A method and device for hiding network structure based on personality information in a social network

Abstract

The application belongs to the field of big data analysis, and particularly relates to a network structure hiding method and device based on personality information in a social network. The network structure hiding method comprises the following steps: S1: converting a user relationship graph in a social network into graph structure data; S2: generating a five-dimensional personality vector of each user through a text analysis tool based on social content published by each user; S3: constructing a personality feature matrix for representing each personality of all users based on each centrality index of the network structure by using a multivariate linear regression method; S4: generating an anonymous k-degree sequence with the largest personality change in the degree component through an evolutionary algorithm; and S5: modifying the original graph G according to the obtained anonymous k-degree sequence, so as to generate a new graph G' satisfying the k-degree anonymous graph. The application solves the problem that information publishing in the existing social network may leak user personality features or private information.

Description

Technical Field

[0001] The present invention belongs to the field of big data analysis, and specifically relates to a method for hiding a network structure based on personality information in a social network, and a user relationship graph preprocessing device based on personality information using the method. Background Art

[0002] In recent years, with the rapid development of mobile communications and network technologies, digital virtual life has gradually become an integral part of people's daily lives. User interactions on social networking platforms generate a vast amount of data. This data can be used to analyze relationships between users and their preferences. By thoroughly analyzing this data, we can generate a user relationship graph in graphical format. In this user relationship graph, each node represents a user, and the edges or links between nodes represent the relationships between users or entities in a richer and more intuitive way.

[0003] The full disclosure of information on social networking platforms also poses certain security risks. For example, cyber attackers can use technologies such as crawlers to obtain large amounts of timely social network relationship data. These attackers can then use this data to understand user relationships and analyze user personality information. By gaining access to the structure of a user's social network and inferring their corresponding personality, attackers can push targeted information to the user. This information push can serve both positive economic purposes, such as advertising, and potentially be used for illicit purposes. Therefore, protecting user privacy on social networks and preventing the inaccurate extraction of user personality traits and preferences has become a pressing technical challenge for social networking service providers. Generally speaking, one approach to protecting user information on social networks is to anonymize some users or their posted content, thereby eliminating the risk of user relationships or user characteristics being easily extracted. However, no method exists for effectively anonymizing information posted on social networks that conceals user personality traits. Summary of the Invention

[0004] In order to overcome the problem that information release in existing social networks may leak user personality characteristics or privacy information, the present invention provides a network structure hiding method based on personality information in social networks, and a user relationship graph preprocessing device based on personality information.

[0005] The present invention is achieved by adopting the following technical solutions:

[0006] A method for hiding a network structure based on personality information in a social network, comprising the following steps:

[0007] S1: Generate a corresponding user relationship graph based on the relationship between users in the social network; convert the user relationship graph into a graph structure data G: G = (V, E).

[0008] S2: Based on the social content posted by each user, a text analysis tool is used to generate a five-dimensional personality vector for each user. The five-dimensional personality vector includes agreeableness, conscientiousness, extraversion, neuroticism, and openness.

[0009] S3: Using multiple linear regression, we construct a personality matrix to characterize the personality traits of all users based on various centrality indicators of the network structure. These centrality indicators include degree centrality, betweenness centrality, closeness centrality, eigenvector centrality, and local clustering coefficient.

[0010] S4: Generate an anonymous k-degree sequence based on the graph G through an evolutionary algorithm that meets the preset anonymity value k and has the largest personality change in the degree component.

[0011] S5: Modify the original graph G accordingly according to the obtained anonymous k-degree sequence, so as to generate a new graph G′ that satisfies the k-degree anonymous graph.

[0012] As a further improvement of the present invention, in the graph G of step S1, V is the user set, E is the link set between users, n=|V| is the number of nodes, and m=|E| is the number of links.

[0013] As a further improvement of the present invention, in step S2, the method for generating the five-dimensional personality vector is as follows:

[0014] Collect all information posted by users on social networks and extract text data from them. Preprocess the text data by removing non-alphabetic characters and converting all capital letters to lowercase. Finally, analyze the preprocessed text data using a word-based text analysis tool to generate scores for the user's personality traits across five dimensions, which form the required five-dimensional personality vector.

[0015] As a further improvement of the present invention, in the centrality index of step S3, degree centrality represents the number of direct connections between the current node and other nodes. i The standardized degree centrality d(v i , A) is calculated as follows:

[0016]

[0017] Among them, A is the adjacency matrix of graph G, with dimension n*n; a ij is the element in the i-th row and j-th column of the adjacency matrix A.

[0018] Betweenness centrality indicates the proportion of the shortest path between other node pairs in which the current node appears, and is used to measure the control ability of the node on the spread of other nodes. i The standardized betweenness centrality b(v i , A) is calculated as follows:

[0019]

[0020] Among them, l jk (i) is the slave node v j To node v k The shortest path passes through node v i The number of l jk Node v j To node v k The number of shortest paths.

[0021] Close centrality indicates the closeness of the current node to all other nodes, which is the inverse of the mean of the node to all other nodes. i The closeness centrality c(v i , A) is calculated as follows:

[0022]

[0023] Among them, d ij For node v i To node v j The shortest distance.

[0024] Eigenvector centrality reflects the importance of a node; node v i The eigenvector centrality v(v i , A) is calculated as follows:

[0025] p is The element in row i of For r max The corresponding n*1 dimensional vector.

