Network community hiding method based on user attribute driving and related device

Abstract

The invention belongs to the field of network communities, and discloses a network community hiding method based on user attribute driving and a related device, and the method comprises the steps: obtaining the edge addition selection probability of each community pair according to the separation degree and attribute similarity of each community pair and the number of nodes of each community in the community pair; obtaining an edge deletion selection probability of each community according to the cohesion, the attribute uniqueness and the node pair number of each community; randomly selecting the community pair as a target community pair according to the edge addition selection probability of each community pair, and selecting a non-edge with the minimum product of the sum of node degrees and (1-attribute similarity) in the target community pair to perform edge addition; randomly selecting the community as a target community according to the edge deletion selection probability of each community, and selecting the edge with the maximum product of the sum of node degrees and (1-attribute similarity) in the target community for edge deletion. According to the method, the limitation of a specific target community mode is broken through, the defect that node attribute information is generally neglected in an existing algorithm is effectively overcome, and the community hiding performance is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the field of network community, and relates to a network community hiding method based on user attribute driving and a related device. BACKGROUND

[0002] In actual graph network communities, node attributes contain rich user information, such as gender, age, and interest preferences. These attributes are of great value for accurately identifying and hiding communities. However, existing technologies mostly focus on disturbing community structures, ignoring the mining and utilization of node attribute information. This makes current methods unable to fully consider the personalized features of users when hiding communities, which may result in poor hiding effects. Once community detection algorithms utilize these ignored attribute information, users are likely to be detected, thereby causing privacy leakage risks and seriously damaging the information security of users. SUMMARY

[0003] The application aims to overcome the shortcomings of the prior art and provide a network community hiding method based on user attribute driving and a related device.

[0004] To achieve the above-mentioned purpose, the application adopts the following technical solutions:

[0005] In a first aspect, the application provides a network community hiding method based on user attribute driving, which includes: obtaining the graph structure, community information, and node attribute information of a network community; obtaining and according to the separation degree, attribute similarity, and node quantity of each community in a community pair, the edge addition selection probability of each community pair is obtained according to the graph structure, community information, and node attribute information of the network community; obtaining and according to the cohesion degree, attribute uniqueness, and node pair number of each community, the edge deletion selection probability of each community is obtained; randomly selecting a community pair as a target community pair according to the edge addition selection probability of each community pair, and adding an edge to the non-edge with the smallest product of the sum of node degrees and (1-attribute similarity) in the target community pair; randomly selecting a community as a target community according to the edge deletion selection probability of each community, and deleting an edge with the largest product of the sum of node degrees and (1-attribute similarity) in the target community.

[0006] Optionally, the obtaining and according to the separation degree, attribute similarity, and node quantity of each community in a community pair, the edge addition selection probability of each community pair includes: obtaining the attribute similarity of each community pair by the following formula: taking the cosine similarity between the average attribute vectors of the two communities in the community pair as the attribute similarity of the community pair.

[0007] Optionally, the obtaining and according to the separation degree of each community pair, the attribute similarity of each community pair and the node number of each community in the community pair, obtaining the edge adding selection probability of each community pair comprises: obtaining the edge adding selection probability of each community pair by the following formula:

[0008] P J(ij) = |X i ||X j | × S(X i , X j ) × CS ij

[0009] Wherein, P J(ij) is the edge adding selection probability of the community pair (X i , X j ); X i is the community numbered i in the network community; X j is the community numbered j in the network community; |X i | is the node number of X i ; |X j | is the node number of X j ; S(X i , X j ) is the separation degree of the community pair (X i , X j ), and CS ij is the attribute similarity of the community pair (X i , X j ).

[0010] Optionally, the obtaining and according to the cohesion degree of each community, the attribute uniqueness of each community and the node pair number, obtaining the edge deleting selection probability of each community comprises: obtaining the attribute uniqueness of each community by the following manner: taking the cosine similarity between the average attribute vector of all communities in the network community and the average attribute vector of the current community as the attribute uniqueness of the current community.

[0011] Optionally, the obtaining and according to the cohesion degree of each community, the attribute uniqueness of each community and the node pair number, obtaining the edge deleting selection probability of each community comprises: obtaining the edge deleting selection probability of each community by the following formula:

[0012] P S(i) = C(X i ) × CS i × |X i ||X i -1|

[0013] Wherein, P S(i) is the edge deleting selection probability of X i ; X i is the community numbered i in the network community; C(X i ) is the cohesion degree of Xi cohesion of the community; CS i |X i ||X i -1| is twice the number of node pairs of the community.

