CN110175634A - A kind of network privacy guard method based on disturbance subgraph - Google Patents
A kind of network privacy guard method based on disturbance subgraph Download PDFInfo
- Publication number
- CN110175634A CN110175634A CN201910371124.3A CN201910371124A CN110175634A CN 110175634 A CN110175634 A CN 110175634A CN 201910371124 A CN201910371124 A CN 201910371124A CN 110175634 A CN110175634 A CN 110175634A
- Authority
- CN
- China
- Prior art keywords
- subgraph
- disturbance
- population
- node
- network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/604—Tools and structures for managing or administering access control systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
- G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
- G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
Abstract
A kind of network privacy guard method based on disturbance subgraph; network is handled by certain mode to obtain multiple populations on the basis of spectrum maximizes algorithm; operation for population, the present invention include two aspects: first aspect is to be operated to obtain the disturbance sub-graph data library of respective objects to population；The second aspect is to be operated to obtain the disturbance sub-graph data library of respective objects to different population.The present invention can generate the disturbance sub-graph data library of respective objects according to corresponding network, realize the protection of network social intercourse privacy.
Description
Technical field
The present invention relates to filed of network information security, data sharing will bring huge income, however, the individual in data is hidden
Private leakage and information security will face a severe challenge, by study based on disturbance subgraph network privacy guard method come probe into as
What preferably realizes the secret protection of network.
Background technique
With infiltration of the computer technology in social networks, a networked society is gradually opened, and individual privacy faces a danger.
How control is carried out to the network information security under big data era, protection individual privacy becomes current research hot spot.Firstly,
In our daily life, complex network is ubiquitous, and the boundless universe that we are lived can regard a complexity as in fact
Network, our daily life also be unable to do without network.We accidentally can leave bulk information on many networks daily,
Such as online social portal network, Sina weibo network, Zhi Hu forum, school net, Facebook etc., and in these networks
In, our information can carry out community by certain algorithms in the form of node and side and divide classification.Secondly, gradually with network
Transparence, our information cannot get effective protection, and substantially any people can obtain these information by certain means, at certain
In a little situations, when user wants to protect the privacy of oneself, that is, the information for being not desired to oneself is easily sorted out, and how effective studies
The privacy information of protection user is just particularly important.
Summary of the invention
Secret protection safety in order to overcome the shortcomings of existing network is poor, and the present invention relates to Web Community's discovery techniques
Field, in particular to a kind of network privacy guard method based on disturbance subgraph.
For achieving the above object, the present invention the following technical schemes are provided:
A kind of network privacy guard method based on disturbance subgraph, the described method comprises the following steps:
S1: for a random network, the division that algorithm carries out community is maximized using spectrum, obtains multiple samples primary；
S2: one sample primary of random selection is handled using population dividing method, obtains the kind of multiple disturbance subgraphs
Group；
S3: being based on evolutionary computation, and the population for disturbing subgraph to one is handled, and generates a new disturbance subgraph kind
Group；
S4: handling the population of new disturbance subgraph, obtains disturbing subgraph library in population；
S5: being based on evolutionary computation, handles the populations of disturbance subgraphs different under sample primary, generate one it is new
Subgraph population is disturbed, the population of new disturbance subgraph is handled, obtains disturbing subgraph library between population.
