WO2021189925A1 - 保护用户隐私的风险检测方法和装置 - Google Patents

保护用户隐私的风险检测方法和装置 Download PDF

Info

Publication number
WO2021189925A1
WO2021189925A1 PCT/CN2020/132868 CN2020132868W WO2021189925A1 WO 2021189925 A1 WO2021189925 A1 WO 2021189925A1 CN 2020132868 W CN2020132868 W CN 2020132868W WO 2021189925 A1 WO2021189925 A1 WO 2021189925A1
Authority
WO
WIPO (PCT)
Prior art keywords
subgraph
graph
node
terminal device
picture
Prior art date
Application number
PCT/CN2020/132868
Other languages
English (en)
French (fr)
Inventor
石磊磊
熊涛
Original Assignee
支付宝(杭州)信息技术有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 支付宝(杭州)信息技术有限公司 filed Critical 支付宝(杭州)信息技术有限公司
Publication of WO2021189925A1 publication Critical patent/WO2021189925A1/zh

Links

Images

Classifications

    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04WWIRELESS COMMUNICATION NETWORKS
    • H04W12/00Security arrangements; Authentication; Protecting privacy or anonymity
    • H04W12/12Detection or prevention of fraud
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F21/00Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
    • G06F21/60Protecting data
    • G06F21/62Protecting access to data via a platform, e.g. using keys or access control rules
    • G06F21/6218Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
    • G06F21/6245Protecting personal data, e.g. for financial or medical purposes
    • HELECTRICITY
    • H04ELECTRIC COMMUNICATION TECHNIQUE
    • H04WWIRELESS COMMUNICATION NETWORKS
    • H04W12/00Security arrangements; Authentication; Protecting privacy or anonymity
    • H04W12/12Detection or prevention of fraud
    • H04W12/121Wireless intrusion detection systems [WIDS]; Wireless intrusion prevention systems [WIPS]