[0014] Optionally, when there are at least two non-edges with the minimum product of the sum of node degrees and (1-attribute similarity) in the target community pair, the non-edge that increases the external degree score of the network community the least is selected for edge addition.

[0015] Optionally, when there are at least two edges with the maximum product of the sum of node degrees and (1-attribute similarity) in the target community, the edge that increases the external degree score of the network community the most is selected for edge deletion.

[0016] In a second aspect, the present application provides a network community hiding system based on user attribute driving, comprising: a data acquisition module, configured to acquire a graph structure, community information and node attribute information of a network community; a budget allocation module, configured to, according to the graph structure, community information and node attribute information of the network community, acquire, for each community pair in the network community, an edge addition selection probability according to the separation degree, attribute similarity and number of nodes of each community in the community pair; acquire, for each community in the network community, an edge deletion selection probability according to the cohesion, attribute uniqueness and number of node pairs of each community; an edge addition module, configured to randomly select a community pair as a target community pair according to the edge addition selection probability of each community pair, and select a non-edge with the minimum product of the sum of node degrees and (1-attribute similarity) in the target community pair for edge addition; and an edge deletion module, configured to randomly select a community as a target community according to the edge deletion selection probability of each community, and select an edge with the maximum product of the sum of node degrees and (1-attribute similarity) in the target community for edge deletion.

[0017] In a third aspect, the present application provides a computer device, comprising a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the network community hiding method based on user attribute driving.

[0018] In a fourth aspect, the present application provides a computer readable storage medium, wherein the computer readable storage medium stores a computer program, and the computer program is executable on a processor to implement the steps of the network community hiding method based on user attribute driving.

[0019] Compared with the prior art, the present application has the following beneficial effects:

[0020] The present application is based on a user attribute driven network community hiding method, which innovatively deeply integrates node attributes and community structure information, determines the edge addition selection probability of each community pair by means of the separation degree, attribute similarity and the number of nodes in each community of the community pair, and determines the edge deletion selection probability of each community by means of the cohesion degree, attribute uniqueness and the number of node pairs of each community. And in the edge disturbance, by introducing edges between high similarity node pairs, the nodes can be more easily transferred to other communities; at the same time, the edges inside the community are deleted, further strengthening the separation between nodes to make the nodes more easily leave the original community. After these processes, it helps to significantly improve the hiding effect of the community under smaller disturbance budget, efficiently avoids the detection of the community detection algorithm, and ensures the safety of the attribute information of the user. Unlike traditional methods, the present application method breaks through the limitation of specific target community mode and can flexibly cope with more extensive community hiding needs, greatly improving the flexibility of hiding any target community. By fully mining the structure information and node attribute information in the community, the drawbacks of the existing algorithms that generally ignore node attribute information are effectively overcome, so that the complex community relationship can be better handled, the community hiding performance is significantly improved, and the user can be effectively prevented from being identified by the detection algorithm, greatly reducing the risk of user privacy leakage, and effectively protecting the privacy security of the network community user. BRIEF DESCRIPTION OF DRAWINGS

[0021] Figure 1 The present application is based on a user attribute driven network community hiding method, which innovatively deeply integrates node attributes and community structure information, determines the edge addition selection probability of each community pair by means of the separation degree, attribute similarity and the number of nodes in each community of the community pair, and determines the edge deletion selection probability of each community by means of the cohesion degree, attribute uniqueness and the number of node pairs of each community. And in the edge disturbance, by introducing edges between high similarity node pairs, the nodes can be more easily transferred to other communities; at the same time, the edges inside the community are deleted, further strengthening the separation between nodes to make the nodes more easily leave the original community. After these processes, it helps to significantly improve the hiding effect of the community under smaller disturbance budget, efficiently avoids the detection of the community detection algorithm, and ensures the safety of the attribute information of the user. Unlike traditional methods, the present application method breaks through the limitation of specific target community mode and can flexibly cope with more extensive community hiding needs, greatly improving the flexibility of hiding any target community. By fully mining the structure information and node attribute information in the community, the drawbacks of the existing algorithms that generally ignore node attribute information are effectively overcome, so that the complex community relationship can be better handled, the community hiding performance is significantly improved, and the user can be effectively prevented from being identified by the detection algorithm, greatly reducing the risk of user privacy leakage, and effectively protecting the privacy security of the network community user.

