Security level identification method and device for mining target portraits by aggregating associated objects

Abstract

The present application relates to a method, device, electronic device and computer-readable medium for identifying the security level of aggregating local associated objects to mine target portraits. The method includes: obtaining an identification identifier of a target to be identified; extracting a relationship sub-network from a relationship network through a graph neural network sampling model based on the identification identifier; obtaining identification information of multiple neighbor nodes in the relationship sub-network; generating relationship feature data based on the identification information of the neighbor nodes through a graph neural network aggregation model; and generating a target portrait of the user based on the relationship feature data. The method, device, electronic device and computer-readable medium for identifying the security level of aggregating local associated objects to mine target portraits involved in the present application can accurately predict the target portrait only through the target's identification identifier, reduce data security risks brought about by information interaction with external data platforms, improve the overall security of the system, and reduce system operating costs.

Description

Technical Field

[0001] The present application relates to the field of computer information processing, and in particular to a method, device, electronic device and computer-readable medium for identifying security levels by aggregating local associated objects to mine target portraits. Background Art

[0002] Target portraits are an effective tool for outlining target objects and linking target demands with design directions. They are widely used in various fields. In practical applications, they often use simple, down-to-earth language to connect target attributes, behaviors, and desired data. As a virtual representation of the actual target, businesses or service providers can analyze target portraits to provide more comprehensive and accurate services to their target.

[0003] In the existing technology, the target portrait is often portrayed through the target relationship network. In the existing relationship network model, first, the relationship network is filtered out by the sub-network of each sample. Feature extraction is to calculate these data by aggregating the feature data of the neighbor platform objects, such as terminal information, account information, mutual evaluation information, transaction records, communication information, etc., to the node V, and finally obtain the neighbor features calculated to represent the node features. Then, with the help of the target data on the third-party platform, the classifier of the traditional model is established to predict the target portrait. In this process, it is necessary to manually set the mining features and to use the data on the third-party data platform. Since the target data is often exchanged with the third-party platform, on the one hand, the information transmission process also brings the risk of information leakage. On the other hand, obtaining the target data through the third-party platform will also consume a lot of time and business costs.

[0004] Therefore, a new method, device, electronic device and computer-readable medium for identifying the security level of aggregating local associated objects to mine target portraits are needed.

[0005] The above information disclosed in this Background section is only for enhancement of understanding of the background of the application and therefore it may contain information that does not form the prior art that is already known to a person of ordinary skill in the art. Summary of the Invention

[0006] In view of this, the present application provides a method, device, electronic device and computer-readable medium for identifying the security level of target portraits by aggregating local associated objects, which can accurately predict the target portrait only through the target's identification identifier, reduce the data security risks brought about by information interaction with external data platforms, improve the overall security of the system, and reduce the system operating costs.

[0007] Other features and advantages of the present application will become apparent from the following detailed description, or may be learned in part by practice of the present application.

[0008] According to one aspect of the present application, a method for identifying a security level by aggregating local associated objects to mine target portraits is proposed, the method comprising: obtaining an identification identifier of a target to be identified that does not belong to a local object; extracting a relationship subnetwork containing the target to be identified from a relationship network according to the identification identifier through a graph neural network sampling model; obtaining identification information of multiple neighbor nodes belonging to the local object in the relationship subnetwork; generating relationship feature data according to the identification information of the neighbor nodes through a graph neural network aggregation model; generating a target portrait of the target to be identified based on the relationship feature data; and determining the security level of the target to be identified based on the target portrait of the target to be identified.

[0009] Optionally, a relational sub-network containing the target to be identified is extracted from the relational network according to the identification identifier, including: extracting multiple neighbor nodes in the relational network according to the identification identifier through a graph neural network sampling model; eliminating neighbor nodes that do not belong to local objects in the neighbor nodes; and / or eliminating neighbor nodes corresponding to directed edges with preset directions in the neighbor nodes; and generating the relational sub-network through the target to be identified and the neighbor nodes.

[0010] Optionally, through a graph neural network aggregation model, relationship feature data is generated based on the identification information of the neighbor nodes, including: assigning weights to the edges between the neighbor nodes based on the identification information; dividing the multiple adjacent nodes into multiple levels based on the network structure of the relationship sub-network; and performing aggregation calculations on the neighbor nodes corresponding to the multiple levels through a graph neural network aggregation model to generate the relationship feature data.

