in the tree model includes: summing at least two encrypted weights in at least two weight information to generate a target score for the object to be processed in the tree model.

[0009] In some embodiments, before summing at least two encrypted weights in at least two weight information to generate a target score for the object to be processed in the tree model, the method also includes: obtaining an encrypted threshold sent by the first end, wherein the encrypted threshold is obtained by encrypting the classification threshold by the first end, and the classification threshold is used to classify the object to be processed; and summing at least two encrypted weights in at least two weight information to generate the target score for the object to be processed in the tree model includes: summing at least two encrypted weights in at least two weight information to obtain the encrypted score of the object to be processed in the tree model; obtaining a target value pair; using the target value pair, respectively performing affine transformation on the encrypted score and the encrypted threshold to obtain a target score corresponding to the encrypted score and a target threshold corresponding to the encrypted threshold; and sending the target score to the first end includes: sending the target score and the target threshold to the first end.

[0010] In some embodiments, the direction information is generated by the first end through the following steps: for a tree included in a tree model, the following processing steps are performed: based on the nodes included in the tree, an initial bit string is generated, wherein the bits included in the initial bit string correspond to the nodes included in the tree, and the bits of the initial bit string do not include bit values; for the bits in the initial bit string corresponding to the tree, the following processing sub-steps are performed: in response to the node corresponding to the bit position being a node managed by the first end, determining the splitting direction of the object to be processed at the node corresponding to the bit position; selecting a candidate bit value corresponding to the determined splitting direction from a set of candidate bit values ​​as the bit value corresponding to the bit position; in response to the node corresponding to the bit position being a node managed by the second end, determining a preset bit value as the bit value corresponding to the bit position; generating a bit string corresponding to the tree and including bit values; generating direction information based on the bit string corresponding to the tree in the tree model.

[0011] In some embodiments, generating direction information based on bit strings corresponding to trees in the tree model includes: converting the bit strings corresponding to the trees in the tree model into numerical values ​​as representation values; and generating direction information based on the converted representation values.

[0012] In some embodiments, after receiving the target score, the first end performs the following steps: classifying the object to be processed based on the target score.

[0013] In a second aspect, an embodiment of the present disclosure provides a device for processing information, which is configured at a second end of a predetermined participating end, the participating end including a first end and a second end, the first end and the second end having the same tree model, the first end and the second end respectively managing part of the nodes in the tree model, the first end having the weight of the leaf node of the tree model, the device including: a first acquisition unit, configured to acquire weight information and direction information from the first end in response to receiving a processing request for an object to be processed sent by the first end, wherein the weight information is used to indicate the weight of the leaf node of the tree model, and the direction information is used to indicate the splitting direction of the object to be processed at the node managed by the first end; a determination unit, configured to determine the leaf node reached by the object to be processed in the tree model based on pre-stored feature information and direction information of the object to be processed; a generation unit, configured to generate a target score of the object to be processed in the tree model based on the weight information corresponding to the determined leaf node; and a sending unit, configured to send the target score to the first end.

[0014] In some embodiments, the tree model includes at least two trees; and the determination unit is further configured to: determine at least two leaf nodes reached by the object to be processed in the tree model based on pre-stored feature information and direction information of the object to be processed; and the generation unit is further configured to: generate a target score for the object to be processed in the tree model based on at least two weight information corresponding to the determined at least two leaf nodes.

[0015] In some embodiments, the weight information includes an encrypted weight obtained by encrypting the weight of a leaf node of a tree model by the first end; and the generation unit includes: a generation module, configured to sum at least two encrypted weights in at least two weight information to generate a target score for the object to be processed in the tree model.

[0016] In some embodiments, the device also includes: a second acquisition unit, configured to acquire an encrypted threshold sent by the first end, wherein the encrypted threshold is obtained by encrypting the classification threshold by the first end, and the classification threshold is used to classify the object to be processed; and the generation module includes: a summation submodule, configured to sum at least two encrypted weights in at least two weight information to obtain an encrypted score of the object to be processed in the tree model; an acquisition submodule, configured to acquire a target value pair; a transformation submodule, configured to use the target value pair to perform affine transformation on the encrypted score and the encryption threshold, respectively, to obtain a target score corresponding to the encrypted score and a target threshold corresponding to the encryption threshold; and the sending unit is further configured to: send the target score and the target threshold to the first end.

[0017] In some embodiments, the direction information is generated by the first end through the following steps: for a tree included in a tree model, the following processing steps are performed: based on the nodes included in the tree, an initial bit string is generated, wherein the bits included in the initial bit string correspond to the nodes included in the tree, and the bits of the initial bit string do not include bit values; for the bits in the initial bit string corresponding to the tree, the following processing sub-steps are performed: in response to the node corresponding to the bit position being a node managed by the first end, determining the splitting direction of the object to be processed at the node corresponding to the bit position; selecting a candidate bit value corresponding to the determined splitting direction from a set of candidate bit values ​​as the bit value corresponding to the bit position; in response to the node corresponding to the bit position being a node managed by the second end, determining a preset bit value as the bit value corresponding to the bit position; generating a bit string corresponding to the tree and including bit values; generating direction information based on the bit string corresponding to the tree in the tree model.

[0018] In some embodiments, generating direction information based on bit strings corresponding to trees in the tree model includes: converting the bit strings corresponding to the trees in the tree model into numerical values ​​as representation values; and generating direction information based on the converted representation values.

[0019] In some embodiments, after receiving the target score, the first end performs the following steps: classifying the object to be processed based on the target score.

[0020] In a third aspect, an embodiment of the present disclosure provides an electronic device, comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by the one or more processors, the one or more processors implement the method of any embodiment of the above-mentioned method for processing information.

[0021] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable medium having a computer program stored thereon, which, when executed by a processor, implements a method of any embodiment of the above-mentioned method for processing information.

[0022] In practice, the interpretability of models is of great significance in many fields. An interpretable model can provide a basis for making a decision. At present, in the field of federated learning, in order to improve the interpretability of the model, the data party usually discloses the features corresponding to the nodes it manages to the business party. However, in the process of using the model for prediction, the data party and the business party usually disclose to each other the split direction of the object to be predicted at the node they manage. Therefore, in the prediction process, the business party can simultaneously obtain the features corresponding to the nodes managed by the data party and the split direction of the object to be predicted under the feature. This will make it possible for the business party to infer the feature threshold corresponding to the node managed by the data party based on the split direction and features, combined with background knowledge or elimination method, and thus lead to the privacy leakage of the data party.