Definitions

  • One or more embodiments of this specification relate to the computer field, and in particular to a risk detection method and device for protecting user privacy.
  • Graph computing is often used to perform risk detection on this type of data, such as anti-money laundering, gambling detection, MLM detection, cheating detection, fraud detection, etc. .
  • the global graph calculation scheme is adopted, and the risk is detected by the global graph on the server.
  • the process of calculating the risk assessment result and transmitting the risk assessment result on the server there is a risk of user risk data privacy leakage.
  • One or more embodiments of this specification describe a risk detection method and device for protecting user privacy, which can effectively protect user privacy.
  • a risk detection method for protecting user privacy includes: a first terminal device determines a first sub-picture according to associated information of a first user corresponding to the first terminal device.
  • the subgraph includes a central node and associated nodes of the central node, the central node corresponds to the first user, and the associated node corresponds to one or more second users that have an associated relationship with the first user;
  • the first terminal device sends the first sub-picture to the server, so that the server updates the global picture according to the first sub-picture, and determines the picture embedding information of the first sub-picture according to the updated global picture ,
  • the graph embedding information includes at least the node feature vector corresponding to each node in the first subgraph; wherein, the global graph is established according to the subgraphs sent by multiple terminal devices;
  • the server receives the graph embedding information of the first subgraph; the first terminal device determines the risk assessment result corresponding to the first subgraph according to the graph embedding information, and the risk assessment
  • the associated information includes transaction information or social activity information.
  • the determining the first subgraph includes: collecting the association relationship records of the first user to generate the first subgraph; or, according to the association relationship records newly generated by the first user in the most recent predetermined period of time, Update the first subgraph that has been generated.
  • the graph embedding information further includes an edge feature vector corresponding to each connected edge in the first subgraph; the first terminal device determines the graph embedding information according to the graph embedding information.
  • the risk assessment result corresponding to the first subgraph includes: the first terminal device determines the risk assessment result corresponding to the first subgraph according to the feature vector of each node and the feature vector of each edge.
  • the first terminal device determining the risk assessment result corresponding to the first sub-graph according to the graph embedding information includes: the first terminal device at least according to each node feature vector, A pre-trained classification model or regression model is used to determine the risk assessment result corresponding to the first subgraph.
  • a risk detection method for protecting user privacy includes: a server receives a first subgraph from a first terminal device corresponding to the first user, the first subgraph including a central node and the central The associated node of the node, the central node corresponds to the first user, and the associated node corresponds to one or more second users that have an associated relationship with the first user; the server updates according to the first subgraph Global graph, and determine the graph embedding information of the first subgraph according to the updated global graph, the graph embedding information includes at least the node feature vector corresponding to each node in the first subgraph; wherein, all The global graph is established based on subgraphs sent by multiple terminal devices; the server sends the graph embedding information of the first subgraph to the first terminal device, so that the first terminal device embeds the graph according to the graph Information to determine the risk assessment result corresponding to the first subgraph, and the risk assessment result is used to make a risk decision.
  • the server updating the global graph according to the first subgraph includes: the server updating the graph structure of the global graph according to the graph structure of the first subgraph; and/or The server updates the node attributes of the target nodes in the global graph according to the node attributes of the target nodes in the first subgraph; and/or, the server updates the node attributes of the target nodes in the global graph according to the The edge attribute updates the edge attribute of the target connecting edge of the global graph.
  • the server determining the graph embedding information of the first subgraph according to the updated global graph includes: the server based on the graph neural network according to the updated global graph A graph neural network (GNN) algorithm or a graph embedding Node2Vec algorithm determines the graph embedding information of the first subgraph.
  • GNN graph neural network
  • the graph embedding information further includes an edge feature vector corresponding to each connected edge in the first subgraph.
  • a risk detection device for protecting user privacy.
  • the device is provided in a first terminal device, and the device includes: a determining unit configured to determine the associated information of the first user corresponding to the first terminal device. , Determine a first subgraph, the first subgraph includes a central node and associated nodes of the central node, the central node corresponds to the first user, and the associated node corresponds to the first user having an association relationship One or more of the second users; the sending unit is configured to send the first sub-picture determined by the determining unit to the server, so that the server updates the global picture according to the first sub-picture, and according to the updated all
  • the global graph determines the graph embedding information of the first sub graph, and the graph embedding information includes at least the node feature vector corresponding to each node in the first sub graph; wherein, the global graph is sent according to multiple terminal devices
  • the receiving unit is configured to receive the picture embedding information of the first sub-picture from the server; the evaluating unit is configured to determine the
  • a risk detection device for protecting user privacy.
  • the device is set in a server, and the device includes: a receiving unit configured to receive a first sub-picture from a first terminal device corresponding to the first user.
  • the first subgraph includes a central node and associated nodes of the central node, the central node corresponds to the first user, and the associated node corresponds to one or more second users that have an associated relationship with the first user;
  • the updating unit is configured to update the global graph according to the first subgraph received by the receiving unit, and determine the graph embedding information of the first subgraph according to the updated global graph, and the graph embedding information includes at least the The node feature vector corresponding to each node in the first subgraph; wherein the global graph is established according to the subgraphs sent by multiple terminal devices; the sending unit is configured to send the update unit determination to the first terminal device The graph embedding information of the first subgraph, so that the first terminal device determines the risk assessment result corresponding to the first subgraph according to the graph embedd
  • a computer-readable storage medium on which a computer program is stored, and when the computer program is executed in a computer, the computer is caused to execute the method of the first or second aspect.
  • a computing device including a memory and a processor, the memory stores executable code, and the processor implements the method of the first or second aspect when the executable code is executed by the processor.
  • the first terminal device does not directly send the user's associated information to the server, but determines the first child based on the first user's associated information corresponding to the first terminal device.
  • the first sub-picture is then sent to the server.
  • the server updates the global picture according to the first sub-picture, and determines the picture embedding information of the first sub-picture according to the updated global picture.
  • the server does not determine the first sub-picture according to the picture embedding information.
  • the first terminal device determines the risk assessment result corresponding to the first sub-graph based on the image embedded information, and the risk assessment result is used to make risk decisions. .
  • the embodiment of this specification uses a graph calculation method combining terminal equipment and a server to detect risks, and improves usability, timeliness, and computing resources.
  • the server since the server is only responsible for generating graphs and embedding information, the risk assessment process and results are all on the terminal device, which effectively protects user privacy.
  • FIG. 1 is a schematic diagram of an implementation scenario of an embodiment disclosed in this specification
  • FIG. 2 shows a schematic diagram of interaction of a risk detection method for protecting user privacy according to an embodiment
  • Fig. 3 shows a schematic diagram of a graph structure of a first subgraph according to an embodiment
  • Fig. 4 shows a schematic diagram of a diagram structure of a global diagram according to an embodiment
  • Fig. 5 shows a schematic block diagram of a risk detection device for protecting user privacy according to an embodiment
  • Fig. 6 shows a schematic block diagram of a risk detection device for protecting user privacy according to another embodiment.
  • Figure 1 is a schematic diagram of an implementation scenario of an embodiment disclosed in this specification. This implementation scenario involves risk detection to protect user privacy.
  • the server has a communication connection with multiple terminal devices. The number of the above multiple terminal devices may be relatively large. Figure 1 only shows three terminal devices as an example. Each terminal device can be associated with its corresponding user. The information establishes the sub-graph, and each terminal device sends the sub-graphs that it has established to the server, and the server establishes the global graph according to the received sub-graphs. It can be understood that both the sub-graph and the global graph are relational network graphs.
  • the association information of the user corresponding to the terminal device changes over time. For example, in the first time period, user A and user B do not have an association relationship, and in the second time period after the first time period, user A and user B have Association relationship, the change of such associated information will be correspondingly reflected as the change of the sub-graph, which will cause the change of the global map accordingly.
  • the first terminal device sends the determined first sub-map to the server, so that the server updates the global map according to the first sub-map, and according to the updated all
  • the global graph determines the graph embedding information of the first sub graph
  • the server sends the graph embedding information to the first terminal device
  • the first terminal device determines the risk assessment result corresponding to the first sub graph according to the graph embedding information, and the risk assessment
  • the results are used to make risk decisions.
  • the above-mentioned risk assessment result is determined by the terminal device, and the server does not know the risk assessment result, and the risk assessment result is not easy to leak, thereby effectively protecting user privacy.
  • the embodiments of this specification do not limit the type of the foregoing server, and the foregoing server may refer to a physical server or a cloud server.
  • the cloud server can provide a simple, efficient, safe and reliable computing service with elastically scalable processing capabilities. Its management method is simpler and more efficient than physical servers. Users can quickly create or release any number of cloud servers without purchasing hardware in advance.
  • Fig. 2 shows an interactive schematic diagram of a risk detection method for protecting user privacy according to an embodiment.
  • the method can be implemented by a combination of a first terminal device and a server based on the implementation scenario shown in Fig. 1.
  • the risk detection method for protecting user privacy in this embodiment includes steps 21-25.