[0022] Figure 2 The present application is based on a user attribute driven network community hiding method, which innovatively deeply integrates node attributes and community structure information, determines the edge addition selection probability of each community pair by means of the separation degree, attribute similarity and the number of nodes in each community of the community pair, and determines the edge deletion selection probability of each community by means of the cohesion degree, attribute uniqueness and the number of node pairs of each community. And in the edge disturbance, by introducing edges between high similarity node pairs, the nodes can be more easily transferred to other communities; at the same time, the edges inside the community are deleted, further strengthening the separation between nodes to make the nodes more easily leave the original community. After these processes, it helps to significantly improve the hiding effect of the community under smaller disturbance budget, efficiently avoids the detection of the community detection algorithm, and ensures the safety of the attribute information of the user. Unlike traditional methods, the present application method breaks through the limitation of specific target community mode and can flexibly cope with more extensive community hiding needs, greatly improving the flexibility of hiding any target community. By fully mining the structure information and node attribute information in the community, the drawbacks of the existing algorithms that generally ignore node attribute information are effectively overcome, so that the complex community relationship can be better handled, the community hiding performance is significantly improved, and the user can be effectively prevented from being identified by the detection algorithm, greatly reducing the risk of user privacy leakage, and effectively protecting the privacy security of the network community user. DETAILED DESCRIPTION

[0023] In order to make the person skilled in the art better understand the present application scheme, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by the person skilled in the art without creative labor should belong to the protection scope of the present application.

[0024] It should be noted that the terms "first", "second", etc. in the description and claims of the present invention and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the numbers used in this way can be interchanged where appropriate, so that the embodiments of the present invention described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0025] The present invention is described in further detail below with reference to the accompanying drawings:

[0026] See also Figure 1 In one embodiment of the present invention, a network community hiding method based on user attribute driving is provided, which innovatively integrates node attributes with community structure information to achieve efficient community hiding.

[0027] Specifically, the network community hiding method based on user attribute drive of the present invention includes the following steps:

[0028] S1: Obtain the graph structure, community information and node attribute information of the network community;

[0029] S2: Based on the graph structure, community information, and node attribute information of the network community, for each community pair within the network community, obtain and calculate the edge addition selection probability of each community pair based on the separation degree, attribute similarity, and the number of nodes in each community within the community pair; for each community within the network community, obtain and calculate the edge deletion selection probability of each community based on the cohesion degree, attribute uniqueness, and number of node pairs of each community;

[0030] S3: Randomly select a community pair as the target community pair based on the edge addition selection probability of each community pair, and select the non-edge with the smallest product of the sum of node degrees and (1-attribute similarity) in the target community pair for edge addition;

[0031] S4: Randomly select a community as the target community according to the edge deletion selection probability of each community, and select the edge with the largest product of the sum of node degrees and (1-attribute similarity) in the target community for edge deletion.

[0032] The application is based on a user attribute driven network community hiding method, which innovatively deeply integrates node attributes and community structure information, determines the edge addition selection probability of each community pair by means of the separation degree, attribute similarity and the number of nodes in each community in the community pair, and determines the edge deletion selection probability of each community by means of the cohesion degree, attribute uniqueness and the number of node pairs of each community. And in the edge disturbance, by introducing edges between high similarity node pairs, the nodes can be more easily transferred to other communities; at the same time, the edges inside the community are deleted, further strengthening the separation between nodes to make the nodes more easily leave the original community. After these treatments, it is helpful to significantly improve the hiding effect of the community under a smaller disturbance budget, efficiently avoid the detection of the community detection algorithm, and ensure the safety of the attribute information of the user.

[0033] Explanatorily, in the current network community hiding method, taking the modularization-based method as an example, when adding edges, it always preferentially selects to add edges between the two communities that can increase the total volume of the community the most. This strategy leads to uneven allocation of disturbance resources, some target communities are paid too much attention, while other communities are easily ignored, increasing the risk of exposure. In addition, when selecting edges to add or delete, the existing method only uses a simple random strategy, lacking an effective differentiation mechanism. This makes some users more likely to be exposed to detection algorithms, further increasing the possibility of privacy leakage, and cannot guarantee the privacy security of users.

[0034] In addition, the modularization method has limitations in measuring community structure and sometimes cannot accurately reflect the true situation of the community. In some network structures that are not obvious, the modularization value may be overestimated, resulting in poor effect of the deception strategy based on modularization. At the same time, the performance and efficiency of the existing method in dealing with complex graph networks and diversified community structures are difficult to achieve the ideal state. They often have difficulty in finding a good balance point between hiding effect and computational cost, either paying a high computational cost for pursuing high hiding effect, or resulting in unsatisfactory hiding effect due to computational resource limitations, which cannot stably protect user privacy and effectively deal with the excessive mining of user privacy by community detection algorithms.