[0023] The method and device for processing information provided by the embodiments of the present disclosure can obtain weight information and direction information from the first end in response to receiving a processing request for an object to be processed sent by the first end, wherein the weight information is used to indicate the weight of the leaf node of the tree model, and the direction information is used to indicate the splitting direction of the object to be processed at the node managed by the first end, and then based on the pre-stored feature information and direction information of the object to be processed, determine the leaf node reached by the object to be processed in the tree model, and then generate the target score of the object to be processed in the tree model based on the weight information corresponding to the determined leaf node, and finally send the target score to the first end, so that the second end can independently complete the prediction process of the model based on the weight information and direction information provided by the first end. The entire process does not need to disclose the splitting direction of the object to be processed on the managed node to the first end, which helps to improve the security of the data on the second end. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Other features, objects and advantages of the present disclosure will become more apparent from the detailed description of non-limiting embodiments made with reference to the following drawings:

[0025] Figure 1 is an exemplary system architecture diagram in which an embodiment of the present disclosure may be applied;

[0026] Figure 2 is a flow chart of one embodiment of a method for processing information according to the present disclosure;

[0027] Figure 3 is a structural schematic diagram of a tree model in an embodiment of a method for processing information according to the present disclosure;

[0028] Figure 4 is a schematic diagram of an application scenario of a method for processing information according to an embodiment of the present disclosure;

[0029] Figure 5 is a flow chart of yet another embodiment of a method for processing information according to the present disclosure;

[0030] Figure 6 is a structural schematic diagram of a tree model in yet another embodiment of the method for processing information according to the present disclosure;

[0031] Figure 7 is a schematic structural diagram of an embodiment of a device for processing information according to the present disclosure;

[0032] Figure 8 It is a structural diagram of a computer system of an electronic device suitable for implementing an embodiment of the present disclosure. DETAILED DESCRIPTION

[0033] The present disclosure is further described in detail below in conjunction with the accompanying drawings and embodiments. It is understood that the specific embodiments described herein are only used to explain the relevant invention, rather than to limit the invention. It is also necessary to explain that, for ease of description, only the parts related to the relevant invention are shown in the accompanying drawings.

[0034] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other. The present disclosure will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0035] Figure 1 An exemplary system architecture 100 is shown to which embodiments of the method for processing information or the apparatus for processing information of the present disclosure may be applied.

[0036] like Figure 1 As shown, the system architecture 100 may include first terminals 101, 102, a network 103, and second terminals 104, 105. The network 103 may provide a medium for a communication link between the first terminals 101, 102 and the second terminals 104, 105. The network 103 may include various connection types, such as wired, wireless communication links, or fiber optic cables.

[0037] The first end 101, 102 and the second end 104, 105 may be participating ends of the federated learning of the model. Specifically, the first end 101, 102 may correspond to the business party in the federated learning process, and the second end 104, 105 may correspond to the data party in the federated learning process. Among them, the business party may be the party that provides both feature data and labels of feature data during the federated learning process. The data party may be the party that only provides feature data. The dimension of the feature data provided by the business party is usually different from the dimension of the data provided by the data party (for example, the business party provides feature data of age features, and the data party provides feature data of height features). The model obtained after the federated learning can be deployed in the first end 101, 102 and the second end 104, 105 respectively. The first end 101, 102 and the second end 104, 105 respectively manage the features corresponding to the feature data provided by them.

[0038] The first ends 101 and 102 can be hardware or software. When the first ends 101 and 102 are hardware, they can be various electronic devices, including but not limited to servers, smart phones, tablet computers, laptop computers, desktop computers, etc. When the first ends 101 and 102 are software, they can be installed in the electronic devices listed above. It can be implemented as multiple software or software modules (for example, multiple software or software modules for providing distributed services), or it can be implemented as a single software or software module. No specific limitation is made here.

[0039] Similarly, the second ends 104 and 105 can be hardware or software. When the second ends 104 and 105 are hardware, they can be various electronic devices, including but not limited to servers, smart phones, tablet computers, laptop computers, desktop computers, etc. When the second ends 104 and 105 are software, they can be installed in the electronic devices listed above. It can be implemented as multiple software or software modules (for example, multiple software or software modules for providing distributed services), or it can be implemented as a single software or software module. No specific limitation is made here.

[0040] It should be noted that the method for processing information provided in the embodiments of the present disclosure is generally executed by the second ends 104 and 105 , and accordingly, the device for processing information is generally disposed in the second ends 104 and 105 .

[0041] It should be understood that Figure 1 The number of the first end, network and second end in the embodiment is only for illustration. According to the implementation requirements, there may be any number of the first end, network and second end.

[0042] Continue to refer Figure 2 , shows a process 200 of an embodiment of a method for processing information according to the present disclosure. The method for processing information comprises the following steps:

[0043] Step 201: in response to receiving a processing request for an object to be processed sent by a first end, obtaining weight information and direction information from the first end.

[0044] In this embodiment, the second end of the execution body of the method for processing information can obtain weight information and direction information from the first end in response to receiving a processing request for the object to be processed sent by the first end through a wired connection or a wireless connection. Among them, the first end and the above-mentioned second end belong to the predetermined participating end. The first end and the second end have the same tree model. The tree model can be various models composed of trees, for example, it can be a gradient boosting tree model. Specifically, the tree model can be a model obtained by the first end and the second end through joint learning training. The first end and the second end can respectively manage some nodes in the tree model.

[0045] For example, please refer to Figure 3 , Figure 3 A structural diagram of a tree model of this embodiment is given. The tree model can be configured at the first end and the second end respectively. Specifically, the tree model can include three nodes and four leaf nodes. The nodes represented by triangles among the three nodes can be nodes managed by the second end, and the nodes represented by squares can be nodes managed by the first end. The four leaf nodes can be represented by circles.

[0046] In this embodiment, the nodes in the tree model managed by the first end and the second end respectively can be determined in the process of training the tree model. Specifically, the nodes managed by the first end and the second end respectively are related to the feature information stored by the first end and the second end respectively. As an example, if the first end stores age information and the second end stores height information, the first end manages the nodes corresponding to the age feature in the tree model, and the second end manages the nodes corresponding to the height feature in the tree model.

[0047] In this embodiment, the first end may have the weight of the leaf node of the tree model. Here, the nodes of the tree model can be split, while the leaf nodes of the tree model cannot be split. The weight of the leaf node can be obtained in the process of training the tree model. Specifically, the weight of a leaf node can be used to indicate the number of training objects falling on the leaf node. The more the number, the greater the weight of the leaf node.