  • the first terminal device determines a first subgraph according to the association information of the first user corresponding to the first terminal device, and the first subgraph includes the central node and the associated nodes of the central node.
  • the central node corresponds to the first user
  • the associated node corresponds to one or more second users that have an associated relationship with the first user.
  • the first subgraph is a relational network graph, which is composed of nodes and connecting edges between nodes. Each node corresponds to a user.
  • the central node may have one or more associated nodes, and the associated node and the central node With connecting edges.
  • FIG. 3 shows a schematic diagram of the graph structure of the first subgraph according to an embodiment.
  • node 1 is a central node
  • node 2 node 3
  • node 4 node 5
  • node 6 are associated nodes of node 1.
  • the associated information includes transaction information or social activity information.
  • association information is not limited to this, and any other information that can reflect the association relationship between users can be used as the association information to construct the relationship network diagram.
  • the determining the first subgraph includes: collecting the association relationship record of the first user to generate the first subgraph; or, updating the generated association relationship record according to the newly generated association relationship record of the first user in the most recent predetermined period of time. The first subgraph.
  • the first terminal device may determine the first subgraph periodically, or determine the first subgraph when there is a risk detection requirement.
  • the first terminal device sends the first sub-picture to the server. It can be understood that the first terminal device may send one or more of the graph structure, node attributes, and edge attributes of the first subgraph to the server.
  • the first terminal device may periodically send the first subgraph to the server, or send the first subgraph to the server when there is a risk detection requirement.
  • the server updates the global graph according to the first subgraph, and determines the graph embedding information of the first subgraph according to the updated global graph.
  • the graph embedding information includes at least the respective corresponding nodes in the first subgraph.
  • FIG. 4 shows a schematic diagram of a graph structure of a global graph according to an embodiment.
  • the relational network diagram surrounded by curve 41 corresponds to the first subgraph determined by the first terminal device
  • the relational network diagram surrounded by curve 42 corresponds to the second terminal
  • the relational network graph surrounded by the curve 43 corresponds to the third subgraph determined by the third terminal device, and each subgraph may have overlapping parts.
  • the actual global map contains a large number of sub-pictures determined by the terminal device, and FIG. 4 is only for illustration.
  • the graph embedding information further includes an edge feature vector corresponding to each connected edge in the first subgraph.
  • the server updating the global graph according to the first subgraph includes: the server updating the graph structure of the global graph according to the graph structure of the first subgraph; and/or, the server Update the node attribute of the target node of the global graph according to the node attribute of the target node of the first subgraph; and/or, the server updates the all-in-one attribute according to the edge attribute of the target connecting edge of the first subgraph The edge attribute of the target connecting edge of the global graph.
  • the server determining the graph embedding information of the first subgraph according to the updated global graph includes: the server based on the graph neural network (graph neural network) according to the updated global graph , GNN) algorithm or graph embedding Node2Vec algorithm to determine the graph embedding information of the first subgraph.
  • translating embedding or an algorithm based on topological features may also be used to determine the above-mentioned image embedding information.
  • the server sends the picture embedding information of the first sub-picture to the first terminal device. It is understandable that the server may actively send the picture embedding information of the first subgraph to the first terminal device; or after receiving the request of the first terminal device, send the picture embedding information of the first subgraph to the first terminal device.
  • the first terminal device determines the risk assessment result corresponding to the first sub-graph according to the image embedded information, and the risk assessment result is used to make risk decisions. It is understandable that the central node of the first subgraph corresponds to the first user, and therefore, the risk assessment result corresponding to the first subgraph can be used as the risk assessment result of the first user and used to make risk decisions on related behaviors of the first user.
  • the graph embedding information includes not only the node feature vector corresponding to each node in the first subgraph, but also the edge feature vector corresponding to each connected edge in the first subgraph;
  • the first terminal device determines the risk assessment result corresponding to the first subgraph according to the feature vector of each node and the feature vector of each edge.
  • the first terminal device uses a pre-trained classification model or regression model to determine the risk assessment result corresponding to the first subgraph at least according to the feature vector of each node.
  • the foregoing classification model or regression model may specifically be a GNN model, a logistic regression (LR) model, a gradient boosting tree XGBoost model, a deep neural network (deep neural networks, DNN) model, etc.
  • LR logistic regression
  • XGBoost gradient boosting tree XGBoost
  • DNN deep neural network
  • the terminal device can perform risk assessment based on the point-and-edge data related to the existing subgraph; the terminal device only needs to store and calculate the transaction (or social network) on the terminal device. Behaviour) The point and edge data related to the generated subgraph, rather than all the data of all terminal devices.
  • the data scale is controllable and suitable for calculation on the terminal device.
  • the server is only responsible for generating graph embedding information, and the risk calculation process and results are all on the terminal device, thus effectively avoiding the leakage of user risk data privacy;
  • the graph embedding information on the terminal device is generating and
  • the calculation is performed on the server, and the global map information is included in the calculation process, so the risk assessment accuracy is high.
  • the calculation of global graph information is decoupled from the calculation logic of local subgraphs, and the node characteristics of the subgraphs are used as the medium to effectively carry out the information transmission and calculation division of the whole graph and subgraphs.
  • the computing resources, storage resources and network resources are affected by In a limited terminal equipment environment, the integrity of the information and the validity of the calculation are guaranteed.
  • a risk detection device for protecting user privacy.
  • the device is set in a first terminal device and used to execute the first terminal in the risk detection method for protecting user privacy provided by the embodiments of this specification.
  • Fig. 5 shows a schematic block diagram of a risk detection device for protecting user privacy according to an embodiment.
  • the device 500 includes: a determining unit 51, a sending unit 52, a receiving unit 53, and an evaluating unit 54.
  • the determining unit 51 is configured to determine a first subgraph according to the association information of the first user corresponding to the first terminal device, the first subgraph including a central node and associated nodes of the central node, the central The node corresponds to the first user, and the associated node corresponds to one or more second users that have an associated relationship with the first user.
  • the sending unit 52 is configured to send the first subgraph determined by the determining unit 51 to the server, so that the server updates the global chart according to the first subgraph, and determines the first subgraph according to the updated global chart.
  • the graph embedding information of a subgraph includes at least the node feature vector corresponding to each node in the first subgraph; wherein the global graph is established based on the subgraphs sent by multiple terminal devices.
  • the receiving unit 53 is configured to receive the picture embedding information of the first sub-picture from the server.
  • the evaluation unit 54 is configured to determine the risk evaluation result corresponding to the first sub-picture according to the picture embedding information received by the receiving unit 53, and the risk evaluation result is used to make a risk decision.
  • the associated information includes: transaction information or social activity information.
  • the determining unit 51 is specifically configured to: collect the association relationship record of the first user and generate the first subgraph; or, according to the association relationship record newly generated by the first user in the most recent predetermined period of time To update the first subgraph that has been generated.
  • the graph embedding information further includes an edge feature vector corresponding to each connected edge in the first sub-graph; the evaluation unit 54 is specifically configured to perform according to the feature vector of each node and each The edge feature vector determines the risk assessment result corresponding to the first subgraph.
  • the evaluation unit 54 is specifically configured to use a pre-trained classification model or regression model to determine the risk evaluation result corresponding to the first subgraph at least according to the feature vector of each node.
  • a risk detection device for protecting user privacy.
  • the device is installed in a server and used to execute the processing flow of the server in the risk detection method for protecting user privacy provided by the embodiments of this specification.
  • Fig. 6 shows a schematic block diagram of a risk detection device for protecting user privacy according to an embodiment.
  • the device 600 includes: a receiving unit 61, an updating unit 62, and a sending unit 63.
  • the receiving unit 61 is configured to receive a first subgraph from a first terminal device corresponding to a first user, where the first subgraph includes a central node and an associated node of the central node, and the central node corresponds to the first Users, the associated node corresponds to one or more second users that have an associated relationship with the first user.
  • the updating unit 62 is configured to update the global graph according to the first subgraph received by the receiving unit 61, and determine the graph embedding information of the first subgraph according to the updated global graph, and the graph embedding information includes at least Node feature vectors corresponding to each node in the first subgraph; wherein, the global graph is established according to subgraphs sent by multiple terminal devices.
  • the sending unit 63 is configured to send the picture embedding information of the first subpicture determined by the updating unit 62 to the first terminal device, so that the first terminal device determines the picture embedding information according to the picture embedding information
  • the risk assessment result corresponding to the first subgraph, and the risk assessment result is used to make a risk decision.
  • the updating unit 62 is specifically configured to: update the graph structure of the global graph according to the graph structure of the first subgraph; and/or, according to the graph structure of the first subgraph
  • the node attribute of the target node updates the node attribute of the target node of the global graph; and/or, the edge of the target connected edge of the global graph is updated according to the edge attribute of the target connected edge of the first subgraph Attributes.
  • the updating unit 62 is specifically configured to determine the graph embedding information of the first subgraph based on the graph neural network GNN algorithm or the graph embedding Node2Vec algorithm according to the updated global graph.
  • the graph embedding information further includes an edge feature vector corresponding to each connected edge in the first subgraph.
  • a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed in a computer, the computer is caused to execute the method described in conjunction with FIG. 2.
  • a computing device including a memory and a processor, the memory stores executable code, and when the processor executes the executable code, it implements what is described in conjunction with FIG. 2 method.