[0035] The method of the application provides a new and more efficient community hiding strategy. Unlike traditional methods, the method breaks through the limitations of specific target community patterns and can flexibly cope with more extensive community hiding needs, greatly improving the flexibility of hiding any target community. By fully exploiting the structural information and node attribute information in the community, the drawbacks of existing algorithms that generally ignore node attribute information are effectively overcome, so that complex community relationships can be better handled, community hiding performance can be significantly improved, and users can be effectively prevented from being identified by detection algorithms, greatly reducing the risk of user privacy leakage and ensuring the privacy security of network community users.

[0036] Explanatorily, when the edges are added and deleted, based on the preset edge adding quantity setting value and edge deleting quantity setting value, when the edge adding quantity and the edge deleting quantity reach the edge adding quantity setting value and the edge deleting quantity setting value, the network community hiding driven by the user attribute is completed.

[0037] In a possible implementation, the graph structure of the network community, the community information and the node attribute information are embodied in a network G=(V, E) and a partition P={X1, X2,..., Xn} of the network G=(V, E), where V represents a node set, E represents an edge set in the network, and the edges connect the nodes to form the basic structure of the network community. L} are set, which lays a foundation for subsequent operations and ensures clear cognition of the environment of the entire network community.

[0038] In a possible implementation, the edge adding selection probability of each community pair is obtained according to the separation degree, the attribute similarity and the node quantity of each community in the community pair, and the edge adding selection probability of each community pair is obtained by the following formula:

[0039] Explanatorily, the value range of the cosine similarity is [-1, 1], and the greater the value is, the higher the similarity is.

[0040] Optionally, the edge adding selection probability of each community pair is obtained according to the separation degree, the attribute similarity and the node quantity of each community in the community pair, and the edge adding selection probability of each community pair is obtained by the following formula:

[0041] P J(ij) =|X i ||X j |×S(X i ,X j )×CS ij

[0042] wherein P is the edge adding selection probability of the community pair (X, X) ; X is the community numbered i in the network community; X is the community numbered j in the network community; |X| is the node quantity of X; |X| is the node quantity of X; S(X, X) is the separation degree of the community pair (X, X), and CS is the attribute similarity of the community pair (X, X). J(ij) i j i j i i j j i j i j ij ​​​​​​​​​​​​​For the community (X i , X j ) attribute similarity.

[0043] Explanation: The probability of adding edges to each community pair is essentially a reasonable allocation of the edge adding budget. Adding cross-community edges will increase the possibility of merging two communities. In order to allocate the edge adding budget fairly and effectively, the possibility of merging two communities after adding new edges and the benefits after the merger need to be considered. The separation degree is S(X i , X j ) measure the inter-community (X i , X j ) The sparseness of the connection. The smaller the separation, the sparser the connection, and the more difficult it is to merge. Therefore, it can be used as a measure of the merging X i and X j An indicator of difficulty. If two communities merge, X i Each node in will get |X j | new friends in the same community. In addition, the attribute similarity between the two communities will affect the formation of new connections after the merger.

[0044] In this implementation, the cosine similarity between the average attribute vectors of two communities within a community pair is used as the attribute similarity of the community pair. The higher the attribute similarity, the easier it is to form a connection after merging. i ||X j |CS ij Reflects the benefits of the merger. Comprehensive costs and benefits, each round randomly selects a community pair (X i , X j ), whose probability is the same as |X i ||X j |×S(X i , X j )×CS ij In this way, the edge addition budget can be more reasonably allocated to each community pair, avoiding the poor hiding effect of some communities due to uneven resource allocation, thereby reducing the risk of user privacy leakage.

[0045] Based on the above design, when performing the edge adding operation, the possibility of community merging (measured by the separation degree) and the benefits brought by the merging (factors such as attribute similarity) can be comprehensively considered, so as to realize reasonable allocation of the budget and avoid excessive concentration of the disturbance resources on a few communities. The core purpose of adding edges is to facilitate the target community to be more easily merged with other communities. The above method accurately measures the possibility of merging of the two communities after adding edges, and clearly defines the benefits brought by the merging, such as the number of new neighbors of the community members after the merging, and the higher the attribute similarity of the two communities, the lower the difficulty of forming new connections after the merging. In this way, the problem of uneven allocation of disturbance resources in the prior art is effectively solved, and it is ensured that each target community can obtain reasonable disturbance resources, thereby comprehensively improving the overall hiding effect and avoiding the situation that individual network community users cause privacy leakage due to unreasonable resource allocation.