[0026] The local clustering coefficient reflects the degree of interconnection between neighboring nodes of a node. i The local clustering coefficient clu(v i , A) is calculated as follows:

[0027]

[0028] Among them, V i For node v i The set of adjacent points of .

[0029] As a further improvement of the present invention, in step S3, a personality feature matrix for characterizing the agreeableness of all users is constructed based on the degree centrality, betweenness centrality, and local clustering coefficient of each node. The formula is as follows:

[0030]

[0031] in, is the residual, which represents the difference between the user's actual agreeable personality value and the agreeable personality value obtained by multiple linear regression, α1, α2, α3, C agre is a constant, is an n*1 dimensional vector of all 1s.

[0032] Based on the degree centrality, closeness centrality, betweenness centrality, and local clustering coefficient of each node, a personality characteristic matrix is ​​constructed to characterize the sense of responsibility of all users. The formula is as follows:

[0033]

[0034] in, is the residual, which represents the difference between the user's real sense of responsibility personality value and the sense of responsibility personality value obtained by multiple linear regression, β1, β2, β3, β4, C cons is a constant.

[0035] Based on degree centrality, closeness centrality, betweenness centrality and eigenvector centrality, a personality trait matrix is ​​constructed to characterize the extraversion of all users. The formula is as follows:

[0036]

[0037] in, is the residual, which represents the difference between the user's actual extraversion personality value and the extraversion personality value obtained by multiple linear regression, γ1, γ2, γ3, γ4, C extr is a constant.

[0038] A personality trait matrix is ​​constructed based on degree centrality, closeness centrality, betweenness centrality, and eigenvector centrality to characterize the neuroticism of all users. The formula is as follows:

[0039]

[0040] in, is the residual, which represents the difference between the user's true neurotic personality value and the appropriate neurotic personality value obtained by multiple linear regression, v1, δ2, δ3, v4, C neur is a constant.

[0041] Based on degree centrality and betweenness centrality, a personality characteristic matrix is ​​constructed to characterize the openness of all users. The formula is as follows:

[0042]

[0043] in, is the residual, which represents the difference between the user's true openness personality value and the openness personality value obtained by multiple linear regression, θ1, θ2, C open is a constant.

[0044] As a further improvement of the present invention, the construction process of the anonymous k-degree sequence is as follows:

[0045] S41: Get the degree of n nodes in the graph G, and construct a degree sequence d = {d1, d2, ..., d n}, where d i is the degree of node i; set the number of iterations in the evolution process epcoh and the anonymity value k.

[0046] S42: Get the degree d of the node in the current graph G i The number of nodes num that is greater than the preset anonymity value k, and the ratio of num to the total number of nodes n is used as the score value radio of the graph G.

[0047] S42: Initialize the system parameters of the evolution process, including the maximum score max_radio, the best degree sequence best_d, and the maximum change in personality in the degree component best_cange. The initial values ​​of the system parameters are as follows:

[0048] max_radio=radio

[0049] best_d=d

[0050] best_cange=0.

[0051] S43: Perform mutation operation according to the preset number of iterations epcoh. In each round of mutation operation, randomly select the degree values ​​of two nodes in the graph G to update. The updating process increases the degree of one node by 1 and decreases the degree of the other node by 1. The degree sequence d after each round of mutation is obtained. i .

[0052] S44: Update the degree sequence d after each round of mutation using the method in step S42 i Corresponding score value radio i and compare it with max_radio: When radio i >max_radio, then for d i Perform normalization to obtain the normalized degree sequence d_nor i , and then calculate the change in the degree component of the hidden personality Δd_nor i .

[0053] S45: Determine the change Δd_nor of each personality trait in each round of mutation i Is it greater than best_cange? If so, update the system parameters. The system parameters are updated as follows:

[0054] max_radio=radio i

[0055] best_d=d i

[0056] best_cange=Δd_nor i .

[0057] S46: Get the d of each node after the system parameters are updated i The number of occurrences c in d i , and then construct a sequence of times c = {c1, c2, ..., c n}; and calculate the anonymity value k_now of the graph G in the current state by the following formula:

[0058] k_now=min(c).

[0059] S47: Determine whether k is satisfied now ≥k, then the best degree sequence best_d is output as the required anonymous k-degree sequence; otherwise, the process returns to step S43 and performs mutation operation on the graph G again until the preset number of iterations epcoh is reached.

[0060] As a further improvement of the present invention, in step S44, each node v in the degree sequence is sequentially i degree d i Normalize and get the normalized degree sequence d_nor:

[0061] d_nor={d_nor1, d_nor2,…, d_nor n};

[0062] The normalized calculation formula used is as follows:

[0063]

[0064] As a further improvement of the present invention, in step S44, when different human features are hidden, the calculation method of the change in the degree component is as follows:

[0065] (1) Change in Agreeableness when Hidden Δd_nor i The calculation formula is as follows:

[0066]

[0067] Among them, d_nor j Represents the values ​​in the normalized degree sequence before the mutation operation, and j is the sequence number of the element in the degree sequence; Represents the values ​​in the normalized degree sequence before the i-th mutation operation iteration, where i represents the number of mutation operation iterations.