[0011] Optionally, the neighbor nodes corresponding to multiple levels are aggregated and calculated through a graph neural network aggregation model to generate the relationship feature data, including: reading the neighbor nodes corresponding to each level one by one; aggregating the neighbor nodes in the current level through the graph neural network aggregation model to generate the layer relationship feature data of the current level; aggregating the neighbor nodes of the next level through the layer relationship feature data of the current level; and generating the relationship feature data after the neighbor nodes corresponding to all levels are calculated.

[0012] Optionally, the neighbor nodes in the current layer are aggregated through a graph neural network aggregation model to generate layer relationship feature data of the current layer, including: assigning weights to the edges between the neighbor nodes corresponding to the current layer according to the identification information; inputting the weights corresponding to the neighbor nodes corresponding to the current layer and the edges between the neighbor nodes into the graph neural network aggregation model; the graph neural network aggregation model performs aggregation calculations based on a multi-layer attention mechanism to generate layer relationship feature data of the current layer.

[0013] Optionally, the method further includes: acquiring multiple contact objects corresponding to multiple historical targets; generating association relationships between multiple historical targets and the multiple contact objects according to identification identifiers; and generating the relationship network through the historical targets, the contact objects and the association relationships therebetween.

[0014] Optionally, the relationship network is generated through the historical targets, the contact objects and the association relationships therebetween, including: treating the historical targets and the contact objects belonging to local objects as first-class nodes; treating the contact objects that do not belong to local objects as second-class nodes; generating directed edges through the first-class nodes, the second-class nodes and the association relationships therebetween; and generating the relationship network through multiple first-class nodes, multiple second-class nodes and their corresponding directed edges.

[0015] Optionally, obtaining identification information of multiple neighbor nodes belonging to local objects in the relationship subnetwork includes: obtaining identification information of multiple neighbor nodes in the relationship subnetwork based on data information of local objects recorded in a local business database; and performing data cleaning on the identification information.

[0016] According to one aspect of the present application, a security level identification device for aggregating local associated objects to mine target portraits is proposed, and the device includes: an identification module for obtaining an identification identifier of a target to be identified that does not belong to a local object; a sampling module for extracting a relationship sub-network containing the target to be identified from a relationship network according to the identification identifier through a graph neural network sampling model; an information module for obtaining identification information of multiple neighbor nodes belonging to local objects in the relationship sub-network; an aggregation module for generating relationship feature data according to the identification information of the neighbor nodes through a graph neural network aggregation model; a portrait module for generating a target portrait of the target to be identified based on the relationship feature data; and an identification module for determining the security level of the target to be identified based on the target portrait of the target to be identified.

[0017] According to one aspect of the present application, an electronic device is proposed, which includes: one or more processors; a storage device for storing one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the method as described above.

[0018] According to one aspect of the present application, a computer-readable medium is provided, on which a computer program is stored. When the program is executed by a processor, the method described above is implemented.

[0019] According to the security level identification method, device, electronic device and computer-readable medium of the present application for aggregating local associated objects to mine target portraits, an identification identifier of a target to be identified that does not belong to a local object is obtained; a relationship sub-network is extracted from a relationship network through a graph neural network sampling model based on the identification identifier; identification information of multiple neighbor nodes in the relationship sub-network is obtained; relationship feature data is generated based on the identification information of the neighbor nodes through a graph neural network aggregation model; a target portrait of the user is generated based on the relationship feature data, and the security level of the target to be identified is determined based on the target portrait of the target to be identified. In this way, the identification information of the local object can be obtained only through the identification identifier of the target to predict the target portrait, thereby reducing data security risks brought about by information interaction with external data platforms, improving the overall security of the system, and reducing system operating costs.

[0020] It should be understood that the foregoing general description and the following detailed description are merely illustrative and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The above and other objects, features, and advantages of the present application will become more apparent by describing in detail exemplary embodiments thereof with reference to the accompanying drawings. The drawings described below are merely some embodiments of the present application, and it is apparent to those skilled in the art that other drawings can be derived from these drawings without inventive effort.

[0022] Figure 1 It is a flowchart of a method for identifying security levels by aggregating local associated objects to mine target portraits according to an exemplary embodiment.