[0046] In a possible implementation, the obtaining and determining of the edge deletion selection probability of each community according to the cohesion degree, attribute uniqueness and node pair number of each community comprises: obtaining the attribute uniqueness of each community by taking the cosine similarity between the average attribute vector of all communities in the network community and the average attribute vector of the current community as the attribute uniqueness of the current community.

[0047] Optionally, the obtaining and determining of the edge deletion selection probability of each community according to the cohesion degree, attribute uniqueness and node pair number of each community comprises: obtaining the edge deletion selection probability of each community by the following formula:

[0048] P S(i) =C(X i )×CS i ×|X i ||X i -1|

[0049] Wherein, P S(i) is the edge deletion selection probability of X i ; X i is the community numbered i in the network community; C(X i ) is the cohesion degree of X i ; CS i is the attribute uniqueness of the community; |X i ||X i -1| is twice the node pair number of the community.

[0050] Explanatorily, the edge deletion selection probability of each community is essentially a reasonable allocation of the edge deletion budget, and deleting edges in the community will increase the possibility of splitting the community. The cohesion degree C i measures the difficulty of splitting, and C iThe higher, the more obvious the community structure, the more disturbance budget is needed to effectively hide the structure. In terms of revenue evaluation, it is considered that if community X i After splitting, its nodes can better hide in other communities, even if they have obtained revenue.

[0051] Specifically, the cosine similarity between the average attribute vector of all communities in the network community and the average attribute vector of the current community is taken as the attribute uniqueness of the current community. The potential number of node pairs in the community is combined to quantify the revenue. The higher the revenue value, the weaker the structural uniqueness of community X i , and its nodes are more likely to hide in other communities. Therefore, a community X i is randomly selected each round, and the probability is proportional to C(X i ) x CS i x |X i ||X i -1|. This edge deletion budget allocation strategy can accurately disturb the community internally, making the community more likely to split, thereby disrupting the original community structure and reducing the likelihood of user privacy information being identified by the community detection algorithm.

[0052] Explanatorily, when selecting a community to delete an edge, the difficulty of splitting (measured by cohesion) and potential revenue (involving attribute uniqueness and number of node pairs) are considered. Through comprehensive analysis of these factors, a community is reasonably selected to delete an edge to achieve the goal of easier community splitting. In this process, the cost and benefit of community edge deletion are clarified. The cost reflects the stability of the community structure itself, which measures the difficulty of community splitting caused by edge deletion; the benefit lies in whether the internal nodes can better hide in other communities after the community splits. The "Matthew effect" that may occur in the budget allocation process is effectively solved, avoiding resource allocation imbalance. By reasonably selecting a community to delete, the stability of the community is reduced while the possibility of node dispersion hiding is increased, thereby significantly reducing the probability of users being detected by the algorithm and further reducing the risk of user privacy leakage, providing more reliable protection for user privacy.

[0053] In a possible implementation, when the sum of node degrees of the target community pair and the product of (1-attribute similarity) of the non-edge with the smallest sum of node degrees of the target community pair and the product of (1-attribute similarity) is at least two: select the non-edge that increases the external degree score of the network community the least when adding an edge.

[0054] Explanatorily, when adding an edge, an edge addition strategy based on node degree and node attribute similarity is adopted. Let the eigenvalues of the Laplacian matrix L of the network graph G be λ1≤λ2≤...λn From the perspective of graph cut, to make the community structure of subgraph G = (X U Y, E) composed of two communities X and Y more difficult to detect, its algebraic connectivity should be increased, i.e. maximizing where, is the perturbed graph after adding edges. However, maximizing the algebraic connectivity λ2 is an NP-hard problem. According to the perturbation theory, if λ2 is isolated, for any non-edge (v2(i)-v2(j)) 2 The first-order approximation is given, where v2 is the Fiedler vector of graph G. However, the method based on the perturbation theory is limited in practical scenarios. Therefore, by defining F(x) = (x-(1 / n)) 2 , the problem of maximizing λ2 is transformed into:

[0055]

[0056] where d k is the degree of the kth node, and m is the number of original edges of graph G. is the edge added between node pair (i, j), and F is the defined function F(x) := (x-(1 / n)) 2 , is the set of non-edges, and (X X Y) is between communities X and Y.