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Security & Cryptography (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Signal Processing (AREA)
  • Health & Medical Sciences (AREA)
  • Bioethics (AREA)
  • General Health & Medical Sciences (AREA)
  • Theoretical Computer Science (AREA)
  • Databases & Information Systems (AREA)
  • Computer Hardware Design (AREA)
  • Software Systems (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Medical Informatics (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)

Abstract

本说明书实施例提供一种保护用户隐私的风险检测方法和装置,方法包括:第一终端设备根据与其对应的第一用户的关联信息,确定第一子图,第一子图包括中心节点和中心节点的关联节点,中心节点对应第一用户,关联节点对应与第一用户具有关联关系的第二用户;第一终端设备向服务器发送第一子图,以使服务器根据第一子图更新全局图,并根据更新后的全局图确定第一子图的图嵌入信息,图嵌入信息至少包括第一子图中各节点分别对应的节点特征向量;全局图根据多个终端设备发送的子图而建立;第一终端设备从服务器接收图嵌入信息;第一终端设备根据图嵌入信息,确定第一子图对应的风险评估结果,用于进行风险决策。能够有效的保护用户隐私。

Description

保护用户隐私的风险检测方法和装置 技术领域
本说明书一个或多个实施例涉及计算机领域,尤其涉及保护用户隐私的风险检测方法和装置。
背景技术
当前,支付、电商和社交等互联网应用场景存在着天然的网络属性,图计算常被用来在该类数据上进行风险检测,譬如反洗钱、赌博检测、传销检测、作弊检测、欺诈检测等。
采用全局图计算的方案,依靠服务器上的全局图对风险进行检测。包括标签传播、社区发现、图嵌入、图神经网络等,这些方法均需要存储并访问全图点边信息,包括全图点边所对应的属性信息与嵌入特征。受终端设备计算资源限制、存储资源与网络资源的限制,在终端设备上通常不能访问到全图点边信息,因此全局图计算不能在终端设备上运行,只能在服务器上计算,带来计算与通讯的延迟。同时,在服务器上计算风险评估结果和传输风险评估结果的过程中,存在用户风险数据隐私泄漏风险。
发明内容
本说明书一个或多个实施例描述了一种保护用户隐私的风险检测方法和装置,能够有效的保护用户隐私。
第一方面,提供了一种保护用户隐私的风险检测方法,方法包括:第一终端设备根据与所述第一终端设备对应的第一用户的关联信息,确定第一子图,所述第一子图包括中心节点和所述中心节点的关联节点,所述中心节点对应所述第一用户,所述关联节点对应与所述第一用户具有关联关系的一个或多个第二用户;所述第一终端设备向服务器发送所述第一子图,以使所述服务器根据所述第一子图更新全局图,并根据更新后的所述全局图确定所述第一子图的图嵌入信息,所述图嵌入信息至少包括所述第一子图中各节点分别对应的节点特征向量;其中,所述全局图根据多个终端设备发送的子图而建立;所述第一终端设备从所述服务器接收所述第一子图的图嵌入信息;所述第一终端设备根据所述图嵌入信息,确定所述第一子图对应的风险评估结果,所述风险评估结果用于进行风险决策。
在一种可能的实施方式中,所述关联信息包括:交易信息或者社交活动信息。
在一种可能的实施方式中,所述确定第一子图包括:采集第一用户的关联关系记录,生成第一子图;或者,根据第一用户在最近预定时长新产生的关联关系记录,更新已生成的第一子图。
在一种可能的实施方式中,所述图嵌入信息还包括,所述第一子图中各连接边分别对应的边特征向量;所述第一终端设备根据所述图嵌入信息,确定所述第一子图对应的风险评估结果,包括:所述第一终端设备根据各节点特征向量和各边特征向量,确定所述第一子图对应的风险评估结果。
在一种可能的实施方式中,所述第一终端设备根据所述图嵌入信息,确定所述第一子图对应的风险评估结果,包括:所述第一终端设备至少根据各节点特征向量,利用预先训练的分类模型或回归模型确定所述第一子图对应的风险评估结果。
第二方面,提供了一种保护用户隐私的风险检测方法,方法包括:服务器从与第一用户对应的第一终端设备接收第一子图,所述第一子图包括中心节点和所述中心节点的关联节点,所述中心节点对应所述第一用户,所述关联节点对应与所述第一用户具有关联关系的一个或多个第二用户;所述服务器根据所述第一子图更新全局图,并根据更新后的所述全局图确定所述第一子图的图嵌入信息,所述图嵌入信息至少包括所述第一子图中各节点分别对应的节点特征向量;其中,所述全局图根据多个终端设备发送的子图而建立;所述服务器向所述第一终端设备发送所述第一子图的图嵌入信息,以使所述第一终端设备根据所述图嵌入信息,确定所述第一子图对应的风险评估结果,所述风险评估结果用于进行风险决策。
在一种可能的实施方式中,所述服务器根据所述第一子图更新全局图,包括:所述服务器根据所述第一子图的图结构更新所述全局图的图结构;和/或,所述服务器根据所述第一子图的目标节点的节点属性更新所述全局图的所述目标节点的节点属性;和/或,所述服务器根据所述第一子图的目标连接边的边属性更新所述全局图的所述目标连接边的边属性。
在一种可能的实施方式中,所述服务器根据更新后的所述全局图确定所述第一子图的图嵌入信息,包括:所述服务器根据更新后的所述全局图,基于图神经网络(graph neural network,GNN)算法或者图嵌入Node2Vec算法确定所述第一子图的图嵌入信息。
在一种可能的实施方式中,所述图嵌入信息还包括,所述第一子图中各连接边分 别对应的边特征向量。
第三方面,提供了一种保护用户隐私的风险检测装置,所述装置设置于第一终端设备,装置包括:确定单元,用于根据与所述第一终端设备对应的第一用户的关联信息,确定第一子图,所述第一子图包括中心节点和所述中心节点的关联节点,所述中心节点对应所述第一用户,所述关联节点对应与所述第一用户具有关联关系的一个或多个第二用户;发送单元,用于向服务器发送所述确定单元确定的第一子图,以使所述服务器根据所述第一子图更新全局图,并根据更新后的所述全局图确定所述第一子图的图嵌入信息,所述图嵌入信息至少包括所述第一子图中各节点分别对应的节点特征向量;其中,所述全局图根据多个终端设备发送的子图而建立;接收单元,用于从所述服务器接收所述第一子图的图嵌入信息;评估单元,用于根据所述接收单元接收的图嵌入信息,确定所述第一子图对应的风险评估结果,所述风险评估结果用于进行风险决策。