[0023] Figure 2 It is a schematic diagram of a security level identification method for aggregating local associated objects to mine target portraits according to another exemplary embodiment.

[0024] Figure 3 It is a schematic diagram of a security level identification method for aggregating local associated objects to mine target portraits according to another exemplary embodiment.

[0025] Figure 4 It is a flowchart of a method for identifying security levels by aggregating local associated objects to mine target portraits according to another exemplary embodiment.

[0026] Figure 5 It is a schematic diagram of a security level identification method for aggregating local associated objects to mine target portraits according to another exemplary embodiment.

[0027] Figure 6It is a schematic diagram of a security level identification method for aggregating local associated objects to mine target portraits according to another exemplary embodiment.

[0028] Figure 7 It is a block diagram of a security level identification device for aggregating local associated objects to mine target portraits according to an exemplary embodiment.

[0029] Figure 8 It is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0030] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be embodied in many forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete and will fully convey the concepts of the example embodiments to those skilled in the art. Like reference numerals in the drawings represent like or similar parts, and thus repetitive description thereof will be omitted.

[0031] The technical abbreviations involved in this application are explained as follows:

[0032] Neural network: A biologically inspired mathematical or computational model in the field of machine learning that mimics the structure and function of biological neural networks.

[0033] Graph neural network: refers to the general term for algorithms that use neural networks to learn graph-structured data, extract and discover features and patterns in graph-structured data, and meet the needs of graph learning tasks such as clustering, classification, prediction, segmentation, and generation.

[0034] Relationship network: A relationship network, sometimes also called a social network, refers to a collection of social actors and the relationships between them. Alternatively, a relationship network is a collection of points (social actors) and the connections between them (the relationships between actors).

[0035] Attention mechanism: If you have a superficial understanding of the Attention mechanism, the core logic is "from paying attention to everything to paying attention to key points" and focusing on key points by assigning weights.

[0036] Embedding: A common encoding method in deep learning that represents a value using a set of vectors.

[0037] Figure 1 The flowchart of a method for identifying a security level by aggregating local associated objects to mine target portraits according to an exemplary embodiment is shown. The method 10 for identifying a security level by aggregating local associated objects to mine target portraits includes at least steps S102 to S110.

[0038] like Figure 1 As shown, in S102, an identification identifier of a target to be identified that is not a local object is obtained. In an embodiment of the present application, the target to be identified may be an individual user, an enterprise user, or a computer node in a computing service cluster, and the local object may be an individual user, an enterprise user, or a computer node in a computing service cluster that has been stored or recorded locally in the system.

[0039] In this embodiment, new users, corporate users who have just logged in or registered, or computer nodes that have newly connected to the computer service cluster need to be identified for security risks to ensure the overall security of the system. However, since the target to be identified has not yet generated any operation information locally, it is impossible to directly determine the security status of the target to be identified. Obtaining the target to be identified through a third party is prone to data loss on the one hand, and on the other hand it will also consume a lot of time and operating costs.

[0040] In one embodiment, when the target to be identified is a user, the identification identifier may be the user's account information, which may be an account ID in a social network or a user identifier on an online instant messaging platform.

[0041] In one embodiment, when the target to be identified is a computer node in a cluster service, the identification identifier may be a hardware identification code of the computer node, or a network name that can uniquely locate the node in the service cluster.

[0042] In S104, a graph neural network sampling model is used to extract a relational subnetwork containing the target to be identified from the relational network based on the identification identifier. For example, the graph neural network sampling model can be used to extract multiple neighboring nodes from the relational network based on the identification identifier; neighboring nodes that do not meet the node strategy are removed from the neighboring nodes; and the relational subnetwork is generated by combining the removed neighboring nodes and the target to be identified.

[0043] More specifically, removing neighbor nodes that do not meet the node policy from the neighbor nodes may be removing neighbor nodes that do not belong to local objects from the neighbor nodes; and / or removing neighbor nodes corresponding to directed edges with preset directions from the neighbor nodes.

[0044] In a specific application scenario, a graph neural network sampling model (GraphSAGE) can be used to extract multiple neighbor nodes from a relational network to generate a subnetwork. In existing technologies, GNNs or GCNs are often used to aggregate neighbor relationships in relational networks. When aggregating neighbor relationships in a relational network, GNNs or GCNs use all neighboring nodes to update a graph network's own nodes. In real-world scenarios, the computational methods used in existing technologies are very difficult, both in terms of resources and in terms of real-time updating of node embedding vectors.