[0068] (2) Change in conscientious personality when it is hidden Δd_nor i The calculation formula is as follows:

[0069]

[0070] (3) Change in Extraversion when Hidden Δd_nor i The calculation formula is as follows:

[0071]

[0072] (4) Change in neurotic personality when hidden Δd_nor i The calculation formula is as follows:

[0073]

[0074] (5) Change in openness when hiding personality Δd_nor i The calculation formula is as follows:

[0075]

[0076] As a further improvement of the present invention, in step S5, the process of modifying the graph G to obtain a new graph G′ that satisfies the k-degree anonymous graph is as follows:

[0077] S51: Get the deviation between the initial degree sequence d and the best degree sequence best_d corresponding to the graph G, node v i Degree deviation The calculation formula is as follows:

[0078]

[0079] S52: Determine the degree deviation of each node Does the value of If yes, then the value Add to the collection to be removed The node v i Put it into set V del ; otherwise the value Add to the collection to be added The node v i Put it into set V add .

[0080] S53: Get the neighbor nodes of each node in the graph G. Among them, node v p The set of neighbor nodes As shown in the following formula:

[0081]

[0082] S54: Get the edge set E in the graph G, and based on the set to be removed Collections to be added Collection V del 、Collection V add Remove or create new edges between each point and its neighboring nodes to complete the modification of graph G.

[0083] As a further improvement of the present invention, the method for adjusting the edges in the network structure in step S54 is as follows:

[0084] S541: Initialize the edge set E′ so that E′=E; and obtain V del 、V add Elements in .

[0085] S542: From Any value in And from V del Find the corresponding node v p ; Then from v p The set of neighbor nodes Select node v q , remove node v p With node v q Links (v p , v q ), (v p , v q ) is removed from E′. The update formula of the removal process E′ is:

[0086] E′=E′\{(v p , v q )}.

[0087] S543: From Any value in And from V add Find the corresponding node v r , construct node v q With node v r Links (v q , v r ), (v q , v r ) is added to E′. The update formula of the addition process E′ is:

[0088] E′=E′∪{(v q , v r )}.

[0089] S544: After removing and building edges, node v p , v q , v r The degree of d′ p , d′ q , d′ r Indicates that the update formula of each node degree is as follows:

[0090] d′ p =d p -1

[0091] d′ q =d q +1-1=d q

[0092] d′ r =d r +1.

[0093] S545: After removing and building edges, node v p , v q , v r The degree deviation is respectively expressed as Indicates that the update formula of degree deviation is as follows:

[0094]

[0095]

[0096]

[0097] S546: Determine whether the following conditions are met: If so, v p From V del Remove, the processing formula is as follows:

[0098]

[0099] V del =V del \{v p};

[0100] Otherwise The value of is modified as follows:

[0101]

[0102] S547: Determine whether the following conditions are met: If so, v r From V add Remove, the processing formula is as follows:

[0103]

[0104] V add =V add \{v r};

[0105] Otherwise The value of is modified as follows:

[0106]

[0107] S548: Judgment Set Are both empty sets? If yes, output the adjusted node set V′ and the current E′ to form the required new graph G′ that satisfies the k-degree anonymous graph; otherwise, repeat the processing of steps S542-S547.

[0108] The present invention also includes a user relationship graph preprocessing device based on personality information, which includes a memory, a processor, and a computer program stored in the memory and runable on the processor. When the processor executes the computer program, it implements the steps of the aforementioned network structure hiding method based on personality information in a social network, and then adjusts the network structure graph G representing the user relationship graph to obtain a new graph G′ that meets the requirements of a k-degree anonymous graph; and anonymizes key information related to user privacy contained in the original graph G.

[0109] The technical solution provided by the present invention has the following beneficial effects:

[0110] This invention combines the network structure of existing social platforms with user personality trait vectors based on platform content to construct a new personality trait matrix. An evolutionary algorithm is then used to generate a graph that maximizes personality trait variation while satisfying k-degree anonymity. This matrix serves as the basis for hiding features within the original network structure. The method provided by this invention uses user personality evaluation as a guide for adjusting the network architecture, thereby fully deconstructing the personality traits hidden within the original network structure and protecting the privacy of each user within the network. BRIEF DESCRIPTION OF THE DRAWINGS

[0111] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0112] Figure 1This is a flowchart of a method for hiding a network structure based on personality information in a social network provided in Example 1 of the present invention.

[0113] Figure 2 This is a step diagram of a method for hiding a network structure based on personality information in a social network provided in Example 1 of the present invention.

[0114] Figure 3 This is a flowchart of the anonymous k-degree sequence construction process in Example 1 of the present invention.

[0115] Figure 4 This is a flowchart of the process of adjusting the original graph G based on the anonymous k-degree sequence in Example 1 of the present invention.

[0116] Figure 5 This is a flowchart of a process for adjusting edges in a network structure based on a set to be removed and a set to be added in embodiment 1 of the present invention. DETAILED DESCRIPTION

[0117] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0118] Example 1

[0119] This embodiment provides a method for hiding the network structure based on personality information in a social network. Figure 1 As shown in the figure, the overall idea of ​​the network structure hiding method is as follows: (1) Based on the known relationships between users and the collected content published by users, the relationship graph between users is analyzed and converted into graph structure data. (2) The content published by users is used as the basic information for analyzing the personality characteristics of users. Then, the personality characteristics of each user's Big Five personality (Ocean), that is, the five-dimensional personality vector, are constructed. (3) A reliable correlation function is constructed between the five-dimensional personality vectors of all users and the centrality index of the network structure through the multivariate linear regression method, that is, the corresponding personality characteristic matrix. (4) Continuously select a user-to-user link to simulate the dismantling or reorganization, evaluate the operation method that has the greatest impact on the user personality vector in the network structure, and use this as a guide for the network structure reorganization method used to achieve the hiding of the user's true personality. The simulation process can be iterated continuously using an evolutionary algorithm, and the required k-degree anonymity sequence is obtained during the iterative process. (5) The k-degree anonymity sequence is used as the basis for the final processing of the original graph structure data. The processing process mainly adjusts the edges contained in the original graph structure (that is, the links between nodes). The format of the original graph structure is as follows: G = (V, E), where V is the user set and E is the link set between users. After processing, a new graph G′ is obtained: G′=(V′, E′).