第四方面,提供了一种保护用户隐私的风险检测装置,所述装置设置于服务器,装置包括:接收单元,用于从与第一用户对应的第一终端设备接收第一子图,所述第一子图包括中心节点和所述中心节点的关联节点,所述中心节点对应所述第一用户,所述关联节点对应与所述第一用户具有关联关系的一个或多个第二用户;更新单元,用于根据所述接收单元接收的第一子图更新全局图,并根据更新后的所述全局图确定所述第一子图的图嵌入信息,所述图嵌入信息至少包括所述第一子图中各节点分别对应的节点特征向量;其中,所述全局图根据多个终端设备发送的子图而建立;发送单元,用于向所述第一终端设备发送所述更新单元确定的所述第一子图的图嵌入信息,以使所述第一终端设备根据所述图嵌入信息,确定所述第一子图对应的风险评估结果,所述风险评估结果用于进行风险决策。
第五方面,提供了一种计算机可读存储介质,其上存储有计算机程序,当所述计算机程序在计算机中执行时,令计算机执行第一或第二方面的方法。
第六方面,提供了一种计算设备,包括存储器和处理器,所述存储器中存储有可执行代码,所述处理器执行所述可执行代码时,实现第一或第二方面的方法。
通过本说明书实施例提供的方法和装置,首先第一终端设备不是直接将用户的关联信息发送给服务器,而是根据与所述第一终端设备对应的第一用户的关联信息,确定第一子图,然后向服务器发送第一子图,服务器根据第一子图更新全局图,并根据更新后的全局图确定第一子图的图嵌入信息,接着服务器不是根据该图嵌入信息确定第一子图对应的风险评估结果,而是向第一终端设备发送该图嵌入信息,最后第一终端设备根 据该图嵌入信息,确定第一子图对应的风险评估结果,风险评估结果用于进行风险决策。由上可见,本说明书实施例,通过终端设备和服务器相结合的图计算方法对风险进行检测,在可用性、时效性和计算资源上做出了改进。同时由于服务器只负责生成图嵌入信息,而风险评估的过程和结果全部在终端设备上,从而有效的保护了用户隐私。
附图说明
为了更清楚地说明本申请实施例的技术方案,下面将对实施例描述中所需要使用的附图作简单地介绍,显而易见地,下面描述中的附图仅仅是本申请的一些实施例,对于本领域普通技术人员来讲,在不付出创造性劳动的前提下,还可以根据这些附图获得其它的附图。
图1为本说明书公开的一个实施例的实施场景示意图;
图2示出根据一个实施例的保护用户隐私的风险检测方法交互示意图;
图3示出根据一个实施例的第一子图的图结构示意图;
图4示出根据一个实施例的全局图的图结构示意图;
图5示出根据一个实施例的保护用户隐私的风险检测装置的示意性框图;
图6示出根据另一个实施例的保护用户隐私的风险检测装置的示意性框图。
具体实施方式
下面结合附图,对本说明书提供的方案进行描述。
图1为本说明书披露的一个实施例的实施场景示意图。该实施场景涉及保护用户隐私的风险检测。参照图1,服务器与多个终端设备具有通信连接,上述多个终端设备的数目可能比较大,图1中仅示出三个终端设备作为示例,各终端设备均可以根据各自对应的用户的关联信息建立子图,各终端设备将各自建立的子图发送给服务器,服务器根据接收的各子图建立全局图,可以理解的是,无论是子图还是全局图均为关系网络图。
终端设备对应的用户的关联信息是随时间变化的,例如,在第一时间周期内用户A和用户B不具有关联关系,在第一时间周期之后的第二时间周期内用户A和用户B具有关联关系,这种关联信息的变化会对应体现为子图的变化,相应地会引起全局图的变化。
参照图1,以第一终端设备要进行风险检测为例,第一终端设备将其确定的第一子图发送给服务器,以使服务器根据第一子图更新全局图,并根据更新后的所述全局图确定第一子图的图嵌入信息,服务器将该图嵌入信息发送给第一终端设备,第一终端设备根据该图嵌入信息,确定第一子图对应的风险评估结果,该风险评估结果用于进行风险决策。
上述风险评估结果是由终端设备确定的,服务器并不知晓该风险评估结果,风险评估结果不易泄露,从而有效的保护了用户隐私。
需要说明的是,本说明书实施例对于上述服务器的类型不做限定,上述服务器可以指物理服务器,也可以指云服务器。其中,云服务器可以提供一种简单高效、安全可靠、处理能力可弹性伸缩的计算服务。其管理方式比物理服务器更简单高效。用户无需提前购买硬件,即可迅速创建或释放任意多台云服务器。
图2示出根据一个实施例的保护用户隐私的风险检测方法交互示意图,该方法可以基于图1所示的实施场景,由第一终端设备和服务器相结合实现该方法。如图2所示,该实施例中保护用户隐私的风险检测方法包括步骤21~步骤25。
首先在步骤21,第一终端设备根据与第一终端设备对应的第一用户的关联信息,确定第一子图,所述第一子图包括中心节点和所述中心节点的关联节点,所述中心节点对应所述第一用户,所述关联节点对应与所述第一用户具有关联关系的一个或多个第二用户。可以理解的是,第一子图为关系网络图,由节点和节点之间的连接边构成,每个节点对应一个用户,中心节点可以具有一个或多个关联节点,关联节点与中心节点之间具有连接边。
图3示出根据一个实施例的第一子图的图结构示意图,参照图3,节点1为中心节点,节点2、节点3、节点4、节点5和节点6为节点1的关联节点。
在一个示例中,所述关联信息包括:交易信息或者社交活动信息。
可以理解的是,关联信息并不限定于此,其他的任何能够体现用户之间关联关系的信息均可作为关联信息,以用于构建关系网络图。
在一个示例中,所述确定第一子图包括:采集第一用户的关联关系记录,生成第一子图;或者,根据第一用户在最近预定时长新产生的关联关系记录,更新已生成的第一子图。
本说明书实施例中,第一终端设备可以周期性确定第一子图,或者在有风险检测 需求时确定第一子图。
然后在步骤22,第一终端设备向服务器发送第一子图。可以理解的是,第一终端设备可以向服务器发送第一子图的图结构、节点属性和边属性中的一项或多项信息。
本说明书实施例中,第一终端设备可以周期性向服务器发送第一子图,或者在有风险检测需求时向服务器发送第一子图。
接着在步骤23,服务器根据所述第一子图更新全局图,并根据更新后的全局图确定第一子图的图嵌入信息,图嵌入信息至少包括所述第一子图中各节点分别对应的节点特征向量;其中,所述全局图根据多个终端设备发送的子图而建立。可以理解的是,全局图是比子图包含节点更为广泛的关系网络图,全局图包含子图。
图4示出根据一个实施例的全局图的图结构示意图,参照图4,曲线41包围的关系网络图对应第一终端设备确定的第一子图,曲线42包围的关系网络图对应第二终端设备确定的第二子图,曲线43包围的关系网络图对应第三终端设备确定的第三子图,各子图之间可以具有重合的部分。可以理解的是,实际的全局图包含大量的终端设备确定的子图,图4仅为示意。
在一个示例中,所述图嵌入信息还包括,所述第一子图中各连接边分别对应的边特征向量。
在一个示例中,所述服务器根据所述第一子图更新全局图,包括:所述服务器根据所述第一子图的图结构更新所述全局图的图结构;和/或,所述服务器根据所述第一子图的目标节点的节点属性更新所述全局图的所述目标节点的节点属性;和/或,所述服务器根据所述第一子图的目标连接边的边属性更新所述全局图的所述目标连接边的边属性。