Further, in the step S2, population dividing method includes following procedure:
Firstly, maximizing clustering algorithm using spectrum obtains K samples primary, C1、C2、C3......CK, random choosing wherein
Take a sample C primaryi, n node is randomly selected from the sample primary；
Then, some node d in n node is randomly selected, obtains selected node d in sample C primaryiUnder it is all
Neighbor node and non-neighbor node form the neighbor node set Φ of node d1With non-neighbours' node set Φ2；
Next, random erasure d and belonging to Φ1Node company side, if it is not, the wheel is skipped, while random increasing
Add a d and belongs to Φ2Node company side, update the neighborhood Φ of d1With non-neighborhood Φ2；It repeats the process T times；
Later, sample C primary is traversediMiddle n node, obtains disturbance subgraph H, and this n node is given and new includes
Nodal scheme and even the label L ' of frontier juncture system1, L '2, L '3......L′n；
Then, C is chosen againiN new node, and repetition n times operate to obtain and belong to sample C primaryiN number of disturbance
Subgraph, and N number of disturbance subgraph is classified as population Gi；
Finally, repeating all operations obtains disturbance subgraph population G corresponding to K samples primary1、G2、G3......GK。
Further, in the step S3, the treatment process for disturbing the population of subgraph is as follows:
S3.1, from population GiIn random successively obtain two disturbance subgraph N1, N2, then will disturb subgraph N1, N2It is each random
The part for being partitioned into half swaps mating, generates two new disturbance subgraph N3, N4, then from disturbance subgraph N3, N4In with
Machine is chosen one and is handled；
S3.2 repeats above step, until traversing entire population Gi。
Further, in the step S4, the process handled the population of new disturbance subgraph is as follows:
The new disturbance subgraph for randomly selecting out is carried out exclusive or addition with former network, obtains a new net by S4.1
Network；
S4.2 is divided using identical community's partitioning algorithm, by the original before obtained new network structure and attack
Network structure is compared, and by comprehensively considering a series of indexs such as node degree, average shortest path length, observation is based on disturbance subgraph
Attack strategies method whether achieved the effect that concealed nodes, will disturbance subgraph carry out label, be put into database and planted
Subgraph library is disturbed between group.
Preferably, the exclusive or in the S4.1 is added, and is made as given a definition.
Wherein, C (i, j) represents the company side of former network node i, j, and R (i, j) is the node i for disturbing subgraph, the company between j
Side, C ' (i, j) represent the company side of network node i, j after attack.
It is as follows that subgraph library generating process is disturbed in the step S5, between population:
Firstly, from different population Gi, GjIn one disturbance subgraph N of each random acquisition1, N2；
Secondly, subgraph N will be disturbed1, N2Each random division goes out the part of half, i.e. chromosome, exchange at random in population
Mating generates two new disturbance subgraph N3, N4；
Then, entire population G is traversedi, the new disturbance subgraph for randomly selecting out is added with former network exclusive or, is obtained
One new network；
It is divided finally, maximizing clustering algorithm using spectrum, by the former net before obtained new network structure and attack
Network structure is compared, and by comprehensively considering a series of indexs such as node degree, average shortest path length, is observed based on disturbance subgraph
Whether attack strategies method has achieved the effect that concealed nodes, and disturbance subgraph is carried out label, is put into database and obtains population
Between disturb subgraph library.
In the step S1, the corresponding feature vector of computing module degree matrix maximum eigenvalue, then according to vector element
Characteristic node is assigned in multiple corporations by recursive call, that is, divide multiple samples primary.
Technical concept of the invention are as follows: handle by network by certain mode on the basis of spectrum maximizes algorithm
To multiple populations.Operation for population, the present invention include two aspects: first aspect is operate to population
To the disturbance sub-graph data library of respective objects.The second aspect is disturbance for being operated to obtain respective objects to different population
Chart database.The present invention can generate the disturbance sub-graph data library of respective objects according to corresponding network, realize network social intercourse privacy
Protection.
Beneficial effects of the present invention are shown: the disturbance sub-graph data library of respective objects can be generated according to corresponding network,
Realize the protection of network social intercourse privacy.
Detailed description of the invention
Fig. 1 is flow diagram of the present invention.
Fig. 2 is disturbance subgraph aggressinogen network method schematic diagram, and (a) is former network, is (b) disturbance subgraph.
Fig. 3 is disturbance subgraph aggressinogen web results schematic diagram.
Specific embodiment
To make the objectives, technical solutions, and advantages of the present invention more comprehensible, with reference to the accompanying drawings and embodiments to this
Invention is described in further detail.It should be appreciated that the specific embodiments described herein are only used to explain the present invention,
And the scope of protection of the present invention is not limited.
Referring to Fig.1~Fig. 3, a kind of network privacy guard method based on disturbance subgraph, comprising the following steps:
S1: for a random network, the division that algorithm carries out community is maximized using spectrum, obtains multiple samples primary；
It is similar using part based on the basic ideas that can be converted to clustering algorithm after community discovery algorithm building similarity
Index constructs similar matrix, maximizes algorithm partition community using spectrum and obtains sample primary.