[0120] Considering that in this embodiment, the personality characteristics of users in the social networking platform are closely related to the network structure, after removing or adding some links between users, the true personality characteristics of each user will be hidden, thereby protecting the user's privacy information on the social networking platform.

[0121] Specifically, if Figure 2 As shown, the network structure hiding method includes the following steps:

[0122] S1: Generate a corresponding user relationship graph based on the relationship between users in the social network; convert the user relationship graph into a graph structure data G: G = (V, E).

[0123] In social platforms, the relationship between users is a typical network structure data. In the graph G representing the user network obtained in this embodiment, V = {v1, v2, v3, ... v i …v n}, n = |V| is the number of nodes. V is the network node set, including all users on the platform. E is the set of links between users, and m = |E| is the number of links. The link set reflects the interactive relationships between all users on the platform.

[0124] S2: Based on the social content posted by each user, a text analysis tool is used to generate a five-dimensional personality vector for each user. The five-dimensional personality vector includes agreeableness, conscientiousness, extraversion, neuroticism, and openness.

[0125] Specifically, the five-dimensional personality vector is generated as follows:

[0126] Collect all information posted by users on social networks and extract text data from them. Preprocess the text data by removing non-alphabetic characters and converting all capital letters to lowercase. Finally, analyze the preprocessed text data using a word-based text analysis tool to generate scores for the user's personality traits across five dimensions, which form the required five-dimensional personality vector.

[0127] In this embodiment, LIWC software is used to analyze the collected user text information to obtain the five-dimensional personality representation vector of each user. * ), the form of its five-dimensional personality representation vector is: [agre i ,cons i , extr i ,neur i , open i ]; This array reflects the numerical manifestation of user i's agreeableness, conscientiousness, extraversion, neuroticism and openness.

[0128] S3: Using multiple linear regression, we construct a personality matrix to characterize the personality traits of all users based on various centrality indicators of the network structure. These centrality indicators include degree centrality, betweenness centrality, closeness centrality, eigenvector centrality, and local clustering coefficient.

[0129] Among them, degree centrality indicates the number of direct connections between the current node and other nodes. i The standardized degree centrality d(v i , A) is calculated as follows:

[0130]

[0131] Among them, A is the adjacency matrix of graph G, with dimension n*n; a ij is the element in the i-th row and j-th column of the adjacency matrix A.

[0132] Betweenness centrality indicates the proportion of the shortest path between other node pairs in which the current node appears, and is used to measure the control ability of the node on the spread of other nodes. i The standardized betweenness centrality b(v i , A) is calculated as follows:

[0133]

[0134] Among them, l jk (i) is the slave node v j To node v k The shortest path passes through node v i The number of l jk Node v j To node v k The number of shortest paths.

[0135] Close centrality indicates the closeness of the current node to all other nodes, which is the inverse of the mean of the node to all other nodes. i The closeness centrality c(v i , A) is calculated as follows:

[0136]

[0137] Among them, d ij For node v i To node v j The shortest distance.

[0138] Eigenvector centrality reflects the importance of a node; node v i The eigenvector centrality v(v i , A) is calculated as follows:

[0139] p is The element in row i of is the n*l dimensional vector corresponding to rmax.

[0140] The local clustering coefficient reflects the degree of interconnection between neighboring nodes of a node. i The local clustering coefficient clu(v i , A) is calculated as follows:

[0141]

[0142] Among them, V i For node v i The set of adjacent points of .

[0143] Next, based on the degree centrality, betweenness centrality, and local clustering coefficient of each node, a personality feature matrix is ​​constructed to characterize the agreeableness of all users. The formula is as follows:

[0144]

[0145] in, is the residual, which represents the difference between the user's actual agreeable personality value and the agreeable personality value obtained by multiple linear regression, α1, α2, α3, C agre is a constant, is an n*1 dimensional vector of all 1s.

[0146] Based on the degree centrality, closeness centrality, betweenness centrality, and local clustering coefficient of each node, a personality characteristic matrix is ​​constructed to characterize the sense of responsibility of all users. The formula is as follows:

[0147]

[0148] in, is the residual, which represents the difference between the user's real sense of responsibility personality value and the sense of responsibility personality value obtained by multiple linear regression, β1, β2, β3, β4, C cons is a constant.

[0149] Based on degree centrality, closeness centrality, betweenness centrality and eigenvector centrality, a personality trait matrix is ​​constructed to characterize the extraversion of all users. The formula is as follows:

[0150]

[0151] in, is the residual, which represents the difference between the user's actual extraversion personality value and the extraversion personality value obtained by multiple linear regression, γ1, γ2, γ3, γ4, C extris a constant.

[0152] A personality trait matrix is ​​constructed based on degree centrality, closeness centrality, betweenness centrality, and eigenvector centrality to characterize the neuroticism of all users. The formula is as follows:

[0153]

[0154] in, is the residual, which represents the difference between the user's true neurotic personality value and the neurotic personality value obtained by multiple linear regression, v1, v2, δ3, v4, C neur is a constant.