[0048] It should be noted that the method of obtaining a tree model through joint learning training is a well-known technology that is currently widely studied and applied, and will not be described in detail here.

[0049] In this embodiment, the object to be processed may be an object to be processed, specifically, various objects that can be processed by the above-mentioned tree model, such as images, texts, videos, audios, etc.

[0050] Specifically, the first end may send a processing request for the object to be processed to the second end in various ways. For example, the first end may send an identifier (such as a number, account number, etc.) of the object to be processed to the second end to request processing of the object to be processed.

[0051] In this embodiment, the weight information can be used to indicate the weight of the leaf node of the tree model. Specifically, the weight information can be obtained by the first end using various methods. For example, the first end can directly determine the information including the weight of the leaf node of the tree model as the weight information; or, the first end can encrypt the weight of the leaf node of the tree model to obtain the encrypted weight, and then determine the information including the encrypted weight as the weight information.

[0052] In this embodiment, the direction information may be used to indicate the splitting direction of the object to be processed at the node managed by the first end.

[0053] It can be understood that the nodes in the tree model correspond to features and feature thresholds. For the object to be processed, the first end can determine the splitting direction of the object to be processed at the node through its corresponding feature information and the features and feature thresholds corresponding to the node.

[0054] As an example, the object to be processed is user 1, and the feature information of user 1 stored at the first end is age 20. The feature corresponding to node A managed by the first end in the tree model is age, and its corresponding feature threshold is 18 years old. It is predetermined that if the age is greater than or equal to 18 years old, split to the right; if the age is less than 18 years old, split to the left. Since user 1 is 20 years old, it can be determined that the split direction of user 1 at node A is right split.

[0055] In this embodiment, after the first end determines the split direction of the object to be processed at the node it manages, various methods can be used to generate direction information. For example, the first end can directly represent the determined split direction with a target identifier (such as the text "left" or the text "right"), and then determine the target identifier as the direction information.

[0056] In some optional implementations of this embodiment, the direction information may be generated by the first end through the following steps:

[0057] First, for the trees included in the tree model, the following processing steps are performed:

[0058] The first step is to generate an initial bit string based on the nodes included in the tree.

[0059] The bits included in the initial bit string correspond to the nodes included in the tree, and the bits in the initial bit string do not include bit values.

[0060] Continuing with the above example, Figure 3 The tree model shown includes three nodes, so the initial bit string may include three bits, and the three bits correspond one-to-one to the three nodes included in the tree model.

[0061] In the second step, for the bit position in the initial bit string corresponding to the tree, perform the following processing sub-steps: in response to the node corresponding to the bit position being a node managed by the first end, determine the splitting direction of the object to be processed at the node corresponding to the bit position; select the candidate bit value corresponding to the determined splitting direction from the candidate bit value set as the bit value corresponding to the bit position; in response to the node corresponding to the bit position being a node managed by the second end, determine the preset bit value as the bit value corresponding to the bit position.

[0062] Among them, the candidate bit value set may be a predetermined bit value set, and the candidate bit value set may include bit values ​​corresponding to the splitting direction. For example, the candidate bit value set may include candidate bit value 0 and candidate bit value 1, wherein candidate bit value 0 may correspond to "split to the left"; candidate bit value 1 may correspond to "split to the right". The preset bit value may be a predetermined bit value used to indicate that the splitting direction cannot be known. It should be noted that the preset bit value may be the same as a candidate bit value in the candidate bit value set. Continuing with the above example, the preset bit value may be 0.

[0063] continue Figure 3 For example, Figure 3 The nodes represented by squares in the tree model are nodes managed by the first end, so the first end can determine the splitting direction of the object to be processed at the node, for example, right splitting, and then the first end can select the candidate bit value 1 corresponding to "right splitting" from the candidate bit value set as the bit value of the bit position corresponding to the node; for two nodes represented by triangles, since these two nodes are nodes managed by the second end, the first end can determine the preset bit value 0 as the bit value of the bit positions corresponding to the two nodes.

[0064] The third step is to generate a bit string including bit values ​​corresponding to the tree.

[0065] The bit string corresponding to the tree can be obtained by adding the determined bit value to the corresponding bit position in the initial bit string.

[0066] Continuing with the above example, based on the determined bit values ​​1, 0, 0, the first end may generate a bit string 100 corresponding to the tree.

[0067] Then, based on the bit string corresponding to the tree in the tree model, the direction information is generated.

[0068] For a tree in the tree model, after the first end generates the bit string corresponding to the tree through the above processing steps, the first end can generate the direction information corresponding to the tree based on the bit string of the tree. Specifically, the first end can use various methods to generate the direction information based on the bit string. For example, the first end can directly determine the bit string corresponding to the tree as the direction information.

[0069] It can be understood that the bit string includes the bit position corresponding to the node managed by the first end, and the bit value on the bit position can indicate the splitting direction of the object to be processed at the node corresponding to the bit position. Therefore, the bit string corresponding to the tree can indicate the splitting direction of the object to be processed at the node managed by the first end. Furthermore, based on the bit string corresponding to the tree, direction information for indicating the splitting direction of the object to be processed at the node managed by the first end can be generated.

[0070] In some optional implementations of this embodiment, based on the bit string corresponding to the tree in the tree model, the first end can generate direction information through the following steps: first, convert the bit string corresponding to the tree in the tree model into a numerical value as a representation value, and generate direction information based on the converted representation value.

[0071] Specifically, the first end may convert the bit string into a numerical value in various ways. For example, if the bit string is 100, the first end may convert it into the numerical value 100; or, the binary bit string 100 may be converted into the decimal numerical value 4.

[0072] This implementation method only uses one numerical value to represent the splitting direction of the object to be processed at the node managed by the first end. Compared with the solution of determining the identifier (such as text, bit value) used to represent the splitting direction of the object to be processed at each node managed by the first end as the direction information, this implementation method uses the expression value as the direction information, which can reduce the amount of data sent from the first end to the second end, help save communication bandwidth, and improve data transmission efficiency.

[0073] Step 202: Based on the pre-stored feature information and direction information of the object to be processed, determine the leaf node reached by the object to be processed in the tree model.

[0074] In this embodiment, after receiving the processing request for the object to be processed sent by the first end, the second end can first extract the pre-stored feature information of the object to be processed. The features represented by the feature information here can be the same as the features corresponding to the nodes managed by the second end in the tree model. For example, the second end manages two nodes in the tree model, and the features corresponding to the two nodes are height and weight, respectively. Then, after receiving the processing request for user 1, the second end can extract the height information and weight information of user 1.

