Data publishing method based on Bayesian network and related equipment
Through the Bayesian network-based data publishing method, a network that complies with privacy budgets is built, differential privacy budgets are allocated and noisy data is added, which solves the problem of privacy leakage in data sharing and achieves efficient privacy protection and data publishing.
Patent Information
- Application Number
- CN202510754441.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-06
- Publication Date
- 2025-08-08
AI Technical Summary
The existing technology is difficult to effectively protect privacy during data sharing and publishing, resulting in the leakage of private information, and at the same time affecting the practicality and publishing efficiency of data.
Using a Bayesian network-based data publishing method, by building a Bayesian network that meets the target privacy budget, the importance of each network node is determined, and the differential privacy budget is allocated according to the importance, and noise data is added to generate the data set to be published.
On the premise of ensuring privacy protection, the practicality and efficiency of data release are improved, the risk of privacy leakage is reduced, and the security and controllability of data release are enhanced.
Smart Images
Figure CN120449208A_ABST
Abstract
Description
Technical Field
[0001] The embodiments of the present invention relate to the technical fields of data privacy processing and data publishing, and more particularly to a Bayesian network-based data publishing method and related equipment. Background Art
[0002] With the rapid development of big data technology, data sharing has become crucial. However, data sharing and publishing can easily lead to privacy leaks. Directly publishing raw data can lead to malicious theft of private information. Current solutions include using differential privacy, data anonymization, and encryption to protect data. However, these protection methods significantly impact the practicality and efficiency of data publishing. Summary of the Invention
[0003] In view of this, the present invention provides a data publishing method and related equipment based on a Bayesian network, which improves the practicality and efficiency of data publishing while meeting the privacy protection of data.
[0004] In a first aspect, an embodiment of the present invention provides a data publishing method based on a Bayesian network. The method comprises:
[0005] Obtaining an original data set and a corresponding target privacy budget; the original data set includes data of multiple data attributes;
[0006] Constructing a Bayesian network that meets the target privacy budget according to the data attributes; the Bayesian network includes a plurality of network nodes;
[0007] Determining the importance of each of the network nodes based on the distribution of a plurality of data in the original data set at each of the network nodes;
[0008] Allocating the target privacy budget to each network node according to the importance of each network node to obtain a differential privacy budget for each network node;
[0009] Adding noise data to the original dataset according to the differential privacy budget to obtain a dataset to be released;
[0010] The to-be-published dataset is published.
[0011] In one possible implementation, constructing a Bayesian network that meets the target privacy budget based on the data attributes includes:
[0012] Calculating the information entropy of each data attribute in the original data set;
[0013] Determine the target data attribute with the maximum information entropy;
[0014] Determine the target data attribute as the first node;
[0015] The Bayesian network is constructed based on the first node.
[0016] In a possible implementation, constructing the Bayesian network based on the first node includes:
[0017] Determining each of the data attributes as a network node;
[0018] Determining a set of parent nodes corresponding to each of the network nodes according to the target privacy budget;
[0019] Obtaining connection relationships and distribution probabilities between the network nodes according to the parent node set corresponding to the network nodes and the head node;
[0020] The Bayesian network is obtained according to the connection relationship between the various network nodes, the distribution probability and the first node.
[0021] In a possible implementation, determining the importance of each of the network nodes based on the distribution of the plurality of data in the original data set at each of the network nodes includes:
[0022] Clustering the network nodes according to the data attributes to obtain a plurality of disjoint clusters;
[0023] Calculating the information entropy of each cluster;
[0024] Determining the importance of each cluster according to the information entropy of each cluster;
[0025] The importance of each network node is determined according to the importance of each cluster and the network nodes included in each cluster.
[0026] In one possible implementation, allocating the target privacy budget to each network node according to the importance of each network node to obtain a differential privacy budget for each network node includes:
[0027] Determine a privacy budget coefficient for each cluster according to the importance of each cluster;
[0028] Allocating the target privacy budget to each cluster according to the privacy budget coefficient of each cluster;
[0029] The privacy budget allocated to each cluster is allocated to the network nodes included in each cluster to obtain the differential privacy budget of each network node.