在一个示例中,所述服务器根据更新后的所述全局图确定所述第一子图的图嵌入信息,包括:所述服务器根据更新后的所述全局图,基于图神经网络(graph neural network,GNN)算法或者图嵌入Node2Vec算法确定所述第一子图的图嵌入信息。
此外,还可以采用翻译嵌入(translating embedding,TransE)或者基于拓扑特征的算法确定上述图嵌入信息。
再在步骤24,服务器向第一终端设备发送第一子图的图嵌入信息。可以理解的是,服务器可以主动向第一终端设备发送第一子图的图嵌入信息;或者在接收到第一终端设备的请求后,向第一终端设备发送第一子图的图嵌入信息。
最后在步骤25,第一终端设备根据图嵌入信息,确定第一子图对应的风险评估结果,风险评估结果用于进行风险决策。可以理解的是,第一子图的中心节点对应第一用户,因此第一子图对应的风险评估结果可以作为第一用户的风险评估结果,用于对第一用户的相关行为进行风险决策。
在一个示例中,所述图嵌入信息不仅包括所述第一子图中各节点分别对应的节点特征向量,还包括,所述第一子图中各连接边分别对应的边特征向量;所述第一终端设备根据各节点特征向量和各边特征向量,确定所述第一子图对应的风险评估结果。
在一个示例中,所述第一终端设备至少根据各节点特征向量,利用预先训练的分类模型或回归模型确定所述第一子图对应的风险评估结果。
上述分类模型或回归模型具体可以为GNN模型、逻辑回归(logistic regression,LR)模型、梯度提升树XGBoost模型、深度神经网络(deep neural networks,DNN)模型等。
在一个示例中,在没有网络连接的环境下,终端设备可以根据其上已有的子图相关的点边数据进行风险评估;在终端设备上只需要存储和计算该终端设备上交易(或者社交行为)所产生的子图相关的点边数据,而非所有终端设备的全部数据,数据规模可控,适合在终端设备上计算。
本说明书实施例提供的方法,服务器只负责生成图嵌入信息,而风险计算的过程和结果全部在终端设备上,从而有效的避免了用户风险数据隐私泄漏;终端设备上的图嵌入信息在生成和更新时,是在服务器进行计算,计算过程中包括了全局图信息,因此风险评估准确性高。将全局图信息的计算与局部子图计算逻辑进行解耦,并以子图的节点特征作为介质,有效的进行全图和子图的信息传递与计算分割,在计算资源、存储资源和网络资源受限的终端设备环境中,保证了信息完整性和计算有效性。
根据另一方面的实施例,还提供一种保护用户隐私的风险检测装置,所述装置设置于第一终端设备,用于执行本说明书实施例提供的保护用户隐私的风险检测方法中第一终端设备的处理流程。图5示出根据一个实施例的保护用户隐私的风险检测装置的示意性框图。如图5所示,该装置500包括:确定单元51、发送单元52、接收单元53、评估单元54。
确定单元51,用于根据与所述第一终端设备对应的第一用户的关联信息,确定第一子图,所述第一子图包括中心节点和所述中心节点的关联节点,所述中心节点对应所 述第一用户,所述关联节点对应与所述第一用户具有关联关系的一个或多个第二用户。
发送单元52,用于向服务器发送所述确定单元51确定的第一子图,以使所述服务器根据所述第一子图更新全局图,并根据更新后的所述全局图确定所述第一子图的图嵌入信息,所述图嵌入信息至少包括所述第一子图中各节点分别对应的节点特征向量;其中,所述全局图根据多个终端设备发送的子图而建立。
接收单元53,用于从所述服务器接收所述第一子图的图嵌入信息。
评估单元54,用于根据所述接收单元53接收的图嵌入信息,确定所述第一子图对应的风险评估结果,所述风险评估结果用于进行风险决策。
可选地,作为一个实施例,所述关联信息包括:交易信息或者社交活动信息。
可选地,作为一个实施例,所述确定单元51,具体用于:采集第一用户的关联关系记录,生成第一子图;或者,根据第一用户在最近预定时长新产生的关联关系记录,更新已生成的第一子图。
可选地,作为一个实施例,所述图嵌入信息还包括,所述第一子图中各连接边分别对应的边特征向量;所述评估单元54,具体用于根据各节点特征向量和各边特征向量,确定所述第一子图对应的风险评估结果。
可选地,作为一个实施例,所述评估单元54,具体用于至少根据各节点特征向量,利用预先训练的分类模型或回归模型确定所述第一子图对应的风险评估结果。
根据另一方面的实施例,还提供一种保护用户隐私的风险检测装置,所述装置设置于服务器,用于执行本说明书实施例提供的保护用户隐私的风险检测方法中服务器的处理流程。图6示出根据一个实施例的保护用户隐私的风险检测装置的示意性框图。如图6所示,该装置600包括:接收单元61、更新单元62、发送单元63。
接收单元61,用于从与第一用户对应的第一终端设备接收第一子图,所述第一子图包括中心节点和所述中心节点的关联节点,所述中心节点对应所述第一用户,所述关联节点对应与所述第一用户具有关联关系的一个或多个第二用户。
更新单元62,用于根据所述接收单元61接收的第一子图更新全局图,并根据更新后的所述全局图确定所述第一子图的图嵌入信息,所述图嵌入信息至少包括所述第一子图中各节点分别对应的节点特征向量;其中,所述全局图根据多个终端设备发送的子图而建立。
发送单元63,用于向所述第一终端设备发送所述更新单元62确定的所述第一子图的图嵌入信息,以使所述第一终端设备根据所述图嵌入信息,确定所述第一子图对应的风险评估结果,所述风险评估结果用于进行风险决策。
可选地,作为一个实施例,所述更新单元62,具体用于:根据所述第一子图的图结构更新所述全局图的图结构;和/或,根据所述第一子图的目标节点的节点属性更新所述全局图的所述目标节点的节点属性;和/或,根据所述第一子图的目标连接边的边属性更新所述全局图的所述目标连接边的边属性。
可选地,作为一个实施例,所述更新单元62,具体用于根据更新后的所述全局图,基于图神经网络GNN算法或者图嵌入Node2Vec算法确定所述第一子图的图嵌入信息。
可选地,作为一个实施例,所述图嵌入信息还包括,所述第一子图中各连接边分别对应的边特征向量。
根据另一方面的实施例,还提供一种计算机可读存储介质,其上存储有计算机程序,当所述计算机程序在计算机中执行时,令计算机执行结合图2所描述的方法。
根据再一方面的实施例,还提供一种计算设备,包括存储器和处理器,所述存储器中存储有可执行代码,所述处理器执行所述可执行代码时,实现结合图2所描述的方法。
本领域技术人员应该可以意识到,在上述一个或多个示例中,本申请所描述的功能可以用硬件、软件、固件或它们的任意组合来实现。当使用软件实现时,可以将这些功能存储在计算机可读介质中或者作为计算机可读介质上的一个或多个指令或代码进行传输。
以上所述的具体实施方式,对本申请的目的、技术方案和有益效果进行了进一步详细说明,所应理解的是,以上所述仅为本申请的具体实施方式而已,并不用于限定本申请的保护范围,凡在本申请的技术方案的基础之上,所做的任何修改、等同替换、改进等,均应包括在本申请的保护范围之内。