[0155] Based on degree centrality and betweenness centrality, a personality characteristic matrix is ​​constructed to characterize the openness of all users. The formula is as follows:

[0156]

[0157] in, is the residual, which represents the difference between the user's true openness personality value and the openness personality value obtained by multiple linear regression, θ1, θ2, C open is a constant.

[0158] S4: Generate an anonymous k-degree sequence based on the graph G through an evolutionary algorithm that meets the preset anonymity value k and has the largest personality change in the degree component.

[0159] For a given graph G and a positive integer k, the graph G is called a k-degree anonymous graph if and only if every node in the graph has at least k-1 nodes with the same degree, and its degree sequence is called a k-degree anonymous degree sequence. This embodiment uses an evolutionary algorithm to find an anonymous k-degree sequence that satisfies the preset anonymity k and can achieve the maximum change in user personality.

[0160] like Figure 3 As shown, the construction process of the anonymous k-degree sequence is as follows:

[0161] S41: Get the degree of n nodes in the graph G, and construct a degree sequence d = {d1, d2, ..., d n}, where d i is the degree of node i; set the number of iterations in the evolution process epcoh and the anonymity value k.

[0162] S42: Get the degree d of the node in the current graph G i The number of nodes num that are greater than the preset anonymity value k, and the ratio of num to the total number of nodes n is used as the score value radio of the graph G:

[0163]

[0164] S42: Initialize the system parameters of the evolution process, including the maximum score max_radio, the best degree sequence best_d, and the maximum change in personality in the degree component best_cange. The initial values ​​of the system parameters are as follows:

[0165] max_radio=radio

[0166] best_d=d

[0167] best_cange=0.

[0168] S43: Perform mutation operation according to the preset number of iterations epcoh. In each round of mutation operation, two nodes v are randomly selected from n nodes. i , v j , and according to d, node v i The degree of v is increased by 1, j The degree of N is reduced by 1, and the i-th (1≤i≤epochandi∈ N * ) times the degree sequence after the change is recorded as d i The corresponding formula is as follows:

[0169] d i =d i +1

[0170] d j =d j -1

[0171] S44: Update the degree sequence d after each round of mutation using the method in step S42 i Corresponding score value radio i and compare it with max_radio: When radio i >max_radio, then for d i Perform normalization to obtain the normalized degree sequence d_nor i , and then calculate the change in the degree component of the hidden personality Δd_nor i .

[0172] Among them, each node v in the degree sequence is sequentially i degree d i By normalizing, we can get the normalized degree sequence d_nor:

[0173] d_nor={d_nor1, d_nor2,…, d_nor n};

[0174] The calculation formula used for normalization is as follows:

[0175]

[0176] After obtaining the normalized sequence, the calculation method for the change in the degree component when different human features are hidden is as follows:

[0177] (1) Change in Agreeableness when Hidden Δd_nor i The calculation formula is as follows:

[0178]

[0179] Among them, d_nor j Represents the values ​​in the normalized degree sequence before the mutation operation, and j is the sequence number of the element in the degree sequence; Represents the values ​​in the normalized degree sequence before the i-th mutation operation iteration, where i represents the number of mutation operation iterations.

[0180] (2) Change in conscientious personality when it is hidden Δd_nor i The calculation formula is as follows:

[0181]

[0182] (3) Change in Extraversion when Hidden Δd_nor i The calculation formula is as follows:

[0183]

[0184] (4) Change in neurotic personality when hidden Δd_nor i The calculation formula is as follows:

[0185]

[0186] (5) Change in openness when hiding personality Δd_nor i The calculation formula is as follows:

[0187]

[0188] S45: Determine the change Δd_nor of each personality trait in each round of mutation i Is it greater than best_cange? If so, update the system parameters. The system parameters are updated as follows:

[0189] max_radio=radio i

[0190] best_d=d i

[0191] best_cange=Δd_nor i .

[0192] S46: Get the d of each node after the system parameters are updated i The number of occurrences c in d i , and then construct a sequence of times c = {c1, c2, ..., c n}; and calculate the anonymity value k_now of the graph G in the current state by the following formula:

[0193] k_now=min(c).

[0194] S47: Determine whether k is satisfied now ≥k, then the best degree sequence best_d is output as the required anonymous k-degree sequence; otherwise, the process returns to step S43 and performs mutation operation on the graph G again until the preset number of iterations epcoh is reached.

[0195] S5: Modify the original graph G accordingly based on the obtained anonymous k-degree sequence, thereby generating a new graph G′ that satisfies the k-degree anonymous graph. Figure 4 As shown in , the process of modifying graph G to obtain a new graph G′ that satisfies the k-degree anonymous graph is as follows:

[0196] S51: Get the deviation between the initial degree sequence d and the best degree sequence best_d corresponding to the graph G, node v i Degree deviation The calculation formula is as follows:

[0197]

[0198] S52: Determine the degree deviation of each node Does the value of If yes, then the value Add to the collection to be removed The node v i Put it into set V del ; otherwise the value Add to the collection to be added The node v i Put it into set V add .

[0199] S53: Get the neighbor nodes of each node in the graph G. Among them, node v p The set of neighbor nodes As shown in the following formula:

[0200]

[0201] S54: Get the edge set E in the graph G, and based on the set to be removed Collections to be added Collection V del 、Collection Vadd Remove or create new edges between each point and its neighboring nodes to complete the modification of graph G.