[0075] In this embodiment, based on the extracted feature information of the object to be processed and the direction information obtained in step 201, the second end can determine the leaf node reached by the object to be processed in the above tree model.

[0076] Specifically, the second end can start from the root node of the tree model and split the object to be processed. Specifically, when the current node is a node managed by the second end, the object to be processed can be split based on the extracted feature information; when the current node is a node managed by the first end, the object to be processed can be split based on the acquired direction information.

[0077] Continuing with the above example, for Figure 3In the tree model shown, the second end can start from the node indicated by the square and split the user 1. Specifically, the following steps are included: first, for the node indicated by the square, the second end can determine that the node is a node managed by the first end, and then based on the direction information "right", it can determine that the split direction of the user 1 at the node is to the right, and then the node indicated by the triangle on the right can be reached; then, for the node indicated by the triangle on the right, the feature corresponding to the node is height, the feature threshold is 170 cm, and it is predetermined that if the height is greater than or equal to 170 cm, it will split to the right; if the height is less than 170 cm, it will split to the left, and then based on the extracted feature information "height 180 cm" of user 1, the second end can determine that the split direction of the user 1 at the node is to the right, and then the node can be reached. Figure 3 The rightmost leaf node in .

[0078] Step 203: Generate a target score of the object to be processed in the tree model based on the weight information corresponding to the determined leaf nodes.

[0079] In this embodiment, after determining the leaf node to which the object to be processed reaches, the second end can extract the weight information corresponding to the determined leaf node from the acquired weight information, and then generate the target score of the object to be processed in the tree model based on the extracted weight information. Among them, the target score can be a numerical value used to characterize the possibility that the object to be processed belongs to a preset category. The higher the target score, the greater the possibility that the object to be processed belongs to the preset category. The preset category can be a category that the tree model can predict. As an example, the tree model is used to predict whether the user is a user to whom information can be recommended, then the preset category can be a recommendable information category; the tree model is used to predict whether the user is obese, then the preset category can be an obesity category.

[0080] Specifically, based on the determined weight information of the leaf node, the second end may generate the target score using various methods. As an example, the determined leaf node includes a leaf node, and the weight information of the leaf node includes the weight of the leaf node, then the second end may directly determine the weight of the leaf node as the target score; or, the weight information of the leaf node includes the encrypted weight of the leaf node, then the second end may determine the encrypted weight of the leaf node as the target score.

[0081] Step 204: Send the target score to the first end.

[0082] In this embodiment, based on the target score obtained in step 203, the second end may send the target score to the first end.

[0083] In some optional implementations of this embodiment, after receiving the target score, the first end may perform the following steps: classifying the object to be processed based on the target score.

[0084] Specifically, the first end may pre-store a classification threshold for classifying the object to be processed, and then after obtaining the target score, the first end may classify the object to be processed based on the target score and the classification threshold.

[0085] As an example, the first end pre-stores a classification threshold of 70 for determining whether a user belongs to the obese class. When the user's score is greater than or equal to 70, it can be determined that the user belongs to the obese class. Furthermore, after the first end obtains the target score "60" corresponding to user 1, since 60 is less than 70, the first end can classify user 1 as non-obese.

[0086] Continue to see Figure 4 , Figure 4 FIG. 1 is a schematic diagram of an application scenario of the method for processing information according to this embodiment. Figure 4 In the application scenario of FIG. 4 , a first server 401 (corresponding to the first end) and a second server 402 (corresponding to the second end) have the same tree model 403, and the tree model 403 includes nodes 4031, 4032, 4033, and leaf nodes 4034, 4035, 4036, and 4037. The first server 401 can manage the node 4031 in the tree model 403. The second server 402 can manage the nodes 4032 and 4033 in the tree model 403. The first server 401 can have weights corresponding to the leaf nodes 4034, 4035, 4036, and 4037 of the tree model 403, respectively.

[0087] In this scenario, the second server 402 can respond to receiving a processing request 405 for the object to be processed sent by the first server 401, and first obtain weight information 4061, weight information 4062, weight information 4063, weight information 4064 and direction information 407 from the first server 401, wherein weight information 4061, 4062, 4063, 4064 can be used to indicate the weights of leaf node 4034, leaf node 4035, leaf node 4036 and leaf node 4037, respectively, and direction information 307 can be used to indicate the splitting direction of the object to be processed at node 4031 managed by the first server 401.

[0088] Next, the second server 402 may determine the leaf node reached by the object to be processed in the tree model 403 , for example, the leaf node 4035 , based on the pre-stored feature information 408 and direction information 407 of the object to be processed.

[0089] Then, the second server 402 may generate a target score 409 for the object to be processed in the tree model 403 based on the weight information 4062 corresponding to the determined leaf node 4035 .

[0090] Finally, the second server 402 may send the target score 409 to the first server 401 , so that the first server 401 can make a decision on the object to be processed based on the target score 409 .

[0091] The method provided by the above-mentioned embodiment of the present disclosure can be executed by the second end among the predetermined participating ends. The second end can independently complete the prediction process of the model based on the weight information and direction information provided by the first end. The entire process does not need to disclose the splitting direction of the object to be processed on the managed node to the first end, which helps to improve the security of the data on the second end.

[0092] Further references Figure 5 , which shows a process 500 of another embodiment of a method for processing information. The process 500 of the method for processing information includes the following steps:

[0093] Step 501: in response to receiving a processing request for an object to be processed sent by a first end, obtaining weight information and direction information from the first end.

[0094] In this embodiment, the second end of the execution body of the method for processing information can obtain weight information and direction information from the first end in response to receiving a processing request for the object to be processed sent by the first end through a wired connection or a wireless connection. Among them, the first end and the above-mentioned second end belong to the predetermined participating end. The first end and the second end have the same tree model. The tree model can be various models composed of trees. Specifically, the tree model can include at least two trees. Here, the tree model can be a model obtained by the first end and the second end through joint learning training. The first end and the second end can respectively manage some nodes in the tree model.

[0095] For example, please refer to Figure 6 , which shows a schematic diagram of the structure of a tree model in this embodiment. The tree model in the figure may include two trees, namely tree 601 and tree 602. The nodes represented by triangles in the two trees may be nodes managed by the second end, and the nodes represented by squares may be nodes managed by the first end. The leaf nodes of the two trees are represented by circles.

[0096] In this embodiment, the nodes in the tree model managed by the first end and the second end respectively can be determined in the process of training the tree model. Specifically, the nodes managed by the first end and the second end respectively are related to the feature information stored by the first end and the second end respectively.