[0045] In this application, the graph neural network sampling model can sample the subnetwork of the node, and can set the relationship depth and the number of neighbors calculated when the node vector is updated as needed. This greatly reduces the resource requirements while controlling the composition size of each node.

[0046] In S106, identification information of multiple neighboring nodes belonging to the local object in the relationship sub-network is obtained. The identification information of the multiple neighboring nodes in the relationship sub-network can be obtained based on the data information of the local object recorded in the local service database; and data cleaning is performed on the identification information. Data cleaning of the identification information can reduce the impact of interference data in the identification information on the present solution and improve recognition accuracy.

[0047] Among them, when the target object is a user, the identification information may include basic information authorized by the user, such as business account information, user terminal device identification information, user location information, etc.; user information may also include behavioral information, such as user page operation data, user service access duration, user service access frequency, etc. The specific content of user information can be determined according to the actual application scenario and is not limited here. More specifically, the user information of the current user can be obtained by embedding points on the web page based on user authorization.

[0048] Among them, when the target object is a node in a computing service cluster, the identification information may include the node's hardware identification, node IP address, node installation software name, node storage capacity and other information. The identification information may also include behavioral information, such as the amount of data processed by the computing node, interactive reading information, business access information, etc.

[0049] In S108, a graph neural network aggregation model is used to generate relationship feature data based on the identification information of the neighboring nodes; and a target profile of the target to be identified is generated based on the relationship feature data. Weights may be assigned to the edges between the neighboring nodes based on the identification information; the multiple adjacent nodes are divided into multiple levels based on the network structure of the relationship subnetwork; and the graph neural network aggregation model is used to perform aggregation calculations on the neighboring nodes corresponding to the multiple levels to generate the relationship feature data.

[0050] Figure 2 It is the general process of extracting relational feature data in the existing technology. Figure 3 This is the process of using graph neural networks to extract and generate relational feature data. In this application, the graph neural network model is used for both sub-relationship extraction and feature extraction. The feature dimension is converted through embedding, which greatly reduces the time of data processing and the process of manual subjective selection.

[0051] The details of "generating relationship feature data based on the identification information of the neighbor nodes through the graph neural network aggregation model" will be Figure 6 The corresponding embodiments are described in detail.

[0052] In this embodiment, the target to be identified is predicted and analyzed using the above-mentioned relationship feature data to generate a target portrait.

[0053] According to the security level identification method of the target portrait mined by aggregating local associated objects of the present application, the identification identifier of the target to be identified is obtained; a relationship sub-network is extracted from the relationship network according to the identification identifier through a graph neural network sampling model; identification information of multiple neighbor nodes in the relationship sub-network is obtained; relationship feature data is generated according to the identification information of the neighbor nodes through a graph neural network aggregation model; and a target portrait of the user is generated according to the relationship feature data. This method can accurately predict the target portrait only through the identification identifier of the target, reduce data security risks brought about by information interaction with external data platforms, improve the overall security of the system, and reduce system operating costs.

[0054] In S110 , the security level of the target to be identified is determined according to the target portrait of the target to be identified.

[0055] In this embodiment, when the identification information of the target to be identified is not obtained, the relationship feature data of the target to be identified is constructed through the identification information of the local object associated with the target to be identified that has been stored or recorded locally, and a target portrait reflecting the target to be identified is obtained from the side, and then the security level of the target to be identified is determined, thereby improving the identification efficiency and reducing the risk of data loss.

[0056] In a specific embodiment, in a financial services scenario, when the target to be identified is a new user, such as a user who accesses through a third-party link, and only the user's social account is known, the technical solution of this application can be used to predict the user's security level before the user applies, using only the user's social account. From a business cost perspective, the cost of obtaining data from third parties can be reduced; from a time-effectiveness perspective, the constructed graph network can be used to predict the profile or risks of customers who have not yet become users in the network.

[0057] It should be clearly understood that this application describes how to form and use specific examples, but the principles of this application are not limited to any details of these examples. On the contrary, based on the teaching of the content disclosed in this application, these principles can be applied to many other embodiments.