Claims (20)

  1. 一种保护用户隐私的风险检测方法,所述方法包括:
    第一终端设备根据与所述第一终端设备对应的第一用户的关联信息,确定第一子图,所述第一子图包括中心节点和所述中心节点的关联节点,所述中心节点对应所述第一用户,所述关联节点对应与所述第一用户具有关联关系的一个或多个第二用户;
    所述第一终端设备向服务器发送所述第一子图,以使所述服务器根据所述第一子图更新全局图,并根据更新后的所述全局图确定所述第一子图的图嵌入信息,所述图嵌入信息至少包括所述第一子图中各节点分别对应的节点特征向量;其中,所述全局图根据多个终端设备发送的子图而建立;
    所述第一终端设备从所述服务器接收所述第一子图的图嵌入信息;
    所述第一终端设备根据所述图嵌入信息,确定所述第一子图对应的风险评估结果,所述风险评估结果用于进行风险决策。
  2. 如权利要求1所述的方法,其中,所述关联信息包括:
    交易信息或者社交活动信息。
  3. 如权利要求1所述的方法,其中,所述确定第一子图包括:
    采集第一用户的关联关系记录,生成第一子图;或者,根据第一用户在最近预定时长新产生的关联关系记录,更新已生成的第一子图。
  4. 如权利要求1所述的方法,其中,所述图嵌入信息还包括,所述第一子图中各连接边分别对应的边特征向量;
    所述第一终端设备根据所述图嵌入信息,确定所述第一子图对应的风险评估结果,包括:
    所述第一终端设备根据各节点特征向量和各边特征向量,确定所述第一子图对应的风险评估结果。
  5. 如权利要求1所述的方法,其中,所述第一终端设备根据所述图嵌入信息,确定所述第一子图对应的风险评估结果,包括:
    所述第一终端设备至少根据各节点特征向量,利用预先训练的分类模型或回归模型确定所述第一子图对应的风险评估结果。
  6. 一种保护用户隐私的风险检测方法,所述方法包括:
    服务器从与第一用户对应的第一终端设备接收第一子图,所述第一子图包括中心节点和所述中心节点的关联节点,所述中心节点对应所述第一用户,所述关联节点对应与所述第一用户具有关联关系的一个或多个第二用户;
    所述服务器根据所述第一子图更新全局图,并根据更新后的所述全局图确定所述第一子图的图嵌入信息,所述图嵌入信息至少包括所述第一子图中各节点分别对应的节点特征向量;其中,所述全局图根据多个终端设备发送的子图而建立;
    所述服务器向所述第一终端设备发送所述第一子图的图嵌入信息,以使所述第一终端设备根据所述图嵌入信息,确定所述第一子图对应的风险评估结果,所述风险评估结果用于进行风险决策。
  7. 如权利要求6所述的方法,其中,所述服务器根据所述第一子图更新全局图,包括:
    所述服务器根据所述第一子图的图结构更新所述全局图的图结构;和/或,
    所述服务器根据所述第一子图的目标节点的节点属性更新所述全局图的所述目标节点的节点属性;和/或,
    所述服务器根据所述第一子图的目标连接边的边属性更新所述全局图的所述目标连接边的边属性。
  8. 如权利要求6所述的方法,其中,所述服务器根据更新后的所述全局图确定所述第一子图的图嵌入信息,包括:
    所述服务器根据更新后的所述全局图,基于图神经网络GNN算法或者图嵌入Node2Vec算法确定所述第一子图的图嵌入信息。
  9. 如权利要求6所述的方法,其中,所述图嵌入信息还包括,所述第一子图中各连接边分别对应的边特征向量。
  10. 一种保护用户隐私的风险检测装置,所述装置设置于第一终端设备,所述装置包括:
    确定单元,用于根据与所述第一终端设备对应的第一用户的关联信息,确定第一子图,所述第一子图包括中心节点和所述中心节点的关联节点,所述中心节点对应所述第一用户,所述关联节点对应与所述第一用户具有关联关系的一个或多个第二用户;
    发送单元,用于向服务器发送所述确定单元确定的第一子图,以使所述服务器根据所述第一子图更新全局图,并根据更新后的所述全局图确定所述第一子图的图嵌入信息,所述图嵌入信息至少包括所述第一子图中各节点分别对应的节点特征向量;其中,所述全局图根据多个终端设备发送的子图而建立;
    接收单元,用于从所述服务器接收所述第一子图的图嵌入信息;
    评估单元,用于根据所述接收单元接收的图嵌入信息,确定所述第一子图对应的风险评估结果,所述风险评估结果用于进行风险决策。
  11. 如权利要求10所述的装置,其中,所述关联信息包括:
    交易信息或者社交活动信息。
  12. 如权利要求10所述的装置,其中,所述确定单元,具体用于:
    采集第一用户的关联关系记录,生成第一子图;或者,根据第一用户在最近预定时长新产生的关联关系记录,更新已生成的第一子图。
  13. 如权利要求10所述的装置,其中,所述图嵌入信息还包括,所述第一子图中各连接边分别对应的边特征向量;
    所述评估单元,具体用于根据各节点特征向量和各边特征向量,确定所述第一子图对应的风险评估结果。
  14. 如权利要求10所述的装置,其中,所述评估单元,具体用于至少根据各节点特征向量,利用预先训练的分类模型或回归模型确定所述第一子图对应的风险评估结果。
  15. 一种保护用户隐私的风险检测装置,所述装置设置于服务器,所述装置包括:
    接收单元,用于从与第一用户对应的第一终端设备接收第一子图,所述第一子图包括中心节点和所述中心节点的关联节点,所述中心节点对应所述第一用户,所述关联节点对应与所述第一用户具有关联关系的一个或多个第二用户;
    更新单元,用于根据所述接收单元接收的第一子图更新全局图,并根据更新后的所述全局图确定所述第一子图的图嵌入信息,所述图嵌入信息至少包括所述第一子图中各节点分别对应的节点特征向量;其中,所述全局图根据多个终端设备发送的子图而建立;
    发送单元,用于向所述第一终端设备发送所述更新单元确定的所述第一子图的图嵌入信息,以使所述第一终端设备根据所述图嵌入信息,确定所述第一子图对应的风险评估结果,所述风险评估结果用于进行风险决策。
  16. 如权利要求15所述的装置,其中,所述更新单元,具体用于:
    根据所述第一子图的图结构更新所述全局图的图结构;和/或,
    根据所述第一子图的目标节点的节点属性更新所述全局图的所述目标节点的节点属性;和/或,
    根据所述第一子图的目标连接边的边属性更新所述全局图的所述目标连接边的边属性。
  17. 如权利要求15所述的装置,其中,所述更新单元,具体用于根据更新后的所述全局图,基于图神经网络GNN算法或者图嵌入Node2Vec算法确定所述第一子图的图嵌入信息。
  18. 如权利要求15所述的装置,其中,所述图嵌入信息还包括,所述第一子图中 各连接边分别对应的边特征向量。
  19. 一种计算机可读存储介质,其上存储有计算机程序,当所述计算机程序在计算机中执行时,令计算机执行权利要求1-9中任一项的所述的方法。
  20. 一种计算设备,包括存储器和处理器,所述存储器中存储有可执行代码,所述处理器执行所述可执行代码时,实现权利要求1-9中任一项的所述的方法。
PCT/CN2020/132868 2020-03-24 2020-11-30 保护用户隐私的风险检测方法和装置 WO2021189925A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202010213835.0A CN111405563B (zh) 2020-03-24 2020-03-24 保护用户隐私的风险检测方法和装置
CN202010213835.0 2020-03-24