[0057] Further derivation shows that this problem is equivalent to:

[0058]

[0059] where d i and d j represent the degrees of nodes i and j, respectively, is the set of non-edges.

[0060] After that, the node attribute information is integrated into the edge adding process. Specifically, when constructing the edge set, the cosine similarity is used to accurately quantify the similarity between node attribute vectors. The higher the cosine similarity, the more likely the corresponding node pair will be hidden in the same community after adding the edge. Therefore, the non-edge with the minimum product of the sum of the degrees of the nodes in the target community pair and (1-attribute similarity) is added to the edge. This is because adding edges between nodes with small attribute differences is more likely to destroy the original community structure, thereby enhancing the community hiding effect and effectively preventing user privacy from being obtained by detection algorithms.

[0061] Explanation: After selecting community pairs, the method considers both node connectivity (sum of node degrees) and attribute similarity when selecting edges to add. Edges are preferentially added to pairs with low connectivity and high attribute similarity. This is because adding edges between such pairs is more likely to disrupt the existing community structure, blurring the boundaries between communities and thus achieving the goal of hiding communities.

[0062] When there are at least two non-edges with the smallest product of the sum of node degrees and (1-attribute similarity) in the target community, the non-edge that minimizes the increase in the external degree score (ODF) of the network community is selected. The ODF of node u is defined as: Among them, u and v represent nodes, S is a subset of nodes, and d u is the degree of node u.

[0063] In a possible implementation, when selecting the edge with the largest product of the sum of the node degrees and (1-attribute similarity) in the target community for edge deletion, when there are at least two edges with the largest product of the sum of the node degrees and (1-attribute similarity) in the target community: the edge with the largest product of the sum of the node degrees and (1-attribute similarity) in the target community that increases the external degree score of the network community the most is selected for edge deletion.

[0064] Explanation: When deleting edges, we use an edge deletion strategy based on edge perturbation based on node degree and node attribute similarity. For edge deletion within a single community X, the goal is to reduce the connectivity within the community. Consider the complement graph of G Deleting an edge in G is equivalent to Add edges to The nodes in G are more closely connected, making community X in G more likely to split. Therefore, maximizing Algebraic connectivity Equivalent to minimizing the maximum eigenvalue of G in, is the perturbation graph after deleting the edge. It can also be approximately transformed into:

[0065]

[0066] Among them, E(X) is the edge set in the X community, and G is the original network structure To delete the edge between (i, j), m is the number of edges in G, λ k is the kth eigenvalue of the Laplace matrix, and F is the defined function F(x):=(x-(1 / n)) 2 .

[0067] Further equivalent to:

[0068] arg max (i,j)∈E(X) (d i +dj )

[0069] wherein E(X) is the edge set in the X community, d i +d j is the sum of the degrees of the node pair (i, j).

[0070] That is, the edge with the largest increase in the ODF is selected from the edges with the largest product of the sum of the degrees and (1-attribute similarity) in the target community, because deleting such an edge will significantly affect the connectivity of the community, greatly increasing the probability of community splitting, thus achieving the expected community hiding target and effectively preventing the leakage of user privacy information.

[0071] wherein when there are at least two edges with the largest product of the sum of the degrees and (1-attribute similarity) in the target community, the edge with the largest increase in the ODF is selected from the edges with the largest product of the sum of the degrees and (1-attribute similarity) in the target community. The edge with the largest increase in the ODF is defined as ODF(u) is the out-degree score of node u, ODF(v) is the out-degree score of node v, d u , d v are the degrees of nodes u and v.

[0072] Explanatorily, in the process of converting the spectral optimization problem into an edge volume optimization problem, the node attribute information is fully utilized, and the edge is selected in combination with the node degree information. In edge addition, the non-edge with the smallest sum of degrees and the highest attribute similarity is selected for addition; in edge deletion, the edge with the largest sum of degrees and the lowest attribute similarity is selected for deletion. Illustratively, heuristic algorithms can also be used to further improve the operation efficiency, such as by sorting to reduce the search space and dynamically updating the data to optimize the algorithm execution process.