[0058] Figure 4 It is a flowchart of a method for identifying security levels by aggregating local associated objects to mine target portraits according to another exemplary embodiment. Figure 4 The process 40 shown is for Figure 1 Supplementary description of the process shown.

[0059] like Figure 4 As shown, in S402, multiple contact objects corresponding to multiple historical targets are obtained. The following describes the generation process of the relationship network through a specific application scenario of an Internet network service platform.

[0060] In S404, associations between the multiple historical targets and the multiple contact objects are generated based on the identification identifiers. An association between an existing user and a new user (non-platform user) can be generated using account information in a social network. For example, if a user shares a platform activity with a new user, and the new user clicks on the activity link to enter the platform, then an association can be considered between the two.

[0061] In S406, the relationship network is generated by using the multiple historical targets, the multiple contact objects, and the relationships therebetween. The local platform objects (local objects) among the multiple historical targets and the multiple contact objects can be treated as first-type nodes; non-local platform objects among the multiple contact objects can be treated as second-type nodes; directed edges are generated by using the first-type nodes, the second-type nodes, and the relationships therebetween; and the directed relationship network is generated by using the multiple first-type nodes, the multiple second-type nodes, and their corresponding directed edges.

[0062] Specifically, users on the platform can be regarded as the first type of nodes in the relationship network, and non-platform users can be regarded as the second type of nodes. Figure 5 As mentioned, Figure 5 In the figure, the first type of nodes are blue nodes, the second type of nodes are gray nodes, and the association relationship between the first type of nodes and the second type of nodes is a directed edge.

[0063] In the relationship network of this application, the network relationship is based only on reverse association relationships. The so-called reverse relationship means that when non-user B is obtained from user A's social relationship network, then with user A as the node, the direction from user A to non-user B is forward, while with non-user B as the node, the direction from user A to non-user B is reverse. The advantage of using a reverse network is that before non-user B becomes a platform user, the existence of non-user B as a node has been obtained from user A's social relationship network using the technology in this application. Then, all users with non-user B in the platform's social relationship network can be used to form a relationship network with B as the node and platform user A and other users as neighbors. When the characteristics of platform users are used to assess the risk of non-user B, the assessment time is advanced compared to the timeliness of the assessment when B becomes a user. The ability to assess the time point in advance not only allows the risk of B to be characterized in advance, but also reduces the cost of third-party data queries on B because the assessment is based on the historical information of user A.

[0064] In one embodiment, weights may be assigned to the edges between the neighbor nodes according to the identification information; the plurality of adjacent nodes may be divided into a plurality of levels according to the network structure of the relationship sub-network; Figure 6 As shown, the neighbor nodes corresponding to multiple levels can be aggregated and calculated through the graph neural network aggregation model to generate the relationship feature data.

[0065] Among them, the neighbor nodes corresponding to multiple levels are aggregated and calculated through the graph neural network aggregation model (GAT) to generate the relationship feature data, including: reading the neighbor nodes corresponding to each level one by one; aggregating the neighbor nodes in the current level through the graph neural network aggregation model to generate the layer relationship feature data of the current level; aggregating the neighbor nodes of the next level through the layer relationship feature data of the current level; and generating the relationship feature data after the neighbor nodes corresponding to all levels are calculated.

[0066] In the prior art, graph neural network aggregation is based on an assumption: all neighbors are equally important to a node. In practical applications, this aggregation approach is unconventional. Within each user's relationships, relationships are unlikely to be equally important. For example, a family relationship is much more important than a following user with whom one has never communicated, or a user with whom one has briefly interacted and commented. Therefore, simply using equal aggregation is clearly not suitable for such network relationships. Therefore, this application introduces GAT, which incorporates a multi-head attention mechanism when calculating aggregation, to address the limitations of graph neural network aggregation.

[0067] More specifically, weights can be assigned to the edges between neighbor nodes corresponding to the current layer based on the identification information; the weights corresponding to the neighbor nodes corresponding to the current layer and the edges between neighbor nodes are input into the graph neural network aggregation model; the graph neural network aggregation model performs aggregation calculations based on the multi-layer attention mechanism to generate layer relationship feature data of the current layer.