[0030] In one possible implementation, the network nodes in the Bayesian network include branch nodes and leaf nodes; and allocating the privacy budgets allocated to the clusters to the network nodes included in the clusters to obtain differential privacy budgets for the network nodes includes:
[0031] Determining a first allocation ratio for the branch nodes; and determining a second allocation ratio for the leaf nodes; wherein the second allocation ratio is greater than the first allocation ratio;
[0032] The differential privacy budget of each network node is obtained according to the target privacy budget allocated to each cluster, the network nodes included in each cluster, the first allocation ratio, and the second allocation ratio.
[0033] In one possible implementation, adding noise data to the original dataset according to the differential privacy budget includes:
[0034] Determining a joint probability distribution of each network node in the Bayesian network;
[0035] Adding Laplace noise to the joint probability distribution of each network node according to the differential privacy budget of each network node to obtain a noisy joint distribution of each network node;
[0036] Obtaining a noise distribution probability according to the noisy joint distribution of each network node;
[0037] Each network node in the Bayesian network is sampled according to the noise distribution probability to obtain the data set to be released.
[0038] In a second aspect, an embodiment of the present invention provides a data publishing device based on a Bayesian network, comprising:
[0039] An acquisition module, configured to acquire an original data set and a corresponding target privacy budget; the original data set includes data of multiple data attributes;
[0040] A network construction module, configured to construct a Bayesian network that meets the target privacy budget based on the data attributes; the Bayesian network includes a plurality of network nodes;
[0041] a processing module, configured to determine the importance of each of the network nodes based on the distribution of a plurality of data in the original data set at each of the network nodes;
[0042] The processing module is further configured to allocate the target privacy budget to each network node according to the importance of each network node, so as to obtain a differential privacy budget of each network node;
[0043] A noise adding module, configured to add noise data to the original dataset according to the differential privacy budget to obtain data to be published;
[0044] A publishing module is used to publish the data set to be published.
[0045] In a third aspect, an embodiment of the present invention provides an electronic device, including:
[0046] at least one processor; and
[0047] at least one memory in communication with the processor, wherein:
[0048] The memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the method described in the first aspect.
[0049] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to execute the method described in the first aspect.
[0050] In this embodiment of the present invention, a Bayesian network is modified to meet the needs of private data release. By assigning a differential privacy budget to each node in the Bayesian network based on its importance, noise data is added based on the differential privacy budget to generate a dataset to be released. This dataset, generated through these steps, ensures privacy protection while improving data practicality and release efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0051] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0052] Figure 1 A flowchart of a data publishing method based on a Bayesian network provided by an embodiment of the present invention;
[0053] Figure 2 A schematic diagram of a Bayesian network provided by an embodiment of the present invention;
[0054] Figure 3 A schematic diagram of the structure of a data publishing device based on a Bayesian network provided by an embodiment of the present invention;
[0055] Figure 4 A schematic structural diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0056] In order to better understand the technical solution of the present application, the embodiments of the present application are described in detail below with reference to the accompanying drawings.
[0057] It should be clear that the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0058] The terms used in the embodiments of the present application are for the purpose of describing specific embodiments only and are not intended to limit the present application. The singular forms "a", "an", "the" and "the" used in the embodiments of the present application and the appended claims are also intended to include plural forms unless the context clearly indicates otherwise.
[0059] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. Furthermore, the character " / " in this document generally indicates an "or" relationship between the associated objects.
[0060] In order to improve the practicality and publishing efficiency of data while ensuring the privacy protection effect of data, the embodiment of the present invention provides a data publishing method based on Bayesian network. Figure 1 A flow chart of a Bayesian network data publishing method provided by an embodiment of the present invention. Figure 1 As shown in , the method includes:
[0061] Step 101: Obtain an original data set and a corresponding target privacy budget. The original data set includes data of multiple data attributes.
[0062] The target privacy budget can be determined based on the overall sensitivity of the original dataset. The higher the sensitivity of the original dataset, the lower the target privacy budget. For example, a raw dataset with a moderate sensitivity level can have a target privacy budget of 1. A raw dataset with a high sensitivity level can have a target privacy budget of 0.1. A raw dataset with a high sensitivity level can have a target privacy budget of 0.01.