[0202] like Figure 5 As shown, the method for adjusting the edges in the network structure is as follows:

[0203] S541: Initialize the edge set E′ so that E′=E; and obtain V del 、V add Elements in .

[0204] S542: From Any value in And from V del Find the corresponding node v p ; Then from v p The set of neighbor nodes Select node v q , remove node v p With node v q Links (v p , v q ), (v p , v q ) is removed from E′. The update formula of the removal process E′ is:

[0205] E′=E′\{(v p , v q )}.

[0206] S543: From Any value in And from V add Find the corresponding node v r , construct node v q With node v r Links (v q , v r ), (v q , v r ) is added to E′. The update formula of the addition process E′ is:

[0207] E′=E′∪{(v q , v r )}.

[0208] S544: After removing and building edges, node v p , v q , v r The degree of d′ p , d′ q , d′ r Indicates that the update formula of each node degree is as follows:

[0209] d′ p =d p -1

[0210] d′ q =d q +1-1=d q

[0211] d′ r =d r +1.

[0212] S545: After removing and building edges, node v p , v q , v r The degree deviation is respectively expressed as Indicates that the update formula of degree deviation is as follows:

[0213]

[0214]

[0215]

[0216] S546: Determine whether the following conditions are met: If so, v p From V del Remove, the processing formula is as follows:

[0217]

[0218] V del =V del \{v p};

[0219] Otherwise The value of is modified as follows:

[0220]

[0221] S547: Determine whether the following conditions are met: If so, v r From V add Remove, the processing formula is as follows:

[0222]

[0223] V add =V add \{v r};

[0224] Otherwise The value of is modified as follows:

[0225]

[0226] S548: Judgment Set Are both empty sets? If yes, output the adjusted node set V′ and the current E′ to form the required new graph G′ that satisfies the k-degree anonymous graph; otherwise, repeat the processing of steps S542-S547.

[0227] The user personality information hiding method proposed in this paper, based on a designed network structure, associates user personality traits with centrality metrics in the inter-user network structure. Using an evolutionary algorithm, a k-degree anonymity sequence is generated, using the weight of the influence of user personality traits as a constraint. Finally, this k-degree anonymity sequence is used as a guide to restructure the original graph reflecting user relationships, resulting in a new graph that hides any leaked user personality traits, thereby protecting user privacy.

[0228] On one hand, the graph anonymization technology employed in this invention converts the original social network graph into a k-degree anonymous graph, thereby protecting users within the social network to a certain extent. After anonymization, even if an attacker successfully learns a user's friendship relationships and the degree of their nodes in the graph, there are at least k nodes with the same degree in the k-degree anonymous graph. Therefore, the attacker cannot determine the true degree of each node in the graph, thus protecting the user's personality across the entire network.

[0229] On the other hand, the present invention takes into account that every user participating in a social network will build relationships with other users on the social network for communication and will publish personal content (pictures, text). The connections between users build a social network, and each person's personality is correlated with the centrality index of the node representing the user in the network. Therefore, by removing some user links and building links between some users, the network structure can be changed to hide each user's true personality.

[0230] Example 2

[0231] Based on Example 1, this embodiment further provides a user relationship graph preprocessing device based on personality information. The preprocessing device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. The preprocessing device is primarily configured to implement the steps of the method for hiding the network structure based on personality information in a social network as described in Example 1, thereby adjusting the network structure graph G representing the user relationship graph to obtain a new graph G′ that meets the k-degree anonymity requirement.

[0232] The purpose of the pre-processing device provided in this embodiment generating the graph G′ is mainly to provide a reference for the platform's decision on anonymizing the information posted by users. When certain users and the content they post pose a risk of leaking the personality characteristics of the user or user group, the platform should protect the security of the content, anonymize it, or set up obstacles for the information to be captured by crawlers; thereby achieving the purpose of protecting key information in user privacy.

[0233] In the actual deployment of this apparatus, the computer device may be a smartphone, tablet computer, laptop computer, desktop computer, rack server, blade server, tower server, or cabinet server (including a standalone server or a server cluster consisting of multiple servers), etc., capable of executing programs. The computer device of this embodiment includes at least, but is not limited to, a memory and a processor that can be interconnected via a system bus.

[0234] In this embodiment, the memory (i.e., readable storage medium) includes flash memory, hard disk, multimedia card, card-type memory (e.g., SD or DX memory), random access memory (RAM), static random access memory (SRAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), programmable read-only memory (PROM), magnetic storage, magnetic disk, optical disk, etc. In some embodiments, the memory may be an internal storage unit of a computer device, such as the hard disk or internal memory of the computer device. In other embodiments, the memory may also be an external storage device of the computer device, such as a plug-in hard disk equipped with the computer device, a smart media card (SMC), a secure digital (SD) card, a flash memory card, etc. Of course, the memory may also include both the internal storage unit of the computer device and its external storage device. In this embodiment, the memory is generally used to store the operating system and various application software installed on the computer device. In addition, the memory may also be used to temporarily store various types of data that have been output or are about to be output.

[0235] In some embodiments, the processor may be a central processing unit (CPU), a controller, a microcontroller, a microprocessor, or other data processing chip. The processor is generally used to control the overall operation of a computer device. In this embodiment, the processor is used to run program code stored in a memory or process data. When the processor executes the computer program, the steps of the method for hiding the network structure based on personality information in a social network as described in Example 1 are implemented, thereby adjusting the network structure graph G representing the user relationship graph to obtain a new graph G′ that meets the requirements of a k-degree anonymous graph; and anonymizing key information related to user privacy contained in the original graph G.