[0068] The security level identification method for mining target portraits by aggregating local associated objects in this application has the following advantages in practical application: the relationship network is established as a reverse communication relationship network. That is, before non-platform user A becomes a platform user, the associated users {a, b, c...} in A's social relationships can be used to profile A, and if the associated users are already platform users, this information can be used to assess A's various risks in advance.

[0069] In terms of attribute feature extraction, only neighbor attributes are extracted and aggregated into node attribute features through calculation. Therefore, in this model, the prediction target information itself is not used. In other words, the model variables use the features of the associated users {a, b, c...} instead of A itself, which can effectively reduce the cost of querying third-party credit data in the business.

[0070] Those skilled in the art will appreciate that all or part of the steps implementing the above embodiments can be implemented as a computer program executed by a CPU. When executed by the CPU, the computer program performs the functions defined in the above method provided herein. The program can be stored in a computer-readable storage medium, such as a read-only memory, a magnetic disk, or an optical disk.

[0071] Furthermore, it should be noted that the aforementioned figures are merely illustrative of the processes included in the methods according to exemplary embodiments of the present application and are not intended to be limiting. It is readily understood that the processes illustrated in the aforementioned figures do not indicate or limit the temporal order of these processes. Furthermore, it is readily understood that these processes may be executed synchronously or asynchronously, for example, in multiple modules.

[0072] The following are device embodiments of the present application, which can be used to implement the method embodiments of the present application. For details not disclosed in the device embodiments of the present application, please refer to the method embodiments of the present application.

[0073] Figure 7 This is a block diagram of a security level identification device for mining target portraits by aggregating local associated objects according to an exemplary embodiment. Figure 7 As shown, the security level identification device 70 for aggregating local associated objects to mine target portraits includes: an identification module 702, a sampling module 704, an information module 706, an aggregation module 708, and a portrait module 710.

[0074] The identification module 702 is used to obtain the identification identifier of the target to be identified that does not belong to the local object;

[0075] The sampling module 704 is configured to extract a relational subnetwork containing the target to be identified from the relational network according to the identification identifier using a graph neural network sampling model;

[0076] The information module 706 is used to obtain identification information of a plurality of neighbor nodes belonging to the local object in the relationship sub-network;

[0077] The aggregation module 708 is used to generate relationship feature data based on the identification information of the neighbor nodes through a graph neural network aggregation model;

[0078] The portrait module 710 is used to generate a target portrait of the target to be identified based on the relationship feature data;

[0079] The identification module 712 is used to determine the security level of the target to be identified based on the target portrait of the target to be identified.

[0080] The sampling module 704 is specifically used to extract multiple neighbor nodes in the relationship network according to the identification identifier through a graph neural network sampling model; eliminate neighbor nodes that do not belong to local objects in the neighbor nodes; and / or eliminate neighbor nodes corresponding to directed edges with preset directions in the neighbor nodes; and generate the relationship sub-network through the target to be identified and the neighbor nodes.

[0081] Aggregation module 708 is specifically used to assign weights to the edges between the neighbor nodes according to the identification information; divide the multiple adjacent nodes into multiple levels according to the network structure of the relationship sub-network; and perform aggregation calculations on the neighbor nodes corresponding to the multiple levels through the graph neural network aggregation model to generate the relationship feature data.

[0082] Aggregation module 708 is specifically used to read the neighbor nodes corresponding to each level one by one; aggregate the neighbor nodes in the current level through the graph neural network aggregation model to generate the layer relationship feature data of the current level; aggregate the neighbor nodes of the next level through the layer relationship feature data of the current level; and generate the relationship feature data after the neighbor nodes corresponding to all levels are calculated.

[0083] Aggregation module 708 is specifically used to assign weights to the edges between neighbor nodes corresponding to the current layer according to the identification information; input the weights corresponding to the neighbor nodes corresponding to the current layer and the edges between the neighbor nodes into the graph neural network aggregation model; the graph neural network aggregation model performs aggregation calculation based on the multi-layer attention mechanism to generate layer relationship feature data of the current layer.

[0084] The identification device also includes: a relationship network generation module, which is used to obtain multiple contact objects corresponding to multiple historical targets; generate association relationships between multiple historical targets and the multiple contact objects based on the identification identifier; and generate the relationship network through the historical targets, the contact objects and the association relationships therebetween.