[0097] In this embodiment, the first end may have the weight of the leaf node of the tree model. Here, the node of the tree model can be split, while the leaf node of the tree model cannot be split. The weight of the leaf node can be obtained in the process of training the tree model.

[0098] In this embodiment, the object to be processed may be an object to be processed. The weight information may be used to indicate the weight of a leaf node of a tree model. The direction information may be used to indicate the splitting direction of the object to be processed at a node managed by the first end.

[0099] Step 502: Based on the pre-stored feature information and direction information of the object to be processed, determine at least two leaf nodes reached by the object to be processed in the tree model.

[0100] In this embodiment, after receiving the processing request for the object to be processed sent by the first end, the second end can first extract the pre-stored feature information of the object to be processed. The features represented by the feature information here can be the same as the features corresponding to the nodes managed by the second end in the tree model. For example, the second end manages two nodes in the tree model, and the features corresponding to the two nodes are height and weight, respectively. Then, after receiving the processing request for user 1, the second end can extract the height information and weight information of user 1.

[0101] In this embodiment, based on the extracted feature information of the object to be processed and the direction information obtained in step 501, the second end can determine at least two leaf nodes reached by the object to be processed in the above tree model.

[0102] It can be understood that the tree model includes at least two trees, and the object to be processed can reach a leaf node in each tree. Therefore, the object to be processed can reach at least two leaf nodes in the tree model.

[0103] As an example, for Figure 6 The tree model shown includes two trees. For tree 601, the second end can determine a leaf node reached by an object to be processed in the tree 601; for tree 602, the second end can determine a leaf node reached by an object to be processed in the tree 602. Furthermore, the second end can determine two leaf nodes reached by the object to be processed in the tree model.

[0104] Specifically, for each of the at least two trees included in the tree model, the second end may split the object to be processed starting from the root node of the tree. Specifically, when the current node is a node managed by the second end, the object to be processed may be split based on the extracted feature information; when the current node is a node managed by the first end, the object to be processed may be split based on the acquired direction information.

[0105] Step 503: Generate a target score of the object to be processed in the tree model based on the at least two weight information corresponding to the at least two leaf nodes determined.

[0106] In this embodiment, after determining at least two leaf nodes reached by the object to be processed, the second end can extract the weight information corresponding to the at least two leaf nodes determined from the acquired weight information, and then generate the target score of the object to be processed in the tree model based on the extracted at least two weight information. Among them, the target score can be a numerical value used to characterize the possibility that the object to be processed belongs to a preset category. The higher the target score, the greater the possibility that the object to be processed belongs to the preset category. The preset category can be a category that the tree model can predict.

[0107] It should be noted that, for each leaf node reached by the object to be processed, the second end can extract a weight information of the leaf node, and further, for at least two leaf nodes, the second end can extract at least two weight information.

[0108] Specifically, based on the at least two weight information of the at least two leaf nodes determined, the second end can generate the target score using various methods. As an example, the second end determines two leaf nodes reached by the object to be processed in the tree model, and the weight information of the two leaf nodes includes the weights of the two leaf nodes. Then, the second end can sum the weights of the two leaf nodes and obtain the summed result as the target score.

[0109] In some optional implementations of this embodiment, the weight information may include an encrypted weight obtained after the first end encrypts the weight of the leaf node of the tree model; and

[0110] The second end may generate a target score of the object to be processed in the tree model through the following steps: summing at least two encrypted weights in at least two weight information to generate a target score of the object to be processed in the tree model.

[0111] In this implementation, the second end may obtain an encrypted score after summing at least two encryption weights, where the encrypted score refers to an encrypted score. Furthermore, the second end may generate a target score based on the encrypted score.

[0112] Specifically, the second end may generate the target score in various ways based on the encrypted score. For example, the second end may directly determine the encrypted score as the target score.

[0113] It should be noted that, here, the first end can encrypt the weight of the leaf node by homomorphic encryption, and then, according to homomorphism, the encryption method of the encrypted score is the same as the encryption method of the encrypted weight. Then, the first end can decrypt the encrypted score based on the encryption method of the weight, and obtain the score of the object to be processed.

[0114] In this implementation, the first end can transmit the encrypted encryption weight to the second end, which can protect the privacy of the first end and improve the security of the data of the first end.

[0115] Step 504: Send the target score to the first end.

[0116] In this embodiment, based on the target score obtained in step 503, the second end may send the target score to the first end so that the first end can classify the object to be processed based on the target score.

[0117] In some optional implementations of this embodiment, before generating the target score of the object to be processed in the tree model, the second end may further perform the following steps:

[0118] First, the encryption threshold sent by the first end is obtained.

[0119] The encryption threshold is obtained by encrypting the classification threshold at the first end. The classification threshold can be used to classify the object to be processed.

[0120] It should be noted that the way in which the first end encrypts the classification threshold value may be the same as the way in which the weight of the leaf node is encrypted.

[0121] Then, at least two encrypted weights in at least two pieces of weight information are summed to obtain an encrypted score of the object to be processed in the tree model.

[0122] Next, obtain the target value pair.

[0123] Here, the target value pair may include two target values. The target values ​​in the target value pair may be predetermined values.

[0124] Optionally, the target value in the target value pair may be a randomly generated value, thereby reducing the possibility of the first end acquiring the target value pair and improving the security of the data on the second end.

[0125] Furthermore, the second end can randomly generate a target value pair each time it needs to send an encrypted score to the first end, which can reduce the possibility of the first end inferring the current target value pair based on historical target value pairs, and help to further improve the security of the data on the second end.

[0126] Then, the target value pair is used to perform affine transformation on the encryption score and the encryption threshold, respectively, to obtain the target score corresponding to the encryption score and the target threshold corresponding to the encryption threshold.

[0127] In geometry, an affine transformation can refer to a linear transformation of a vector space followed by a translation to transform it into another vector space. Here, the two target values ​​in the target value pair can correspond to the linear transformation and translation of the affine transformation respectively.

[0128] As an example, the encryption score is S, the encryption threshold is T, and the target value pair is (a, b). Then, the target score S' can be obtained by performing an affine transformation on the encryption score S using the target value pair (a, b), where S' = S·a+b; the target threshold T' can be obtained by performing an affine transformation on the encryption threshold T using the target value pair (a, b), where T' = T·a+b.

[0129] Finally, the second end may send the obtained target score and target threshold to the first end, so that the first end classifies the object to be processed based on a comparison result obtained by comparing the target score and the target threshold.