[0236] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for hiding network structure based on personality information in a social network, characterized in that: The network structure hiding method comprises the following steps: S1: Generate a corresponding user relationship graph based on the relationship between users in the social network; convert the user relationship graph into a graph structure data graph G: G = (V, E); S2: Based on the social content posted by each user, a five-dimensional personality vector of each user is generated using a text analysis tool; the five-dimensional personality vector includes agreeableness, conscientiousness, extraversion, neuroticism, and openness; S3: Using a multiple linear regression method, a personality characteristic matrix is ​​constructed based on various centrality indicators of the network structure to characterize the personality of all users. The centrality indicators include degree centrality, betweenness centrality, closeness centrality, eigenvector centrality, and local clustering coefficient. S4: Based on the graph G, an anonymous k-degree sequence is generated by an evolutionary algorithm, which satisfies a preset anonymity value k and has the largest personality change in degree components. The anonymous k-degree sequence is constructed as follows: S41: Get the degrees of n nodes in the graph G and construct a degree sequence d = {d1, d2, ..., d n }, where d i is the degree of node i; set the number of iterations epcoh and the anonymity value k in the evolution process; S42: Get the degree d of the node in the current graph G i The number of nodes num that are greater than the preset anonymity value k, and the ratio of num to the total number of nodes n is used as the score value radio of the graph G; S42: Initialize the system parameters of the evolution process, including the maximum score max_radio, the best degree sequence best_d, and the maximum change in personality in the degree component best_cange. The initial values ​​of the system parameters are as follows: max_radio=radio best_d=d best_cange = 0; S43: Perform mutation operation according to the preset number of iterations epcoh. In each round of mutation operation, randomly select the degree values ​​of two nodes in the graph G to update. The updating process increases the degree of one node by 1 and decreases the degree of the other node by 1. The degree sequence d after each round of mutation is obtained. i ; S44: Update the degree sequence d after each round of mutation using the method in step S42 i Corresponding score value radio i and compare it with max_radio: When radio i >max_radio, then d i Perform normalization to obtain the normalized degree sequence d_nor i , and then calculate the change in the degree component of the hidden personality Δd_nor i ; S45: Determine the change Δd_nor of each personality trait in each round of mutation i Is it greater than best_cange? If so, update the system parameters. The system parameters are updated as follows: max_radio=radio i best_d=d i best_cange=Δd_nor i ; S46: Get the d of each node after the system parameters are updated i The number of occurrences c in d i , and then construct a frequency sequence c={c1,c2,…,c n }; and calculate the anonymity value k_now of the graph G in the current state by the following formula: k_now=min(c) S47: Determine whether k is satisfied now ≥k, then the best degree sequence best_d is output as the required anonymous k-degree sequence; otherwise, the process returns to step S43 and performs mutation operation on the graph G again until the preset number of iterations epcoh is reached; S5: Modify the original graph G accordingly according to the acquired anonymous k-degree sequence, thereby generating a new graph G′ that satisfies the k-degree anonymous graph.

2. The method for hiding a network structure based on personality information in a social network according to claim 1, wherein: In the graph G of step S1, V is the user set, E is the link set between users, n = |V| is the number of nodes, and m = |E| is the number of links.

3. The method for hiding a network structure based on personality information in a social network according to claim 2, wherein: In step S2, the five-dimensional personality vector is generated as follows: All information posted by users on social networks is collected and text data from the information is extracted; the text data is preprocessed to remove non-alphabetic characters and uniformly adjust uppercase letters to lowercase; finally, a text analysis tool based on word measurement is used to analyze the preprocessed text data to generate personality score values ​​across five dimensions of the user, which is the required five-dimensional personality vector.

4. The method for hiding a network structure based on personality information in a social network according to claim 3, wherein: In the centrality index of step S3, degree centrality indicates the number of direct connections between the current node and other nodes; node v i The standardized degree centrality d(v i ,A) is calculated as follows: Among them, A is the adjacency matrix of graph G, with dimension n*n; a ij is the element in the i-th row and j-th column of the adjacency matrix A; Betweenness centrality indicates the proportion of the current node appearing in the shortest path between other node pairs, which is used to measure the control ability of the node on the propagation of other nodes; node v i The standardized betweenness centrality b(v i ,A) is calculated as follows: Among them, l jk (i) is the slave node v j To node v k The shortest path passes through node v i The number of l jk Node v j To node v k The number of shortest paths; Close centrality indicates the closeness of the current node to all other nodes, which is the inverse of the mean of the node to all other nodes. i The closeness centrality c(v i ,A) is calculated as follows: Among them, d ij For node v i To node v j The shortest distance; Eigenvector centrality reflects the importance of a node; node v i The eigenvector centrality v(v i ,A) is calculated as follows: The local clustering coefficient reflects the degree of interconnection between neighboring nodes of a node. i The local clustering coefficient clu(v i ,A) is calculated as follows: Among them, V i For node v i The set of adjacent points of ; Based on the degree centrality, betweenness centrality, and local clustering coefficient of each node, a personality feature matrix is ​​constructed to characterize the agreeableness of all users. The formula is as follows: in, is the residual, which represents the difference between the user's actual agreeable personality value and the agreeable personality value obtained by multiple linear regression; α1, α2, α3, C agre is a constant, is an n*1 dimensional all-1 vector; Based on the degree centrality, closeness centrality, betweenness centrality, and local clustering coefficient of each node, a personality characteristic matrix is ​​constructed to characterize the sense of responsibility of all users. The formula is as follows: in, is the residual, which represents the difference between the user's real sense of responsibility personality value and the sense of responsibility personality value obtained by multiple linear regression; β1,β2,β3,β4,C cons is a constant; Based on degree centrality, closeness centrality, betweenness centrality and eigenvector centrality, a personality trait matrix is ​​constructed to characterize the extraversion of all users. The formula is as follows: in, is the residual, which represents the difference between the user's actual extraversion personality value and the extraversion personality value obtained by multiple linear regression; γ1,γ2,γ3,γ4,C extr is a constant; A personality trait matrix is ​​constructed based on degree centrality, closeness centrality, betweenness centrality, and eigenvector centrality to characterize the neuroticism of all users. The formula is as follows: in, is the residual, which represents the difference between the user's actual neurotic personality value and the neurotic personality value obtained by multiple linear regression; δ1, δ2, δ3, δ4, C neur is a constant; Based on degree centrality and betweenness centrality, a personality characteristic matrix is ​​constructed to characterize the openness of all users. The formula is as follows: in, is the residual, which represents the difference between the user's real open personality value and the open personality value obtained by multiple linear regression; θ1, θ2, C open is a constant.