The present invention carrys out memory node information in a manner of Laplacian Matrix.
Firstly, all nodes in specified bilateral network to be given to the label L comprising nodal scheme1, L2, L3......。
Then adjacency matrix is formed: if there is even side, A between node i and jij=1, it is otherwise 0.Resettle one
A modularity expression formula divided for solving network, as follows:
Wherein, ciIt is group belonging to the i of vertex or corporations, δ (m, n) is the kronecker δ function, ki, kjIt is node i, node
The degree of j, B are modularity matrixes
And
BijIt has the property that
Consider to partition the network into two-part problem again, obtains formula
It can be written as
Wherein, s is by element siThe vector of composition, B are by element BijThe matrix of n × n of composition
It finds for giving modularity matrix B, keeps modularity Q maximum, obtain corresponding vector s.The element of s can only value
±1.It is different from corresponding Graph partition problem.Value be+1 or -1 number of elements be not it is fixed, i.e., the scale of corporations is not by about
Beam.
The optimization problem approximatively solves the problem using method of relaxation.Relax the restriction condition, that is, he only meet it is as follows
A constraint condition:
It adds single Lagrangian β to constrain it, by siIt differentiates and carrys out maximizing
It is available by derivative operation
Therefore, maximum modularity in order to obtain, it should enable the corresponding feature of maximum eigenvalue that s is modularity matrix to
Measure u1。
Because the value of s element is limited to si=± 1.But it can choose and be as closely as possible to u1, that is, make as inferior
Product maximizes:
Wherein [u1]iIt is u1I-th of element.When each single item in summation formula is non-negative, maximum value is obtained, that is,
Therefore it is based on the above thought, obtains following algorithm: the corresponding spy of maximum (positive number) characteristic value of computing module degree matrix
Vector is levied, then node is assigned in multiple corporations by recursive call according to the characteristic of vector element, that is, is divided more
A sample primary.
S2: one sample primary of random selection is handled using population dividing method, obtains the kind of multiple disturbance subgraphs
Group, process are as follows:
S2.1 maximizes clustering algorithm with spectrum above and obtains K samples primary, C1、C2、C3......CK, wherein with
Machine chooses a sample C primaryi, from the sample C primaryiIn randomly select n node.Randomly choose the n node of sample primary
In some node d, obtain selected node d in sample C primaryiUnder all neighbor nodes and non-neighbor node, formed section
The neighbor node set Φ of point d1With non-neighbours' node set Φ2。
S2.2, random erasure d and belong to Φ1Node company side (if it is not, the wheel is skipped), while random increasing
One d and belong to Φ2Node company side, update the neighborhood Φ of d1With non-neighborhood Φ2, repeat described process T times.
S2.3 repeats above-mentioned all steps until traversing n node all in the sample primary, obtains disturbance subgraph H,
And this n node is given to the new label L ' comprising nodal scheme with even frontier juncture system1, L '2, L '3......L′n。
S2.4 repeats above-mentioned all steps, obtains belonging to sample C primaryiN number of disturbance subgraph, if N number of disturbance subgraph
For population Gi。
S2.5 repeats above-mentioned all steps, obtains K population G1、G2、G3......CK。
S3: the thought based on evolutionary computation, the population for disturbing subgraph to one are handled, and generate new disturbance
Scheme population, process is as follows:
S3.1, from population GiIn random successively obtain two disturbance subgraph N1, N2, then will disturb subgraph N1, N2It is each random
The part (i.e. chromosome) for being partitioned into half swaps mating, generates two new disturbance subgraph N3, N4, then from disturbance
Scheme N3, N4In randomly select one and handled.
S3.2 repeats above step, until traversing entire population Gi。
S4: handling the population of new disturbance subgraph, obtains disturbing subgraph library in population, process is as follows:
The new disturbance subgraph for randomly selecting out is carried out exclusive or addition with former network, obtains a new net by S4.1
Network.