Publications (1)

Publication Number Publication Date
WO2021189925A1 true WO2021189925A1 (zh) 2021-09-30

Family

ID=71436611

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/132868 WO2021189925A1 (zh) 2020-03-24 2020-11-30 保护用户隐私的风险检测方法和装置

Country Status (2)

Country Link
CN (1) CN111405563B (zh)
WO (1) WO2021189925A1 (zh)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111405563B (zh) * 2020-03-24 2021-07-13 支付宝(杭州)信息技术有限公司 保护用户隐私的风险检测方法和装置
CN114662204B (zh) * 2022-04-07 2023-03-31 清华大学 基于图神经网络的弹性杆系结构体系数据处理方法及装置

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190215697A1 (en) * 2017-12-18 2019-07-11 Korea University Research And Business Foundation Apparatus and method for managing risk of malware behavior in mobile operating system and recording medium for perform the method
CN110245269A (zh) * 2019-05-06 2019-09-17 阿里巴巴集团控股有限公司 获取关系网络图中节点的动态嵌入向量的方法和装置
CN110309367A (zh) * 2018-03-05 2019-10-08 腾讯科技(深圳)有限公司 信息分类的方法、信息处理的方法和装置
CN110796269A (zh) * 2019-09-30 2020-02-14 北京明略软件系统有限公司 一种生成模型的方法、装置、信息处理的方法及装置
CN111405563A (zh) * 2020-03-24 2020-07-10 支付宝(杭州)信息技术有限公司 保护用户隐私的风险检测方法和装置

Family Cites Families (9)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101860883B (zh) * 2010-05-14 2012-10-24 南京邮电大学 一种基于物联网的多代理异常检测方法
US9996811B2 (en) * 2013-12-10 2018-06-12 Zendrive, Inc. System and method for assessing risk through a social network
US10607226B2 (en) * 2015-04-14 2020-03-31 Samsung Electronics Co., Ltd. System and method for fraud detection in a mobile device
CN107645483B (zh) * 2016-07-22 2021-03-19 创新先进技术有限公司 风险识别方法、风险识别装置、云风险识别装置及系统
CN109934706B (zh) * 2017-12-15 2021-10-29 创新先进技术有限公司 一种基于图结构模型的交易风险控制方法、装置以及设备
CN109657918B (zh) * 2018-11-19 2023-07-18 平安科技(深圳)有限公司 关联评估对象的风险预警方法、装置和计算机设备
CN109949176B (zh) * 2019-03-28 2022-07-15 南京邮电大学 一种基于图嵌入的社交网络中异常用户检测方法
CN110276193B (zh) * 2019-05-17 2023-08-22 创新先进技术有限公司 风险特征输出方法、应用运行控制方法、系统及装置
CN110210227B (zh) * 2019-06-11 2021-05-14 百度在线网络技术(北京)有限公司 风险检测方法、装置、设备和存储介质

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20190215697A1 (en) * 2017-12-18 2019-07-11 Korea University Research And Business Foundation Apparatus and method for managing risk of malware behavior in mobile operating system and recording medium for perform the method
CN110309367A (zh) * 2018-03-05 2019-10-08 腾讯科技(深圳)有限公司 信息分类的方法、信息处理的方法和装置
CN110245269A (zh) * 2019-05-06 2019-09-17 阿里巴巴集团控股有限公司 获取关系网络图中节点的动态嵌入向量的方法和装置
CN110796269A (zh) * 2019-09-30 2020-02-14 北京明略软件系统有限公司 一种生成模型的方法、装置、信息处理的方法及装置
CN111405563A (zh) * 2020-03-24 2020-07-10 支付宝(杭州)信息技术有限公司 保护用户隐私的风险检测方法和装置

Also Published As

Publication number Publication date
CN111405563B (zh) 2021-07-13
CN111405563A (zh) 2020-07-10

Similar Documents

Publication Publication Date Title
JP6686209B2 (ja) ネットワーク内の分散データベースのための方法及び装置
TWI703460B (zh) 分散式圖嵌入方法、裝置、設備及系統
CN103620585B (zh) 虚拟身份管理器
US9104803B2 (en) On-demand software test environment generation
US20130290226A1 (en) System and method for social graph and graph assets valuation and monetization
CN110163004A (zh) 一种区块链生成的方法、相关设备及系统
EP4127998A1 (en) Distributed and blockchain-based ledgers for data cloud services
WO2021189925A1 (zh) 保护用户隐私的风险检测方法和装置
Nzoukou et al. A unified framework for measuring a network's mean time-to-compromise
CN112799708B (zh) 联合更新业务模型的方法及系统
US20150220927A1 (en) Method, apparatus and system for providing transaction indemnification
US10542105B2 (en) Geo-locating individuals based on a derived social network
Al-Qamash et al. Cloud, fog, and edge computing: A software engineering perspective
CN110224859B (zh) 用于识别团伙的方法和系统
WO2021195618A1 (en) System and method for integration and validation
JP6793848B2 (ja) オンライントランザクションの成功率の向上
WO2022267769A1 (zh) 图数据生成的方法及装置
CN116167868A (zh) 基于隐私计算的风险识别方法、装置、设备以及存储介质
CN112181599B (zh) 模型训练方法、装置及存储介质
CN104394162B (zh) 一种考虑服务质量可信性的网络服务选取方法
CN115955362B (zh) 基于区块链的数据存储及通信方法、装置、设备和介质
US11947410B2 (en) Broadband error remediation engine in a data analytics system
Le et al. Computational intelligence towards trusted cloudlet based fog computing
US20240046147A1 (en) Systems and methods for administrating a federated learning network
US20230012202A1 (en) Graph computing over micro-level and macro-level views

Legal Events

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

Ref document number: 20926588

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20926588

Country of ref document: EP

Kind code of ref document: A1