[0130] Here, it should be noted that the encryption method for the classification threshold is the same as the encryption method for the weight, which can be specifically homomorphic encryption. Furthermore, the encryption method corresponding to the encrypted score obtained by summing the encrypted weights is also the same as the encryption method corresponding to the encrypted threshold. The encrypted score is homomorphic to the unencrypted score, and the encrypted threshold is homomorphic to the classification threshold. According to the homomorphism, the comparison result of the encrypted score and the encrypted threshold can be the same as the comparison result of the unencrypted score and the classification threshold.

[0131] In addition, since the present implementation method uses the target numerical value to perform the same affine transformation on the encrypted score and the encrypted threshold, and the affine transformation is a linear transformation, the affine transformation of the present implementation method does not change the comparison result between the score and the threshold, that is, the comparison result between the target score and the target threshold can be the same as the comparison result between the encrypted score and the encrypted threshold, that is, the same as the comparison result between the unencrypted score and the classification threshold.

[0132] For example, the unencrypted score is 80, the classification threshold is 90, and the encrypted score and encryption threshold can be obtained after homomorphic encryption. The target value pair is (3,4), then the target score 28 (28=3·8+4) can be obtained by affine transformation of the encrypted score 8 using the target value pair; the target threshold 31 (28=3·9+4) can be obtained by affine transformation of the encrypted threshold 9 using the target value pair. It is not difficult to see that homomorphic encryption and affine transformation do not change the comparison results of the score and threshold.

[0133] In practice, since the first end holds an encryption method for encrypting weights, that is, a decryption method for decrypting encrypted scores, and holds the weights of leaf nodes, if the second end directly sends the encrypted scores to the first end, the first end may infer the splitting results of the object to be processed in the tree model based on the actual scores obtained after decryption, and then infer the splitting direction of the object to be processed at the node managed by the second end based on the features managed by itself and the features disclosed by the second end, resulting in the privacy leakage of the second end. However, the second end in this implementation method can use the target value pair to perform an affine transformation on the encrypted score before sending it to the first end. Since the first end cannot know the inverse transformation of the affine transformation, the first end cannot directly obtain the actual score, but can only obtain the comparison result between the actual score and the classification threshold. Furthermore, the first end cannot infer the splitting result of the object to be processed. In this way, the privacy of the second end can be protected and the security of the data of the second end can be further improved.

[0134] from Figure 5 It can be seen that Figure 2 Compared with the corresponding embodiment, the process 500 of the method for processing information in this embodiment highlights that the tree model includes at least two trees, and then based on the pre-stored feature information and direction information of the object to be processed, at least two leaf nodes reached by the object to be processed in the tree model are determined, and based on at least two weight information corresponding to the at least two determined leaf nodes, a step of generating a target score for the object to be processed in the tree model. Therefore, the target score in the scheme described in this embodiment is obtained based on at least two leaf nodes corresponding to at least two trees. Compared with the scheme obtained based on one leaf node corresponding to one tree, the scheme of this embodiment can reduce the possibility that the first end can infer the splitting direction of the object to be processed on the node managed by the second end based on the target score, which helps to further reduce the risk of privacy leakage at the second end.

[0135] Further references Figure 7 As an implementation of the methods shown in the above figures, the present disclosure provides an embodiment of a device for processing information. Figure 2 Corresponding to the method embodiment shown, the device can be configured at the second end of a predetermined participating end, the participating end includes a first end and a second end, the first end and the second end have the same tree model, the first end and the second end respectively manage part of the nodes in the tree model, and the first end has the weight of the leaf node of the tree model.

[0136] like Figure 7As shown, the apparatus 700 for processing information in this embodiment includes: a first acquisition unit 701, a determination unit 702, a generation unit 703 and a push unit 704. The first acquisition unit 701 is configured to, in response to receiving a processing request for an object to be processed sent by the first end, acquire weight information and direction information from the first end, wherein the weight information is used to indicate the weight of a leaf node of a tree model, and the direction information is used to indicate the splitting direction of the object to be processed at a node managed by the first end; the determination unit 702 is configured to determine the leaf node reached by the object to be processed in the tree model based on pre-stored feature information and direction information of the object to be processed; the generation unit 703 is configured to generate a target score of the object to be processed in the tree model based on the weight information corresponding to the determined leaf node; and the sending unit 704 is configured to send the target score to the first end.

[0137] In this embodiment, the first acquisition unit 701 of the device 700 for processing information can obtain weight information and direction information from the first end in response to receiving a processing request for the object to be processed sent by the first end through a wired connection or a wireless connection. The first end and the second end belong to the same predetermined participating end. The first end and the second end have the same tree model. The tree model can be various models composed of trees. Specifically, the tree model can be a model obtained by the first end and the second end through joint learning training. The first end and the second end can respectively manage some nodes in the tree model.

[0138] In this embodiment, the first end may have the weight of the leaf node of the tree model. The object to be processed may be an object to be processed. The weight information may be used to indicate the weight of the leaf node of the tree model. The direction information may be used to indicate the splitting direction of the object to be processed at the node managed by the first end.

[0139] In this embodiment, after receiving the processing request for the object to be processed sent by the first end, the determining unit 702 may first extract the pre-stored feature information of the object to be processed. The features represented by the feature information here may be the same as the features corresponding to the nodes managed by the second end in the tree model.

[0140] In this embodiment, based on the extracted feature information of the object to be processed and the direction information acquired by the first acquisition unit 701 , the determination unit 702 may determine the leaf node reached by the object to be processed in the above tree model.

[0141] In this embodiment, after determining the leaf node to be processed, the generation unit 703 can extract the weight information corresponding to the leaf node determined by the determination unit 702 from the weight information obtained by the first acquisition unit 701, and then generate the target score of the object to be processed in the tree model based on the extracted weight information. The target score can be a numerical value used to characterize the possibility that the object to be processed belongs to a preset category. The higher the target score, the greater the possibility that the object to be processed belongs to the preset category. The preset category can be a category that the tree model can predict.

[0142] In this embodiment, based on the target score obtained by the generating unit 703, the sending unit 704 can send the target score to the first end, so that the first end classifies the objects to be processed based on the target score.

[0143] In some optional implementations of this embodiment, the tree model includes at least two trees; and the determination unit 702 is further configured to: determine at least two leaf nodes reached by the object to be processed in the tree model based on pre-stored feature information and direction information of the object to be processed; and the generation unit 703 is further configured to: generate a target score for the object to be processed in the tree model based on at least two weight information corresponding to the determined at least two leaf nodes.