S4.2 is divided using identical community's partitioning algorithm, by the original before obtained new network structure and attack
Network structure is compared, and by comprehensively considering a series of indexs such as node degree, average shortest path length, observation is based on disturbance subgraph
Attack strategies method whether achieved the effect that concealed nodes, will disturbance subgraph carry out label, be put into database and planted
Subgraph library is disturbed between group.
It is wherein added about the exclusive or in S4.1, the present invention makes as given a definition.
Wherein, C (i, j) represents the company side of former network node i, j, and R (i, j) is the node i for disturbing subgraph, the company between j
Side, C ' (i, j) represent the company side of network node i, j after attack
Exclusive or is added concrete methods of realizing and does not connect side between two nodes as shown in Fig. 2, dotted line represents.
As shown in Figure 2, former network is compared with disturbance subgraph, and the node of the disturbance node of subgraph 1 and 4,4 and 3 does not connect side, and former
Network has, and disturbing subgraph and the former node of network 1 and 3 has even side, according to the criterion that exclusive or is added, the 1 of obtained attacking network
With 4 nodes, 4 and 3 nodes still have even side, and 1 and 3 nodes do not connect side, and arithmetic result is as shown in Figure 3.
S5: the thought based on evolutionary computation handles the population with disturbance subgraphs different under sample primary, generates one
A new disturbance subgraph population.The population of new disturbance subgraph is handled, obtains disturbing subgraph library between population.
The present invention has also been proposed further imagination on the basis of based on obtaining disturbing subgraph library in population, i.e., will be former
The range of choice of disturbance subgraph is expanded between population out of population, is implemented the steps of:
S5.1, from different population Gi, GjIn one disturbance subgraph N of each random acquisition1, N2, subgraph N will be disturbed1, N2It is each random
It is partitioned into the part of half, i.e. chromosome, exchange mating at random in population is carried out, generates two new disturbance subgraph N3, N4。
S5.2 traverses entire population Gi, the new disturbance subgraph for randomly selecting out is added with former network exclusive or, is obtained
One new network.
S5.3 maximizes clustering algorithm using spectrum and is divided, by the former net before obtained new network structure and attack
Network structure is compared, and by comprehensively considering a series of indexs such as node degree, average shortest path length, is observed based on disturbance subgraph
Whether attack strategies method has achieved the effect that concealed nodes, and disturbance subgraph is carried out label, is put into database and obtains population
Between disturb subgraph library.
Claims (7)
1. a kind of network privacy guard method based on disturbance subgraph, which is characterized in that the described method comprises the following steps:
S1: for a random network, the division that algorithm carries out community is maximized using spectrum, obtains multiple samples primary；
S2: one sample primary of random selection is handled using population dividing method, obtains the population of multiple disturbance subgraphs；
S3: being based on evolutionary computation, and the population for disturbing subgraph to one is handled, and generates a new disturbance subgraph population；
S4: handling the population of new disturbance subgraph, obtains disturbing subgraph library in population；
S5: being based on evolutionary computation, handles the population with disturbance subgraphs different under sample primary, generates a new disturbance
Subgraph population handles the population of new disturbance subgraph, obtains disturbing subgraph library between population.