[0063] The original dataset can be obtained by sampling information from different dimensions of the sample. For example, if the sample consists of 100 individuals, the original dataset can be obtained by sampling information from multiple dimensions such as age, health status, average monthly income, and average monthly expenditure. The sampling dimensions such as age, health status, and average monthly income are data attributes.
[0064] Step 102: Construct a Bayesian network that meets the target privacy budget based on the data attributes. The Bayesian network includes multiple network nodes.
[0065] A Bayesian network is a probabilistic graphical network model that represents conditional dependencies between variables using a directed acyclic graph. Each network node represents a random variable, and edges represent direct dependencies. Each network node in a Bayesian network has a conditional probability table describing the probability distribution of the variable (i.e., network node) under the state of its parent node. If network node A points to network node B through a directed edge, then network node A is called the parent node of network node B. Figure 2 A schematic diagram of a Bayesian network provided by an embodiment of the present invention. Figure 2 As shown in , the original data set includes 5 data attributes, and the resulting Bayesian network contains 5 network nodes, namely age, exposure to toxins, gender, smoking, and cancer.
[0066] Optionally, an exponential mechanism can be used to construct a k-degree Bayesian network that meets the target privacy budget. Specifically, when using the exponential mechanism to construct a k-degree Bayesian network, the degree of dependence between various data attributes can be determined based on information entropy. The larger the value of the information entropy of a data attribute, the greater the degree of influence of the data attribute on other data attributes. The data attribute with the largest information entropy is determined as the first node of the Bayesian network. The first node is a network node without any parent node. Specifically, the information entropy of each data attribute in the original data set is first calculated. The information entropy can be calculated using the following formula (1).
[0067] Formula (1):
[0068] Among them, H(X) is the information entropy of data attribute X, p(x i ) is the probability that the data attribute X takes the value xi. n is the number of data attribute types in the original data set.
[0069] Afterwards, the data attribute with the largest information entropy is determined as the first node. Finally, a Bayesian network is constructed based on the first node. In the process of constructing the Bayesian network based on the first node, each data attribute is determined as a network node. Afterwards, the parent node set corresponding to each network node is determined based on the target privacy budget. And based on the parent node set corresponding to each network node and the above-mentioned first node, the connection relationship and distribution probability between each network node are obtained. Among them, the purpose of this step is to select AP pairs and corresponding probability distributions based on the target privacy budget. Among them, AP pairs refer to data attribute-parent node sets. Finally, the Bayesian network is obtained based on the connection relationship, distribution probability and first node between each network node.
[0070] In a specific example, the process of constructing a k-degree Bayesian network based on the original data set is as follows: First, initialize the Bayesian network N to be empty, set V to represent the set of data attributes whose parent node set is fixed, and the initial set of set V is empty. Then select the first node, calculate the information entropy of each data attribute according to formula (1), add the data attribute with the largest information entropy as the initial node A1 to the set V, and add the AP pair of this data attribute: (A1, ∅) to the Bayesian network N. Then, select other AP pairs. For data attribute A i , select the data attribute A in a differentially private manner through the exponential mechanism i The specific method is:
[0071] (1) Initialize the AP candidate set Ψ to ∅.
[0072] (2) For each data attribute Ai∈A\V, and Π i ∈ . Set AP to (A i ,Π i ) is added to the AP pair candidate set Ψ. Among them, Defines the set of all subsets of V whose size is min(k,|V|).
[0073] (3) Proportional to The sampling probability of selecting AP pairs from the AP pair candidate set Ψ is i ,Π i ). Where I is the mutual information, which is used as the score function. ε1 is the privacy budget. △ is the global sensitivity. When A i or Π i When it is binary data, it can be calculated by formula (2). i and Π i When all are non-binary data, it is calculated using formula (3).
[0074] Formula (2):
[0075] Formula (3):
[0076] (4) Network node A i Add to the set V, (A i ,Π i ) is added to the Bayesian network N.
[0077] (5) Repeat steps (2) to (4) until all data attributes are selected.
[0078] (6) Obtain the complete Bayesian network N.