[0073] Overall, the edge addition and edge deletion strategies of the present application: on the one hand, the edge selection strategy based on intuitive statistical information such as node degree and node attribute information can more accurately perturb the community structure, making the community structure present stronger fuzziness in front of the detection algorithm, effectively improving the deception effect, greatly reducing the risk of being discovered by the user by the detection algorithm, and thus reducing the possibility of privacy leakage. On the other hand, the efficient algorithm design significantly reduces the computational cost and greatly improves the execution efficiency of the algorithm in the network. For example, when processing social network data, the target community can be quickly hidden while maintaining a low consumption of computing resources. This successfully overcomes the problem of insufficient performance and efficiency of existing methods in network applications, and provides strong support for user privacy protection in large-scale network environments.

[0074] The following is an apparatus embodiment of the present application, which can be used to execute the method embodiment of the present application. For details not disclosed in the apparatus embodiment, please refer to the method embodiment of the present application.

[0075] Referring to Figure 2 In another embodiment of the present application, a network community hiding system based on user attribute driving is provided, which can be used to implement the network community hiding method based on user attribute driving described above. Specifically, the network community hiding system based on user attribute driving comprises a data acquisition module, a budget allocation module, an edge adding module and an edge deleting module.

[0076] The data acquisition module is configured to acquire the graph structure, community information and node attribute information of the network community; the budget allocation module is configured to, for each community pair in the network community, acquire and obtain, according to the separation degree, attribute similarity and node quantity of each community in the community pair, an edge adding selection probability of the community pair; for each community in the network community, acquire and obtain, according to the cohesion degree, attribute uniqueness and node pair number of the community, an edge deleting selection probability of the community; the edge adding module is configured to randomly select a community pair as a target community pair according to the edge adding selection probability of each community pair, and add an edge to a non-edge with the smallest product of the sum of node degrees and (1-attribute similarity) in the target community pair; and the edge deleting module is configured to randomly select a community as a target community according to the edge deleting selection probability of each community, and delete an edge with the largest product of the sum of node degrees and (1-attribute similarity) in the target community.

[0077] The foregoing embodiment of the network community hiding method based on user attribute driving involves all related contents of each step, which can be cited to the function description of the function module corresponding to the network community hiding system based on user attribute driving in the embodiment of the present application, and will not be described here.

[0078] The division of the modules in the embodiment of the present application is illustrative, and is only a logical function division. In actual implementation, another division mode can be used. In addition, each function module in each embodiment of the present application can be integrated in one processor, or can be physically separated, or two or more modules can be integrated in one module. The integrated module can be realized in the form of hardware or in the form of a software function module.

[0079] In still another embodiment of the present application, a computer device is provided, which comprises a processor and a memory, the memory being configured to store a computer program, the computer program comprising program instructions, and the processor being configured to execute the program instructions stored in the computer storage medium. The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions in the computer storage medium to implement a corresponding method flow or a corresponding function; the processor in the embodiments of the present application can be used for the operation of the network community hiding method driven by user attributes.

[0080] In still another embodiment of the present application, the present application further provides a storage medium, specifically a computer readable storage medium (Memory), which is a memory device in a computer device, and is configured to store programs and data. It can be understood that the computer readable storage medium herein can include an internal storage medium in the computer device, and of course can also include an extended storage medium supported by the computer device. The computer readable storage medium provides a storage space, and the storage space stores an operating system of the terminal. Furthermore, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and these instructions can be one or more computer programs (including program codes). It should be noted that the computer readable storage medium herein can be a high-speed RAM memory, or a non-volatile memory such as at least one disk memory. One or more instructions stored in the computer readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the network community hiding method driven by user attributes in the above embodiments.

[0081] Those skilled in the art will appreciate that embodiments of the application can be devised for a method, a system, or a computer program product. Accordingly, the present application can be embodied in the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROMs, optical storage devices, etc.) embodying computer readable program code.

[0082] The present application is described in reference to the flowchart illustrations and / or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0083] These computer program instructions can also be stored in a computer- readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instructions which implement the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0084] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams. Figure 1 one or more functions specified in the flowchart illustrations and / or block diagrams.

[0085] Finally, it should be noted that the above-mentioned embodiments are merely intended for describing and illustrating, not limiting the technical solutions of the present application. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that the specific embodiments of the present application can be modified or replaced by equivalents without departing from the spirit and scope of the present application, and any modifications or equivalent replacements without departing from the spirit and scope of the present application should be covered in the protection scope of the claims of the present application.