[0144] In some optional implementations of this embodiment, the weight information includes the encrypted weight obtained after the first end encrypts the weight of the leaf node of the tree model; and the generation unit 703 includes: a generation module (not shown in the figure), which is configured to sum at least two encrypted weights in at least two weight information to generate a target score for the object to be processed in the tree model.

[0145] In some optional implementations of the present embodiment, the device 700 also includes: a second acquisition unit (not shown in the figure), configured to acquire the encrypted threshold sent by the first end, wherein the encrypted threshold is obtained by encrypting the classification threshold by the first end, and the classification threshold is used to classify the object to be processed; and the generation module includes: a summation submodule (not shown in the figure), configured to sum at least two encrypted weights in at least two weight information to obtain the encrypted score of the object to be processed in the tree model; an acquisition submodule (not shown in the figure), configured to acquire a target value pair; a transformation submodule (not shown in the figure), configured to use the target value pair to perform affine transformation on the encrypted score and the encrypted threshold, respectively, to obtain a target score corresponding to the encrypted score and a target threshold corresponding to the encrypted threshold; and the sending unit 704 is further configured to: send the target score and the target threshold to the first end.

[0146] In some optional implementations of the present embodiment, the direction information is generated by the first end through the following steps: for a tree included in a tree model, the following processing steps are performed: based on the nodes included in the tree, an initial bit string is generated, wherein the bits included in the initial bit string correspond to the nodes included in the tree, and the bits of the initial bit string do not include bit values; for the bits in the initial bit string corresponding to the tree, the following processing sub-steps are performed: in response to the node corresponding to the bit position being a node managed by the first end, determining the splitting direction of the object to be processed at the node corresponding to the bit position; selecting a candidate bit value corresponding to the determined splitting direction from a set of candidate bit values ​​as the bit value corresponding to the bit position; in response to the node corresponding to the bit position being a node managed by the second end, determining a preset bit value as the bit value corresponding to the bit position; generating a bit string corresponding to the tree and including bit values; and generating direction information based on the bit string corresponding to the tree in the tree model.

[0147] In some optional implementations of this embodiment, generating direction information based on the bit strings corresponding to the trees in the tree model includes: converting the bit strings corresponding to the trees in the tree model into numerical values ​​as representation values; and generating direction information based on the converted representation values.

[0148] In some optional implementations of this embodiment, after receiving the target score, the first end may perform the following steps: classifying the objects to be processed based on the target score.

[0149] It is understood that the units described in the device 700 are similar to those described in the reference Figure 2 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the device 700 and the units included therein, and will not be described in detail here.

[0150] The device 700 provided in the above-mentioned embodiment of the present disclosure can be configured on the second end, and the second end configured with the device 700 can independently complete the prediction process of the model based on the weight information and direction information provided by the first end. The entire process does not need to disclose the splitting direction of the object to be processed on the managed node to the first end, which helps to improve the security of the data on the second end.

[0151] Reference below Figure 8 , which shows an electronic device (eg, Figure 1 The terminal device in the embodiment of the present disclosure may include but is not limited to mobile terminals such as mobile phones, notebook computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. Figure 8 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0152] like Figure 8 As shown, the electronic device 800 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 801, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 802 or a program loaded from a storage device 808 into a random access memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the electronic device 800 are also stored. The processing device 801, the ROM 802, and the RAM 803 are connected to each other via a bus 804. An input / output (I / O) interface 805 is also connected to the bus 804.

[0153] Typically, the following devices may be connected to the I / O interface 805: input devices 806 including, for example, a touch screen, a touch pad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; output devices 807 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; storage devices 808 including, for example, a magnetic tape, a hard disk, etc.; and communication devices 809. The communication device 809 may allow the electronic device 800 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 8 The electronic device 800 is shown with various devices, but it should be understood that it is not required to implement or possess all the devices shown. More or fewer devices may be implemented or possessed instead.

[0154] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication device 809, or installed from a storage device 808, or installed from a ROM 802. When the computer program is executed by the processing device 801, the above-mentioned functions defined in the method of the embodiment of the present disclosure are executed.

[0155] It should be noted that the computer-readable medium described in the present disclosure may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. The computer readable signal medium may also be any computer readable medium other than a computer readable storage medium, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.

[0156] The computer-readable medium may be included in the electronic device; or it may exist independently without being assembled into the electronic device. The computer-readable medium carries one or more programs. When the one or more programs are executed by the electronic device, the electronic device: in response to receiving a processing request for the object to be processed sent by the first end, obtain weight information and direction information from the first end, wherein the weight information is used to indicate the weight of the leaf node of the tree model, and the direction information is used to indicate the splitting direction of the object to be processed at the node managed by the first end; based on the pre-stored feature information and direction information of the object to be processed, determine the leaf node reached by the object to be processed in the tree model; based on the weight information corresponding to the determined leaf node, generate the target score of the object to be processed in the tree model; and send the target score to the first end.

[0157] Computer program code for performing the operations of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages, such as Java, Smalltalk, C++, and conventional procedural programming languages, such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0158] The flow chart and block diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product according to various embodiments of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a module, a program segment or a part of a code, and the module, the program segment or a part of the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some implementations as replacements, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0159] The units involved in the embodiments described in the present disclosure may be implemented by software or hardware. The units described may also be set in a processor, for example, it may be described as: a processor includes a first acquisition unit, a determination unit, a generation unit, and a sending unit. The names of these units do not constitute a limitation on the units themselves in some cases, for example, the sending unit may also be described as a "unit for sending a target score".

[0160] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

Claims

1. A method for processing information, the method being executed by a second terminal among predetermined participating terminals, the participating terminals comprising a first terminal and a second terminal, the first terminal and the second terminal having the same tree model, the first terminal and the second terminal respectively managing some nodes in the tree model, the first terminal having the weight of a leaf node of the tree model, the method comprising: In response to receiving a processing request for an object to be processed sent by a first end, obtaining weight information and direction information from the first end, wherein the weight information is used to indicate the weight of a leaf node of the tree model, and the direction information is used to indicate the splitting direction of the object to be processed at the node managed by the first end; the object to be processed includes at least one of an image, text, video, and audio; Based on the pre-stored feature information of the object to be processed and the direction information, determining the leaf node reached by the object to be processed in the tree model; Based on the determined weight information corresponding to the leaf node, generating a target score of the object to be processed in the tree model; The target score is transmitted to the first end.

