CN114363464A - Method and system for suppressing fraud information propagation - Google Patents
Method and system for suppressing fraud information propagation Download PDFInfo
- Publication number
- CN114363464A CN114363464A CN202111658983.4A CN202111658983A CN114363464A CN 114363464 A CN114363464 A CN 114363464A CN 202111658983 A CN202111658983 A CN 202111658983A CN 114363464 A CN114363464 A CN 114363464A
- Authority
- CN
- China
- Prior art keywords
- fraud
- resource
- resources
- information
- fraud information
- 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
- 238000000034 method Methods 0.000 title claims abstract description 52
- 238000011084 recovery Methods 0.000 claims abstract description 26
- 239000011159 matrix material Substances 0.000 claims description 47
- 239000002245 particle Substances 0.000 claims description 25
- 239000013598 vector Substances 0.000 claims description 23
- 238000005457 optimization Methods 0.000 claims description 19
- 230000006870 function Effects 0.000 claims description 16
- 238000006243 chemical reaction Methods 0.000 claims description 14
- 230000036039 immunity Effects 0.000 claims description 13
- 208000015181 infectious disease Diseases 0.000 claims description 13
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000001914 filtration Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000013468 resource allocation Methods 0.000 claims description 6
- 230000007704 transition Effects 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 230000001900 immune effect Effects 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 238000004891 communication Methods 0.000 claims description 2
- 238000004422 calculation algorithm Methods 0.000 abstract description 10
- 230000002028 premature Effects 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000008439 repair process Effects 0.000 description 2
- 238000012163 sequencing technique Methods 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000009792 diffusion process Methods 0.000 description 1
- 230000008014 freezing Effects 0.000 description 1
- 238000007710 freezing Methods 0.000 description 1
- 230000002068 genetic effect Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 108090000765 processed proteins & peptides Proteins 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000007480 spreading Effects 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a method for suppressing fraud information propagation, which comprises the following steps: analyzing the network information to acquire fraud information; constructing a fraud information propagation network model according to the fraud information; obtaining anti-fraud social resources, wherein the anti-fraud social resources comprise immune resources, protection resources and recovery resources; optimizing the anti-fraud social resources to obtain a first priority resource, a second priority resource and a third priority resource; the first priority resource, the second priority resource and the third priority resource are subjected to node notification through a fraud information propagation network model, so that fraud information propagation is restrained; the invention introduces the concept of resource priority to preprocess the solution space, reduces the retrieval range and improves the retrieval efficiency; by means of hierarchical learning, diversity in spatial search is improved, premature trapping is avoided, performance of a group intelligent algorithm is improved, and the problem of phishing information propagation containment is solved more efficiently.
Description
Technical Field
The invention relates to the research field of anti-fraud and evolution calculation of a network, in particular to a method and a system for suppressing fraud information propagation.
Background
The rampant of fraud information not only causes property loss of people, but also disturbs normal social and economic orders, and becomes the critical crime field of the country. In the internet era with highly developed information, the channels for spreading fraud information are very diverse, including but not limited to short message fraud, telephone fraud, internet fraud, offline fraud, etc., and thus the social resources such as police force, public right, anti-fraud propaganda, etc. required for use are greatly increased. In this scenario, anti-fraud is increasingly demanding on computer technology.
Anti-fraud as a whole project requires a great deal of social resources to be mobilized. This mainly refers to the investment in anti-fraud activities. With the existing propagation resource model, we can roughly divide anti-fraud social resources into three types: (1) the immune resources, namely police carry out targeted case explanation and community push according to the detected fraud information which is widely spread or has a diffusion tendency at present, so as to improve the immunity of a specific group to the specific fraud information, for example, entering the community to specifically explain alarm cases such as 'credit card fraud aiming at college student groups' and 'financial fraud aiming at the elderly group'; (2) the protection resources mainly refer to the improvement of the overall protection consciousness, including the realization of popular science education work realized by means of 'anti-fraud public numbers' and the realization of anti-fraud technologies like 'fraud telephone tracing and interception', and the like, and finally the individual protection capability is improved; (3) the resources are recovered, namely the fraud crime facts are detected, the property loss of the people is recovered, and the order of the individuals and the social life is recovered. Each type of resource contains different degrees of human and material resource investment and also corresponds to different effects. On the premise of limited financial budget and human resources, it is a deliberate task to fully, effectively and efficiently schedule these resources and to suppress the spread of fraud information to the maximum extent.
The existing resource scheduling methods comprise a traditional global optimization method, a heuristic optimization method and a meta-heuristic method. The global optimization method represented by convex optimization and quasi-convex optimization has rigorous mathematical derivation and theoretical proof, is suitable for solving the optimization problem with regular solution space, other complex problems need to be mapped into the convex optimization problem by an approximation method, and the process takes the loss of the precision of the solution as cost; heuristic methods are designed aiming at problem characteristics, the aim is to find out an approximate optimal solution with acceptable time efficiency, and the methods have poor mobility and adaptability due to the strong personalized definition characteristic aiming at the cultural and literature characteristics; the meta-heuristic method is a parallelizable heuristic method inspired by the biological evolution of the nature and not directly related to the problem, is not constrained by restrictive assumptions such as continuity and conductability of a search space, and can find an approximate global optimal solution from a discrete, multi-peak and noisy high-dimensional problem, so that the method is very good at solving a complex optimization problem.
Disclosure of Invention
The invention mainly aims to overcome the defects and shortcomings of the prior art and provides a method and a system for suppressing fraud information transmission. The ultimate optimization goal is to achieve a minimization of the impact of fraud information propagation, where the impact is primarily assessed by the propagation decay rate of the information. In the inventive algorithm, priority planning and hierarchical learning methods are first applied to generate probabilities of resource selection, and then the trade-off of resources of different priorities is determined according to the probabilities when constructing a solution.