[0079] Step 103: Determine the importance of each network node based on the distribution of a number of data in the original data set at each network node.
[0080] When determining the importance of each network node, clustering can be used to reduce data dimensionality. Specifically, each network node is clustered based on data attributes to obtain multiple non-overlapping clusters. Clustering methods can use partitioning, hierarchical methods, density-based methods, model-based methods, and other methods.
[0081] Next, the information entropy of each cluster is calculated. Next, the importance of each cluster is determined based on the information entropy of each cluster. Finally, the importance of each network node is determined based on the importance of each cluster and the network nodes contained in each cluster.
[0082] In a specific example, the process of clustering using the Markov blanket algorithm specifically includes:
[0083] The Markov blanket MB(x) of an attribute variable X (i.e., a network node X in a Bayesian network N) can be expressed as the set of parent nodes Pa(x) and child nodes Ch(x) of X, as well as the set of parent nodes of the child nodes of X. That is, MB(x)=Pa(x)∪Ch(x)∪{Pa(y)|y∈Ch(x)}.
[0084] When performing clustering, we can first initialize the set S so that it contains all the network nodes in the Bayesian network N. Then, we randomly select a network node X, add its corresponding MB(x) and the network node X itself to the cluster, and delete the network node from the set S. We repeat this process until the set S is empty. At this point, it means that all network nodes have been added, and t mutually exclusive clusters are obtained. These clusters are denoted as CL1, CL2, ...CL t Each cluster consists of a cluster head (network node X) and its Markov blanket members.
[0085] Afterwards, the allocation of the differential privacy budget for subsequent steps is determined based on the conditional independence between cluster attributes and the cluster importance based on the probability distribution of the original dataset. Specifically, the importance of each cluster can be calculated using Formula (4).
[0086] Formula (4):
[0087] Among them, F (CL i ) is cluster CL i The importance factor of cluster CL i F (CL i The higher the value of ), the more important the cluster is. The numerator of formula (4) represents the cluster CL i The denominator represents the sum of the information entropy of all network nodes in the Bayesian network N.
[0088] Step 104 : Allocate the target privacy budget to each network node according to the importance of each network node to obtain the differential privacy budget of each network node.
[0089] The privacy budget coefficient for each cluster can be determined based on its importance. The target privacy budget is then allocated to each cluster based on its privacy budget coefficient. Finally, the privacy budget allocated to each cluster is distributed to the network nodes within that cluster, yielding the differential privacy budget for each network node.
[0090] Step 105: Add noise data to the original dataset according to the differential privacy budget to obtain the dataset to be released.
[0091] In some embodiments, since the value of a network node is partially or fully deterministic of its child nodes, it is possible to add no noise or a small amount of controllable noise to the branch nodes in the Bayesian network. An appropriate amount of noise is added to the leaf nodes to ensure that the Bayesian network N is less affected by noise during sampling. Among them, a branch node refers to a network node with at least one child node, which is the starting point (parent node) of certain edges. Branch nodes play the role of "forking" or "transmitting information" in the network. A leaf node refers to a network node without child nodes, which is the end point of certain edges but has no subsequent edges (child nodes). The leaf node is the "end" of the Bayesian network.
[0092] Based on the above, we can first determine the first allocation ratio for branch nodes and the second allocation ratio for leaf nodes. The second allocation ratio is greater than the first allocation ratio, and the algebraic sum of the first and second allocation ratios is 1. Next, we can calculate the differential privacy budget for each network node based on the target privacy budget allocated to each cluster, the network nodes included in each cluster, the first allocation ratio, and the second allocation ratio.
[0093] In some embodiments, adaptive Laplace noise can be used for noise addition. Specifically, the joint probability distribution of each network node in the Bayesian network is determined. Laplace noise is then added to the joint probability distribution of each network node based on the differential privacy budget of each network node to obtain a noisy joint distribution of each network node. A noise distribution probability is then obtained based on the noisy joint distribution of each network node. Finally, each network node in the Bayesian network is sampled based on the noise distribution probability to obtain a dataset to be published.