2. the network privacy guard method according to claim 1 based on disturbance subgraph, which is characterized in that the step S2
In, population dividing method includes following procedure:
Firstly, maximizing clustering algorithm using spectrum obtains K samples primary, C1、C2、C3……CK, one is randomly selected wherein
Sample C primaryi, n node is randomly selected from the sample primary；
Then, some node d in n node is randomly selected, obtains selected node d in sample C primaryiUnder all neighbours
Node and non-neighbor node form the neighbor node set Φ of node d1With non-neighbours' node set Φ2；
Next, random erasure d and belonging to Φ1Node company side, if it is not, the wheel is skipped, while random increasing by one
D and belong to Φ2Node company side, update the neighborhood Φ of d1With non-neighborhood Φ2；It repeats the process T times；
Later, sample C primary is traversediMiddle n node obtains disturbance subgraph H, and it includes node mark that this n node, which is given new,
Number and even frontier juncture system label L '1, L '2, L '3……L′n；
Then, C is chosen againiN new node, and repetition n times operate to obtain and belong to sample C primaryiN number of disturbance subgraph,
And N number of disturbance subgraph is classified as population Gi；
Finally, repeating all operations obtains disturbance subgraph population G corresponding to K samples primary1、G2、G3……GK。
3. the network privacy guard method according to claim 1 or 2 based on disturbance subgraph, which is characterized in that the step
In rapid S3, the treatment process for disturbing the population of subgraph is as follows:
S3.1, from population GiIn random successively obtain two disturbance subgraph N1, N2, then will disturb subgraph N1, N2Each random division
The part of half swaps mating out, generates two new disturbance subgraph N3, N4, then from disturbance subgraph N3, N4In select at random
One is taken to be handled；
S3.2 repeats above step, until traversing entire population Gi。
4. the network privacy guard method according to claim 1 or 2 based on disturbance subgraph, which is characterized in that the step
In rapid S4, the process handled the population of new disturbance subgraph is as follows:
The new disturbance subgraph for randomly selecting out is carried out exclusive or addition with former network, obtains a new network by S4.1；
S4.2 is divided using identical community's partitioning algorithm, by the former network before obtained new network structure and attack
Structure is compared, and by comprehensively considering a series of indexs such as node degree, average shortest path length, observes attacking based on disturbance subgraph
It hits whether strategy process has achieved the effect that concealed nodes, disturbance subgraph is subjected to label, is put into database between obtaining population
Disturb subgraph library.
5. the network privacy guard method according to claim 4 based on disturbance subgraph, which is characterized in that in the S4.1
Exclusive or be added, make as given a definition.
Wherein, C (i, j) represents the company side of former network node i, j, and R (i, j) is the node i for disturbing subgraph, the company side between j, C '
(i, j) represents the company side of network node i, j after attack.
6. the network privacy guard method according to claim 1 or 2 based on disturbance subgraph, which is characterized in that the step
It is as follows that subgraph library generating process is disturbed in rapid S5, between population:
Firstly, from different population Gi, GjIn one disturbance subgraph N of each random acquisition1, N2；
Secondly, subgraph N will be disturbed1, N2Each random division goes out the part of half, i.e. chromosome, carries out exchange friendship at random in population
Match, generates two new disturbance subgraph N3, N4；
Then, entire population G is traversedi, the new disturbance subgraph for randomly selecting out is added with former network exclusive or, obtains one
New network；
It is divided finally, maximizing clustering algorithm using spectrum, by the former network knot before obtained new network structure and attack
Structure is compared, and by comprehensively considering a series of indexs such as node degree, average shortest path length, observes the attack based on disturbance subgraph
Whether strategy process has achieved the effect that concealed nodes, and disturbance subgraph is carried out label, is put into database and obtains disturbing between population
Mover picture library.
7. the network privacy guard method according to claim 1 based on disturbance subgraph, which is characterized in that the step S1
In, the corresponding feature vector of computing module degree matrix maximum eigenvalue is adjusted then according to the characteristic of vector element by circulation
With node is assigned in multiple corporations, that is, divides multiple samples primary.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910371124.3A CN110175634B (en) | 2019-05-06 | 2019-05-06 | Network privacy protection method based on disturbance subgraph |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910371124.3A CN110175634B (en) | 2019-05-06 | 2019-05-06 | Network privacy protection method based on disturbance subgraph |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110175634A true CN110175634A (en) | 2019-08-27 |