2. The method according to claim 1, wherein: The tree model includes at least two trees; as well as The pre-stored feature information and direction information of the object to be processed, determining the leaf node reached by the object to be processed in the tree model comprises: Based on the pre-stored feature information of the object to be processed and the direction information, determining at least two leaf child nodes reached by the object to be processed in the tree model; and The step of generating a target score of the object to be processed in the tree model based on the determined weight information corresponding to the leaf node includes: Based on the at least two pieces of weight information corresponding to the determined at least two leaf nodes, a target score of the object to be processed in the tree model is generated.

3. The method according to claim 2, wherein: The weight information includes an encrypted weight obtained by encrypting the weight of the leaf node of the tree model by the first end; as well as The step of generating a target score of the object to be processed in the tree model based on the at least two weight information corresponding to the at least two leaf nodes determined includes: At least two encrypted weights in the at least two weight information are summed to generate a target score of the object to be processed in the tree model.

4. The method according to claim 3, wherein: Before summing at least two encrypted weights in the at least two weight information to generate a target score of the object to be processed in the tree model, the method further includes: Acquire an encryption threshold sent by the first end, wherein the encryption threshold is obtained by encrypting a classification threshold by the first end, and the classification threshold is used to classify the object to be processed; and The summing up at least two encrypted weights in the at least two weight information to generate a target score of the object to be processed in the tree model comprises: Summing at least two encrypted weights in the at least two weight information to obtain an encrypted score of the object to be processed in the tree model; Get the target value pair; Using the target value pair, respectively perform affine transformation on the encrypted score and the encrypted threshold to obtain a target score corresponding to the encrypted score and a target threshold corresponding to the encrypted threshold; and The sending the target score to the first end comprises: The target score and the target threshold are sent to the first end.

5. The method according to any one of claims 1 to 4, wherein: The direction information is generated by the first end through the following steps: For the tree included in the tree model, the following processing steps are performed: Based on the nodes included in the tree, an initial bit string is generated, wherein the bits included in the initial bit string correspond to the nodes included in the tree, and the bits of the initial bit string do not include bit values; For the bits in the initial bit string corresponding to the tree, perform the following processing sub-steps: In response to the node corresponding to the bit being a node managed by the first end, determining a split direction of the object to be processed at the node corresponding to the bit; Selecting a candidate bit value corresponding to the determined splitting direction from the candidate bit value set as the bit value corresponding to the bit position; In response to the node corresponding to the bit position being a node managed by the second end, determining the preset bit value as the bit value corresponding to the bit position; Generate a bit string including bit values ​​corresponding to the tree; Direction information is generated based on the bit strings corresponding to the trees in the tree model.

6. The method according to claim 5, wherein: The generating direction information based on the bit string corresponding to the tree in the tree model includes: Converting the bit string corresponding to the tree in the tree model into a numerical value as a representation value; Based on the converted performance value, direction information is generated.

7. The method according to claim 1, wherein: After receiving the target score, the first end performs the following steps: Based on the target score, the object to be processed is classified.

8. A device for processing information, the device being configured at a second end of a predetermined participating end, the participating end comprising a first end and a second end, the first end and the second end having the same tree model, the first end and the second end respectively managing some nodes in the tree model, the first end having the weight of a leaf node of the tree model, the device comprising: a first acquisition unit, configured to, in response to receiving a processing request for an object to be processed sent by a first end, acquire weight information and direction information from the first end, wherein the weight information is used to indicate the weight of a leaf node of the tree model, and the direction information is used to indicate a splitting direction of the object to be processed at a node managed by the first end; the object to be processed includes at least one of an image, text, video, and audio; a determining unit configured to determine a leaf node reached by the object to be processed in the tree model based on pre-stored feature information of the object to be processed and the direction information; A generating unit configured to generate a target score of the object to be processed in the tree model based on the weight information corresponding to the determined leaf node; A sending unit is configured to send the target score to the first end.

9. The device according to claim 8, wherein: The tree model includes at least two trees; and The determining unit is further configured to: Based on the pre-stored feature information of the object to be processed and the direction information, determining at least two leaf nodes reached by the object to be processed in the tree model; and The generating unit is further configured to: Based on the at least two pieces of weight information corresponding to the determined at least two leaf nodes, a target score of the object to be processed in the tree model is generated.

10. The device according to claim 9, wherein: The weight information includes the encrypted weight obtained by encrypting the weight of the leaf node of the tree model by the first end; as well as The generating unit comprises: The generating module is configured to sum at least two encrypted weights in at least two weight information to generate a target score of the object to be processed in the tree model.

11. The device according to claim 10, wherein: The device also includes: A second acquisition unit is configured to acquire an encryption threshold sent by the first end, wherein the encryption threshold is obtained by the first end after encrypting a classification threshold, and the classification threshold is used to classify the object to be processed; and The generated modules include: A summing submodule is configured to sum at least two encrypted weights in at least two weight information to obtain an encrypted score of the object to be processed in the tree model; An acquisition submodule, configured to acquire a target value pair; a transformation submodule configured to use the target value pair to perform affine transformation on the encryption score and the encryption threshold, respectively, to obtain a target score corresponding to the encryption score and a target threshold corresponding to the encryption threshold; and The sending unit is further configured to: The target score and target threshold are sent to the first end.

12. The device according to any one of claims 8 to 11, wherein: The direction information is generated by the first end through the following steps: For trees included in a tree model, the following processing steps are performed: Based on the nodes included in the tree, an initial bit string is generated, wherein the bits included in the initial bit string correspond to the nodes included in the tree, and the bits of the initial bit string do not include bit values; For the bits in the initial bit string corresponding to the tree, perform the following processing sub-steps: In response to the node corresponding to the bit being a node managed by the first end, determining a splitting direction of the object to be processed at the node corresponding to the bit; Selecting a candidate bit value corresponding to the determined splitting direction from the candidate bit value set as the bit value corresponding to the bit position; In response to the node corresponding to the bit position being a node managed by the second end, determining the preset bit value as the bit value corresponding to the bit position; Generate a bit string including bit values ​​corresponding to the tree; Direction information is generated based on the bit strings corresponding to the trees in the tree model.

13. The device according to claim 12, wherein: The generating direction information based on the bit string corresponding to the tree in the tree model includes: Convert the bit string corresponding to the tree in the tree model into a numerical value as the representation value; Based on the converted performance value, direction information is generated.

14. The device according to claim 8, wherein: After receiving the target score, the first end performs the following steps: Based on the target score, the object to be processed is classified.

15. An electronic device, comprising: one or more processors; a storage device having one or more programs stored thereon, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

16. A computer readable medium having a computer program stored thereon, wherein: When the program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images