5. The method for hiding network structure based on personality information in a social network according to claim 1, characterized in that: In step S44, each node v in the degree sequence is sequentially i degree d i Normalize and get the normalized degree sequence d_nor: d_nor={d_nor1,d_nor2,…,d_nor n }; The normalized calculation formula used is as follows:

6. The method for hiding network structure based on personality information in a social network according to claim 5, characterized in that: In step S44, when different human features are hidden, the change in the degree component is calculated as follows: (1) Change in Agreeableness when Hidden Δd_nor i The calculation formula is as follows: Among them, d_nor j Represents the values ​​in the normalized degree sequence before the mutation operation, and j is the sequence number of the element in the degree sequence; Represents the values ​​in the normalized degree sequence before the i-th mutation operation iteration, where i represents the number of iterations of the mutation operation; (2) Change in conscientious personality when it is hidden Δd_nor i The calculation formula is as follows: (3) Change in Extraversion when Hidden Δd_nor i The calculation formula is as follows: (4) Change in neurotic personality when hidden Δd_nor i The calculation formula is as follows: (5) Change in openness when hiding personality Δd_nor i The calculation formula is as follows:

7. The method for hiding network structure based on personality information in a social network according to claim 1, characterized in that: In step S5, the process of modifying the graph G to obtain a new graph G′ that satisfies the k-degree anonymous graph is as follows: S51: Get the deviation between the initial degree sequence d and the best degree sequence best_d corresponding to the graph G, node v i Degree deviation The calculation formula is as follows: S52: Determine the degree deviation of each node Does the value of If yes, then the value Add to the collection to be removed The node v i Put it into set V del ; otherwise the value Add to the collection to be added The node v i Put it into set V add ; S53: Get the neighbor nodes of each node in the graph G; where node v p The set of neighbor nodes As shown in the following formula: S54: Get the edge set E in the graph G, and based on the set to be removed Collections to be added Collection V del 、Collection V add Remove or create new edges between each point and its neighboring nodes to complete the modification of graph G.

8. The method for hiding network structure based on personality information in a social network according to claim 7, characterized in that: The method for adjusting the edges in the network structure in step S54 is as follows: S541: Initialize the edge set E′ so that E′=E; and obtain V del 、V add Elements in S542: From Any value in And from V del Find the corresponding node v p ; Then from v p The set of neighbor nodes Select node v q , remove node v p With node v q Links (v p ,v q ), (v p ,v q ) is removed from E′; The update formula of the removal process E′ is: E′=E′\{(v p ,v q )}; S543: From Any value in And from V add Find the corresponding node v r , construct node v q With node v r Links (v q ,v r ), (v q ,v r ) is added to E′; the update formula of the addition process E′ is: E′=E′∪{(v q ,v r )}; S544: After removing and building edges, node v p ,v q ,v r The degree of d p ′,d q ′,d r ′ indicates that the updating formula of each node degree is as follows: d p ′=d p -1 d q ′=d q +1-1=d q d r ′=d r +1; S545: After removing and building edges, node v p ,v q ,v r The degree deviation is respectively expressed as Indicates that the update formula of degree deviation is as follows: S546: Determine whether the following conditions are met: If so, v p From V del Remove, the processing formula is as follows: V del =V del \{v p }; Otherwise The value of is modified as follows: S547: Determine whether the following conditions are met: If so, v r From V add Remove, the processing formula is as follows: V add =V add \{v r }; Otherwise The value of is modified as follows: S548: Judgment Set Are both empty sets? If yes, output the adjusted node set V′ and the current e′ to form the required new graph G′ that satisfies the k-degree anonymous graph; otherwise, repeat the processing of steps S542-S547.

9. A user relationship graph preprocessing device based on personality information, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, it implements the steps of the method for hiding the network structure based on personality information in a social network as described in any one of claims 1 to 8, and then adjusts the network structure graph G representing the user relationship graph to obtain a new graph G′ that meets the k-degree anonymous graph requirements; and anonymizes key information related to user privacy contained in the original graph G.

Images