| CN110175634B CN110175634B (en) | 2021-05-18 |
Family
ID=67691138
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910371124.3A Active CN110175634B (en) | 2019-05-06 | 2019-05-06 | Network privacy protection method based on disturbance subgraph |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110175634B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105869053A (en) * | 2016-03-23 | 2016-08-17 | 西安电子科技大学 | Two-stage memetic based social network influence maximizing method |
| CN106203172A (en) * | 2016-07-25 | 2016-12-07 | 浙江工业大学 | Weighting sensitivity limit method for secret protection on network shortest path |
| CN107179954A (en) * | 2017-04-25 | 2017-09-19 | 内蒙古科技大学 | A kind of distributed community network method for secret protection of holding node accessibility |
| CN108234493A (en) * | 2018-01-03 | 2018-06-29 | 武汉大学 | The space-time crowdsourcing statistical data dissemination method of secret protection under insincere server |
| CN109063836A (en) * | 2018-06-28 | 2018-12-21 | 浙江工业大学 | A kind of privacy link protection method based on disturbance of evolving |
-
2019
- 2019-05-06 CN CN201910371124.3A patent/CN110175634B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105869053A (en) * | 2016-03-23 | 2016-08-17 | 西安电子科技大学 | Two-stage memetic based social network influence maximizing method |
| CN106203172A (en) * | 2016-07-25 | 2016-12-07 | 浙江工业大学 | Weighting sensitivity limit method for secret protection on network shortest path |
| CN107179954A (en) * | 2017-04-25 | 2017-09-19 | 内蒙古科技大学 | A kind of distributed community network method for secret protection of holding node accessibility |
| CN108234493A (en) * | 2018-01-03 | 2018-06-29 | 武汉大学 | The space-time crowdsourcing statistical data dissemination method of secret protection under insincere server |
| CN109063836A (en) * | 2018-06-28 | 2018-12-21 | 浙江工业大学 | A kind of privacy link protection method based on disturbance of evolving |
Non-Patent Citations (3)
| Title |
|---|
| SHANQING YU等: "Target Defense Against Link-Prediction-Based Attacks via Evolutionary Perturbations", 《SOCIAL AND INFORMATION NETWORKS》 * |
| 卢春雨: "基于扰动矩阵的社会网络隐私保护方法研究", 《中国优秀硕士论文全文数据库》 * |
| 王小号等: "基于谱约束和敏感区划分的社会网络隐私保护扰动方法", 《计算机应用》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110175634B (en) | 2021-05-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Milli et al. | Model reconstruction from model explanations | |
| Mopuri et al. | Fast feature fool: A data independent approach to universal adversarial perturbations | |
| Tian et al. | Federboost: Private federated learning for gbdt | |
| Keikha et al. | Influence maximization across heterogeneous interconnected networks based on deep learning | |
| Chen et al. | Multiscale evolutionary perturbation attack on community detection | |
| CN107871087A (en) | The personalized difference method for secret protection that high dimensional data is issued under distributed environment | |
| Baek et al. | Personalized subgraph federated learning | |
| Tu et al. | gl2vec: Learning feature representation using graphlets for directed networks | |
| CN111475838B (en) | Deep neural network-based graph data anonymizing method, device and storage medium | |
| Vasa et al. | Deep learning: Differential privacy preservation in the era of big data | |
| Zhu et al. | A fine-grained differentially private federated learning against leakage from gradients | |
| Lunardi et al. | Comparative study of genetic and discrete firefly algorithm for combinatorial optimization | |
| Hu et al. | A new algorithm CNM-Centrality of detecting communities based on node centrality | |
| Geng et al. | Improved gradient inversion attacks and defenses in federated learning | |
| Khanday et al. | A comparative analysis of identifying influential users in online social networks | |
| CN110175634A (en) | A kind of network privacy guard method based on disturbance subgraph | |
| Li et al. | FDPBoost: Federated differential privacy gradient boosting decision trees | |
| Avram et al. | Adversarial perturbations to manipulate the perception of power and influence in networks | |
| Mekouar et al. | Community outlier detection in social networks based on graph matching | |
| Wei et al. | GNN-Ensemble: Towards Random Decision Graph Neural Networks | |
| Eshtay et al. | NMVSA greedy solution for vertex cover problem | |
| Khan et al. | Synthetic Identity Detection using Inductive Graph Convolutional Networks | |
| Wang et al. | Federated Knowledge Graph Completion via Latent Embedding Sharing and Tensor Factorization | |
| Yin et al. | Structural-based graph publishing under differential privacy | |
| WU et al. | FedInverse: Evaluating Privacy Leakage in Federated Learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20230412 Address after: Room 201, Building 3, No. 575, Huxi Road, Changhe Street, Binjiang District, Hangzhou City, Zhejiang Province, 310051 Patentee after: Hangzhou Jiangshang Printing Technology Co.,Ltd. Address before: The city Zhaohui six districts Chao Wang Road Hangzhou City, Zhejiang province 310014 18 Patentee before: JIANG University OF TECHNOLOGY |