Claims

1. A network community hiding method based on user attribute drive, characterized in that: include: Obtain the graph structure, community information and node attribute information of the network community; According to the graph structure, community information and node attribute information of the network community, for each community pair in the network community, the edge addition selection probability of each community pair is obtained based on the separation degree, attribute similarity and the number of nodes in each community in the community pair; For each community in the network community, obtain the edge deletion selection probability of each community based on the cohesion, attribute uniqueness and number of node pairs of each community; According to the edge addition selection probability of each community pair, a community pair is randomly selected as the target community pair, and the non-edge with the smallest product of the sum of node degrees and (1-attribute similarity) in the target community pair is selected for edge addition; A community is randomly selected as the target community according to the edge deletion selection probability of each community, and the edge with the largest product of the sum of node degrees and (1-attribute similarity) in the target community is selected for edge deletion.

2. The method for hiding a network community based on user attribute drive according to claim 1, characterized in that: The step of obtaining the edge addition selection probability of each community pair based on the separation degree, attribute similarity, and number of nodes in each community within the community pair includes: The attribute similarity of each community pair is obtained by the following formula: the cosine similarity between the average attribute vectors of the two communities within the community pair is taken as the attribute similarity of the community pair.

3. The method for hiding a network community based on user attribute drive according to claim 1, characterized in that: The step of obtaining the edge addition selection probability of each community pair based on the separation degree, attribute similarity, and number of nodes in each community within the community pair includes: The edge addition selection probability of each community pair is obtained by the following formula: P J(ij) =|X i ||X j |×S(X i ,X j )×CS ij Among them, P J(ij) For the community (X i , X j )’s edge adds selection probability; X i is the community numbered i in the network community; X j is the community numbered j in the network community; |X i | for X i Number of nodes; |X j | for X j The number of nodes; S(X i , X j ) is the community's i , X j ) separation, CS ij For the community (X i , X j ) attribute similarity.

4. The method for hiding a network community based on user attribute drive according to claim 1, characterized in that: The step of obtaining the edge deletion selection probability of each community based on the cohesion, attribute uniqueness, and number of node pairs of each community includes: The attribute uniqueness of each community is obtained in the following way: the cosine similarity between the average attribute vectors of all communities in the network community and the average attribute vector of the current community is used as the attribute uniqueness of the current community.

5. The method for hiding a network community based on user attribute drive according to claim 1, characterized in that: The step of obtaining the edge deletion selection probability of each community based on the cohesion, attribute uniqueness, and number of node pairs of each community includes: The edge deletion selection probability of each community is obtained by the following formula: P S(i) =C(X i )×CS i ×|X i ||X i -1| Among them, P S(i) For X i The edge deletion selection probability of X i is the community numbered i in the network community; C(X i ) is X i Cohesion; CS i is the uniqueness of community attributes; |X i ||X i -1| is twice the number of node pairs in the community.

6. The method for hiding a network community based on user attribute drive according to claim 1, characterized in that: When selecting the non-edge with the smallest product of the sum of node degrees and (1-attribute similarity) in the target community pair for edge addition, if there are at least two non-edges with the smallest product of the sum of node degrees and (1-attribute similarity) in the target community pair: The target community is selected to add the non-edge that has the smallest product of the sum of node degrees and (1-attribute similarity) among the non-edges that can minimize the increase in the external degree score of the network community.

7. The method for hiding a network community based on user attribute drive according to claim 1, characterized in that: When selecting the edge with the largest product of the sum of node degrees and (1-attribute similarity) in the target community for edge deletion, if there are at least two edges with the largest product of the sum of node degrees and (1-attribute similarity) in the target community: The edge with the largest product of the sum of node degrees and (1-attribute similarity) in the target community that increases the external degree score of the network community the most is selected for edge deletion.

8. A network community hiding system based on user attribute drive, characterized by: include: Data acquisition module, used to obtain the graph structure, community information and node attribute information of the network community; The budget allocation module is used to obtain the edge addition selection probability of each community pair in the network community based on the graph structure, community information and node attribute information of the network community and the separation degree, attribute similarity and number of nodes in each community in the community pair; For each community in the network community, obtain the edge deletion selection probability of each community based on the cohesion, attribute uniqueness and number of node pairs of each community; An edge adding module is used to randomly select a community pair as a target community pair based on the edge adding selection probability of each community pair, and select the non-edge with the smallest product of the sum of node degrees and (1-attribute similarity) in the target community pair for edge adding; The edge deletion module is used to randomly select a community as the target community according to the edge deletion selection probability of each community, and select the edge with the largest product of the sum of node degrees and (1-attribute similarity) in the target community for edge deletion.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the steps of the network community hiding method based on user attribute driving are implemented as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the network community hiding method based on user attribute driving are implemented as claimed in any one of claims 1 to 7.

Images