[0094] In a specific example, the process of adding noise to the original data set specifically includes:
[0095] First, a privacy budget is allocated to each cluster. According to the principle that the smaller the privacy budget, the higher the privacy protection level, the privacy budget coefficient P of each cluster is calculated based on the importance factor F obtained above and formula (5).
[0096] Formula (5):
[0097] Add the adaptive Laplace distribution to the probability distribution of the multiple AP pairs obtained in step 102 to obtain a probability distribution with distribution. The first allocation ratio of the branch nodes is α, and the second allocation ratio of the leaf nodes is β. α + β = 1. Traverse clusters CL1, CL2, ... CL t All network nodes in i ∈CL j , if the network node A i is a branch node, then the privacy budget ε of the network node i =α∙P(CL j )∙ε. Where ε is the target privacy budget. If the network node is a leaf node, the privacy budget of the network node is ε i =β∙P(CL j )∙ε. The obtained (A i , ε i ) is added to the personalized privacy budget set E.
[0098] Then, initialize the noise probability distribution P*=∅. First, calculate the original joint distribution of each AP pair according to the Bayesian network N, denoted as Pr[A i ,Π i ]. Then, add Laplace noise Lap((2(dk)) / (n ε i )) Get the joint distribution Pr* [A i ,Π i ]=Pr[A i,Π i ]+Lap((2(dk)) / (n ε i )). Among them, it is necessary to change Pr* [A i ,Π i ] is set to 0 to normalize the noisy joint distribution.
[0099] Then, for ∀i=k+1 to d, according to Pr* [A i ,Π i ]Calculate Pr* [A k+1 |Π (k+1) ], ⋯,Pr* [A d |Π d ] and add it to the noise probability distribution P*.
[0100] For ∀i=1 to k, according to Pr* [A (k+1) ,Π (k+1) ] calculated to obtain Pr* [A1 |Π1], ⋯,Pr* [A k |Π k ] and add it to the noise probability distribution P*.
[0101] Finally, for the original dataset D, each data attribute is sampled in turn according to its Bayesian network N and noise probability distribution P* to obtain the data to be released.
[0102] Step 106: Publish the dataset to be published.
[0103] In this embodiment of the present invention, the selection of initial nodes in the Bayesian network is optimized, and data dependencies are mined through information entropy, thereby improving the accuracy and scientific nature of the Bayesian network. Furthermore, a privacy budget allocation strategy based on the independence and importance of data attributes is introduced, enabling dynamic allocation of the privacy budget. This improves the practicality of data release while ensuring data privacy protection. Furthermore, clustering data attributes using a Markov blanket algorithm effectively simplifies the modeling process of complex dependencies in high-dimensional data, significantly improving the efficiency and applicability of Bayesian network models.
[0104] Furthermore, the embodiments of the present invention further reduce the risk of privacy leakage and enhance the security and controllability of data release by adding dynamic noise according to the dependencies between different network nodes.
[0105] Corresponding to the above-mentioned data publishing method based on the Bayesian network, an embodiment of the present invention provides a data publishing device based on the Bayesian network. Figure 3 The following is a schematic diagram of the structure of a data publishing device based on a Bayesian network provided by an embodiment of the present invention. Figure 3 As shown in , the device includes:
[0106] Acquisition module 301 , network construction module 302 , processing module 303 , noise adding module 304 and publishing module 305 .
[0107] The acquisition module 301 is used to acquire an original data set and a corresponding target privacy budget. The original data set includes data of multiple data attributes.
[0108] The network construction module 302 is used to construct a Bayesian network that meets the target privacy budget based on data attributes. The Bayesian network includes multiple network nodes.
[0109] The processing module 303 is configured to determine the importance of each network node based on the distribution of a number of data in the original data set at each network node.
[0110] The processing module 303 is further configured to allocate the target privacy budget to each network node according to the importance of each network node, so as to obtain a differential privacy budget of each network node.
[0111] The noise adding module 304 is used to add noise data to the original data set according to the differential privacy budget to obtain the data to be released.
[0112] Publishing module 305, used to publish the dataset to be published
[0113] Figure 3 The data publishing device based on Bayesian network provided by the embodiment shown can be used to execute Figures 1 to 2 The technical solution of the method embodiment shown, its implementation principle and technical effects can be further referred to the relevant description in the method embodiment.