[0085] The relationship network generation module is specifically used to treat the historical target and the contact object belonging to the local object as the first type of nodes; treat the contact object that does not belong to the local object as the second type of node; generate directed edges through the first type of nodes, the second type of nodes and the association relationships between them; and generate the relationship network through multiple first type nodes, multiple second type nodes and their corresponding directed edges.

[0086] The information module 706 is specifically configured to obtain identification information of a plurality of neighbor nodes in the relationship sub-network according to data information of the local object recorded in the local service database; and perform data cleaning on the identification information.

[0087] According to the security level identification device for mining target portraits by aggregating local associated objects of the present application, an identification identifier of the target to be identified is obtained; a relationship sub-network is extracted from the relationship network through a graph neural network sampling model according to the identification identifier; identification information of multiple neighbor nodes in the relationship sub-network is obtained; relationship feature data is generated according to the identification information of the neighbor nodes through a graph neural network aggregation model; and a target portrait of the user is generated according to the relationship feature data. This method can accurately predict the target portrait only through the identification identifier of the target, reduce data security risks brought about by information interaction with external data platforms, improve the overall security of the system, and reduce system operating costs.

[0088] like Figure 8 As shown, an embodiment of the present invention provides an electronic device, including a processor 810, a communication interface 820, a memory 830 and a communication bus 840, wherein the processor 810, the communication interface 820, and the memory 830 communicate with each other through the communication bus 840;

[0089] Memory 830, for storing computer programs;

[0090] The processor 810 is configured to implement the data allocation authority adjustment method based on video facial expressions of any of the above embodiments when executing the program stored in the memory 830 .

[0091] In an electronic device provided by an embodiment of the present invention, a processor 810 obtains the target's initial data allocation permissions and access information by executing a program stored on a memory 830; determines the video text content through the access information; establishes a real-time video link with the target, and displays the video text content based on the video connection to generate video data; recognizes the user's facial expressions in the video data to determine the corresponding permission adjustment coefficient; and adjusts the user's data allocation permissions based on the initial permissions and the permission adjustment coefficient.

[0092] The communication bus 840 mentioned in the electronic device can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The communication bus 840 can be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, the figure shows only one thick line, but this does not mean that there is only one bus or only one type of bus.

[0093] The communication interface 820 is used for communication between the electronic device and other devices.

[0094] The memory 830 may include a random access memory 830 (RAM) or a non-volatile memory 830, such as at least one disk storage 830. Alternatively, the memory 830 may be at least one storage device located away from the processor 810.

[0095] The above-mentioned processor 810 can be a general-purpose processor 810, including a central processing unit 810 (CPU), a network processor 810 (NP), etc.; it can also be a digital signal processor 810 (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components.

[0096] An embodiment of the present invention provides a computer-readable storage medium storing one or more programs, wherein the one or more programs can be executed by one or more processors to implement the security level identification method for mining target profiles by aggregating local associated objects according to any of the above embodiments. For example, the method may include obtaining an identification identifier of a target to be identified; extracting a relational subnetwork from a relational network based on the identification identifier using a graph neural network sampling model; obtaining identification information of multiple neighboring nodes in the relational subnetwork; generating relational feature data based on the identification information of the neighboring nodes using a graph neural network aggregation model; and generating a target profile of the user based on the relational feature data.

[0097] In the above embodiments, all or part of the embodiments can be implemented by software, hardware, firmware, or any combination thereof. When implemented using software, all or part of the embodiments can be implemented in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions according to the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired (e.g., coaxial cable, optical fiber, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) method. The computer-readable storage medium can be any available medium that can be accessed by a computer or a data storage device such as a server or data center that includes one or more available media. The available medium can be a magnetic medium (e.g., a floppy disk, a hard disk, a tape), an optical medium (e.g., a DVD), or a semiconductor medium (e.g., a solid-state drive (SSD)).

[0098] While the exemplary embodiments of the present application have been specifically illustrated and described above, it should be understood that the present application is not limited to the detailed structures, configurations, or implementations described herein; rather, the present application is intended to encompass various modifications and equivalent configurations within the spirit and scope of the appended claims.