[0114] Figure 4 A schematic diagram of the structure of an electronic device provided by an embodiment of the present invention is shown in FIG. Figure 4 As shown, the electronic device may include at least one processor and at least one memory in communication with the processor, wherein the memory stores program instructions that can be executed by the processor, and the processor calls the program instructions to execute the instructions in this specification. Figures 1 to 2 The illustrated embodiment provides a data publishing method based on a Bayesian network.
[0115] like Figure 4 As shown, the electronic device is implemented as a general-purpose computing device. Components of the electronic device may include, but are not limited to, one or more processors 410, a communication interface 420, and a memory 430, and a communication bus 440 connecting different system components (including the memory 430, the communication interface 420, and the processor 410).
[0116] Communication bus 440 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures. Examples of these architectures include, but are not limited to, the Industry Standard Architecture (ISA) bus, the Micro Channel Architecture (MAC) bus, the Enhanced ISA bus, the Video Electronics Standards Association (VESA) local bus, and the Peripheral Component Interconnection (PCI) bus.
[0117] Electronic devices typically include a variety of computer system readable media. These media can be any available media that can be accessed by the electronic device, including volatile and non-volatile media, removable and non-removable media.
[0118] Memory 430 may include computer-readable media in the form of volatile memory, such as random access memory (RAM) and / or cache memory. The electronic device may further include other removable or non-removable, volatile or non-volatile computer storage media. Memory 430 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of various embodiments of this specification.
[0119] A program / utility having a set (at least one) of program modules may be stored in memory 430. Such program modules include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. The program modules generally implement the functions and / or methods of the embodiments described herein.
[0120] The processor 410 executes various functional applications and data processing by running the programs stored in the memory 430, such as implementing the Figures 1 to 2 The illustrated embodiment provides a data publishing method based on a Bayesian network.
[0121] The embodiment of this specification provides a computer program product, which includes a computer program. When the computer program is executed by a processor, the computer program performs the following operations: Figures 1 to 2 The illustrated embodiment provides a data publishing method based on a Bayesian network.
[0122] The embodiment of this specification provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer instructions, wherein the computer instructions enable the computer to execute the present specification. Figures 1 to 2 The illustrated embodiment provides a data publishing method based on a Bayesian network.
[0123] The computer-readable storage medium described above may take the form of any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM) or flash memory, optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.
[0124] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.
[0125] In the description of this specification, the reference terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" mean that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of this specification. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art can combine and combine different embodiments or examples described in this specification and features of different embodiments or examples without contradiction.
[0126] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include at least one such feature. Throughout this specification, "plurality" means at least two, such as two or three, unless otherwise specifically defined.
[0127] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of this specification includes alternative implementations in which functions may be performed out of the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of this specification belong.
[0128] The word "if," as used herein, may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting," depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to the determination" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)," depending on the context.
[0129] It should be noted that the devices involved in the embodiments of this specification may include but are not limited to personal computers (Personal Computer; hereinafter referred to as: PC), personal digital assistants (Personal Digital Assistant; hereinafter referred to as: PDA), wireless handheld devices, tablet computers (Tablet Computer), mobile phones, MP3 displays, MP4 displays, etc.
[0130] In the several embodiments provided in this specification, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed may be through some interface, indirect coupling or communication connection of devices or units, which may be electrical, mechanical or other forms.
[0131] In addition, the functional units in the various embodiments of this specification may be integrated into a single processing unit, each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or hardware plus software functional units.
[0132] The aforementioned integrated unit implemented as a software functional unit can be stored in a computer-readable storage medium. The software functional unit is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a connector, or a network device, etc.) or a processor to execute some of the steps of the methods described in various embodiments of this specification. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.
[0133] The above description is only a preferred embodiment of this specification and is not intended to limit this specification. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of this specification should be included in the scope of protection of this specification.
[0134] In this specification, reference can be made to the same or similar parts between the various embodiments. In particular, for the device embodiment and the terminal embodiment, since they are basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the description in the method embodiment.
Claims
1. A data publishing method based on Bayesian network, characterized in that: include: Obtain the original dataset and the corresponding target privacy budget; The original data set includes data of multiple data attributes; Constructing a Bayesian network that meets the target privacy budget according to the data attributes; the Bayesian network includes a plurality of network nodes; Determining the importance of each of the network nodes based on the distribution of a plurality of data in the original data set at each of the network nodes; Allocating the target privacy budget to each network node according to the importance of each network node to obtain a differential privacy budget for each network node; Adding noise data to the original dataset according to the differential privacy budget to obtain a dataset to be released; The to-be-published dataset is published.
2. The method according to claim 1, characterized in that Constructing a Bayesian network that meets the target privacy budget based on the data attributes, including: Calculating the information entropy of each data attribute in the original data set; Determine the target data attribute with the maximum information entropy; Determine the target data attribute as the first node; The Bayesian network is constructed based on the first node.
3. The method according to claim 2, characterized in that The constructing of the Bayesian network based on the first node includes: Determining each of the data attributes as a network node; Determining a set of parent nodes corresponding to each of the network nodes according to the target privacy budget; Obtaining connection relationships and distribution probabilities between the network nodes according to the parent node set corresponding to the network nodes and the head node; The Bayesian network is obtained according to the connection relationship between the various network nodes, the distribution probability and the first node.
4. The method according to claim 1, wherein The determining the importance of each of the network nodes based on the distribution of the plurality of data in the original data set at each of the network nodes includes: Clustering the network nodes according to the data attributes to obtain a plurality of disjoint clusters; Calculating the information entropy of each cluster; Determining the importance of each cluster according to the information entropy of each cluster; The importance of each network node is determined according to the importance of each cluster and the network nodes included in each cluster.
5. The method according to claim 4, characterized in that Allocating the target privacy budget to each network node according to the importance of each network node to obtain a differential privacy budget for each network node includes: Determine a privacy budget coefficient for each cluster according to the importance of each cluster; Allocating the target privacy budget to each cluster according to the privacy budget coefficient of each cluster; The privacy budget allocated to each cluster is allocated to the network nodes included in each cluster to obtain the differential privacy budget of each network node.
6. The method according to claim 5, characterized in that The network nodes in the Bayesian network include branch nodes and leaf nodes; allocating the privacy budgets allocated to the clusters to the network nodes included in the clusters to obtain differential privacy budgets of the network nodes includes: Determining a first allocation ratio for the branch nodes; and determining a second allocation ratio for the leaf nodes; wherein the second allocation ratio is greater than the first allocation ratio; The differential privacy budget of each network node is obtained according to the target privacy budget allocated to each cluster, the network nodes included in each cluster, the first allocation ratio, and the second allocation ratio.
7. The method according to claim 1, characterized in that Adding noise data to the original data set according to the differential privacy budget includes: Determining a joint probability distribution of each network node in the Bayesian network; Adding Laplace noise to the joint probability distribution of each network node according to the differential privacy budget of each network node to obtain a noisy joint distribution of each network node; Obtaining a noise distribution probability according to the noisy joint distribution of each network node; Each network node in the Bayesian network is sampled according to the noise distribution probability to obtain the data set to be released.
8. A data publishing device based on Bayesian network, characterized in that: include: The acquisition module is used to obtain the original data set and the corresponding target privacy budget; The original data set includes data of multiple data attributes; A network construction module, configured to construct a Bayesian network that meets the target privacy budget based on the data attributes; the Bayesian network includes a plurality of network nodes; a processing module, configured to determine the importance of each of the network nodes based on the distribution of a plurality of data in the original data set at each of the network nodes; The processing module is further configured to allocate the target privacy budget to each network node according to the importance of each network node, so as to obtain a differential privacy budget of each network node; A noise adding module, configured to add noise data to the original dataset according to the differential privacy budget to obtain data to be published; A publishing module is used to publish the data set to be published.
9. An electronic device, characterized in that: include: at least one processor; as well as at least one memory in communication with the processor, wherein: The memory stores program instructions that can be executed by the processor, and the processor can execute the method according to any one of claims 1 to 7 by calling the program instructions.
10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores computer instructions, and the computer instructions enable the computer to execute the method according to any one of claims 1 to 7.