Claims

1. A security level identification method for mining target portraits by aggregating local associated objects, characterized in that: include: Get multiple contact objects corresponding to multiple historical targets; generating association relationships between the plurality of historical targets and the plurality of contact objects according to the identification identifier; The historical target and the contact object belonging to the local object are regarded as first-type nodes; Treating the contact object that is not a local object as a second type of node; Generate directed edges through the first type of nodes, the second type of nodes and the association relationships between them; Generate a relationship network through multiple first-type nodes, multiple second-type nodes and their corresponding directed edges; Obtaining the identification identifier of the target to be identified that does not belong to the local object; Extracting a relational subnetwork containing the target to be identified from the relational network according to the identification identifier using a graph neural network sampling model; Obtaining identification information of a plurality of neighbor nodes belonging to the local object in the relationship subnetwork; Generate relationship feature data based on the identification information of the neighbor nodes through a graph neural network aggregation model; generating a target profile of the target to be identified based on the relationship feature data; The security level of the target to be identified is determined based on the target profile of the target to be identified.

2. The method according to claim 1, wherein By using a graph neural network sampling model, a relational subnetwork containing the target to be identified is extracted from the relational network according to the identification identifier, including: Extracting a plurality of neighbor nodes from the relationship network according to the identification mark using a graph neural network sampling model; Eliminate neighbor nodes that do not belong to local objects from the neighbor nodes; and / or eliminate neighbor nodes corresponding to directed edges with preset directions from the neighbor nodes; The relationship sub-network is generated by using the target to be identified and the neighboring nodes.

3. The method according to claim 1, wherein Generate relationship feature data based on the identification information of the neighbor nodes through a graph neural network aggregation model, including: assigning weights to edges between the neighbor nodes according to the identification information; Dividing the plurality of adjacent nodes into a plurality of levels according to the network structure of the relationship sub-network; Aggregation calculations are performed on neighbor nodes corresponding to multiple levels through a graph neural network aggregation model to generate the relationship feature data.

4. The method according to claim 3, wherein Aggregate and calculate neighbor nodes corresponding to multiple levels through a graph neural network aggregation model to generate the relationship feature data, including: Read the neighbor nodes corresponding to each level one by one; Aggregate neighbor nodes in the current layer through the graph neural network aggregation model to generate layer relationship feature data of the current layer; Aggregate neighbor nodes in the next layer through the layer relationship feature data of the current layer; After the neighbor nodes corresponding to all levels are calculated, the relationship feature data is generated.

5. The method according to claim 4, wherein The neighbor nodes in the current layer are aggregated through the graph neural network aggregation model to generate the layer relationship feature data of the current layer, including: Assign weights to the edges between neighbor nodes corresponding to the current layer based on the identification information; Input the weights corresponding to the neighbor nodes corresponding to the current layer and the edges between the neighbor nodes into the graph neural network aggregation model; The graph neural network aggregation model performs aggregation calculations based on a multi-layer attention mechanism to generate layer relationship feature data of the current level.

6. The method according to claim 1, wherein Obtaining identification information of a plurality of neighbor nodes belonging to the local object in the relationship subnetwork includes: Acquire identification information of a plurality of neighbor nodes in the relationship subnetwork according to data information of the local object recorded in the local service database; Perform data cleansing on identifying information.

7. A security level identification device for aggregating local associated objects to mine target portraits, characterized in that: include: A relationship network generation module is used to obtain multiple contact objects corresponding to multiple historical targets; Generate association relationships between multiple historical targets and the multiple contact objects based on the identification identifiers; treat the historical targets and the contact objects belonging to the local objects as first-type nodes; treat the contact objects that do not belong to the local objects as second-type nodes; generate directed edges through the first-type nodes, the second-type nodes, and the association relationships therebetween; and generate a relationship network through the multiple first-type nodes, the multiple second-type nodes, and their corresponding directed edges; An identification module, used to obtain an identification identifier of a target to be identified that does not belong to a local object; A sampling module, configured to extract a relational subnetwork containing the target to be identified from the relational network according to the identification identifier using a graph neural network sampling model; An information module, configured to obtain identification information of a plurality of neighbor nodes belonging to a local object in the relationship subnetwork; An aggregation module, configured to generate relationship feature data based on the identification information of the neighbor nodes through a graph neural network aggregation model; A portrait module, configured to generate a target portrait of the target to be identified based on the relationship feature data; The identification module is used to determine the security level of the target to be identified based on the target portrait of the target to be identified.

8. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs; 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 6.

9. A computer-readable medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 6 is implemented.

Images