A first object of the present invention is to provide a method of suppressing the spread of fraud information;
it is a second object of the present invention to provide a system for suppressing the spread of fraud information.
The purpose of the invention is realized by the following technical scheme:
a method of suppressing the spread of fraud information, comprising the steps of:
analyzing the network information to acquire fraud information;
constructing a fraud information propagation network model according to the fraud information;
obtaining anti-fraud social resources, wherein the anti-fraud social resources comprise immune resources, protection resources and recovery resources;
optimizing the anti-fraud social resources to obtain a first priority resource, a second priority resource and a third priority resource;
and carrying out node notification on the first priority resource, the second priority resource and the third priority resource through a fraud information propagation network model so as to suppress fraud information propagation.
Further, the analyzing the communication information to obtain the fraud information specifically includes: and screening the network information through a fraud analysis system to acquire fraud information.
Further, the building of the fraud information propagation network model according to the fraud information specifically includes: constructing a state space of the user according to the phishing information, and assuming that the users in the network are nodes, each node i will be in a conversion of four states, including: susceptible state S, exposed state E, cheated state I, and recovered state R;
the probability vectors corresponding to the different states are represented as:
where t denotes the time, i denotes the user number,
respectively, the probabilities that the user i is in S, E, I, R four states at this time point, and the sum of the four probabilities is 1.
Further, there is a limit condition for the node to perform conversion in four states, specifically: the target person of the artificial fraud information in the susceptible state generates immunity to the fraud information with a first probability after receiving the fraud information, wherein the first probability is an immunity rate thetai(ii) a If the person in susceptible state does not generate immunity for the first time, the person is converted into exposure state with a second probability, wherein the second probability is infection rate ui(ii) a The exposed person, upon receiving the fraud information, enters the fraud state with a third probability, the third probability beingProbability is conversion xii(ii) a After the person in the cheated state is identified as being cheated, stopping loss according to a fourth probability, and entering a recovery state, wherein the fourth probability is the recovery rate
Persons in a recovering state are immune to fraud information;
the state transition process corresponding to the four states is as follows:
wherein ,θiFor the immunological rate, uiIs the infection rate, xiiIn order to achieve a high conversion rate,
for the recovery rate, t represents the time, i represents the user number,
respectively representing the probability that the user i is in S, E, I, R four states at the moment;
in the model, the linear upper bound corresponding to the change of the state E and the state I can be represented by a probability matrix L, and the specific expression is as follows:
wherein ,
denotes an identity matrix, T ═ diag ([ θ [ ])1,...,θN]) Matrix representing diagonal element as node immunity rate, U ═ diag ([ U ] s)1,...,uN]) Matrix representing off-diagonal element as node infection rate, F ═ diag ([ ξ [ ]1,...,ξN]) A matrix representing diagonal elements as nodal conversion, D ═ diag ([ δ [ ]1,...,δN]) Element representing diagonal element as node recovery rateA peptide;
the real part of the maximum eigenvalue of the matrix L represents the exponential growth rate of the probability of the population in these two states, denoted as λ (L); the goal to be optimized is to minimize the growth rate of the population in the E and I states, i.e., min (λ (L)).
Further, the obtaining of anti-fraud social resources includes immune resources, protection resources, and recovery resources, and specifically includes: immune resource R1For increasing the immune rate thetaiProtection resource R2For reducing the infection rate uiAnd recovering the resource R3For increasing the recovery rate deltai；
By R ═ τri]Representing resource allocation matrices by τriWhen the resource matrix indicates that the r-th resource is allocated to the i-th node, the format of the resource matrix is:
wherein ,τriIndicating that the r resource is allocated to the ith node;
different resources have different costs, using c1(·)、c2(·)、c3(. represents resources R, respectively)1、R2、R3Unit cost of (2); the total cost is the sum of the costs of the various types of resources, expressed as:
wherein, i is 1,2, and N represents a node number, and c is used as1(·)、c2(·)、c3(. represents resources R, respectively)1、R2、R3Unit cost of (2).
Further, the optimizing the anti-fraud social resource to obtain a first priority resource, a second priority resource, and a third priority resource specifically includes:
after initializing parameters and a population, dividing the sorted individuals into NG groups, wherein the higher the fitness value of the individual is, the higher the group is, namely the smaller the serial number of the group is, and otherwise, the lower the group is, namely the larger the serial number of the group is; the initialized parameters include: the size NP of the population, the number NG of the groups and the dimension D of the solution;
by NPgIndicates the total number of individuals in group g, in the pre-NG-1 group, NPgFloor (NP/NG); in the last group, NPg＝NP-(NG-1)×floor(NP/NG)；
The position and velocity of the r-th particle in the g-th group are respectively expressed as
And
wherein g 1,2, NG denotes the group in which the particles are located, and r 1,2, NPgThe number of the particle in the group is represented, D is 1,2, and D represents the dimension of the problem;
the individuals of the low-level group need to learn from two listed individuals randomly drawn from the high-level group and update the velocity Vg,r。
Further, the update speed is specifically as follows:
assuming two sample individuals and located in the g ' and g ' groups, respectively, and (1 ≦ g ' ≦ g-1), the velocity update formula is:
wherein ,
which is indicative of the velocity of the particles,
denotes the position of the particle, r1 and r2Is two value ranges of [0,1 ]]W represents the degree of influence of the past speed on the current speed, c is a constant, NG represents the number of particle groups, g 'and g' represent two chartsThe group of the sample particles, r 'and r' represent the intra-group numbers of the two sample particles;
further, still include: constructing a candidate solution for assisting in updating the location of the individual; the method comprises the following specific steps:
updating individual locations by means of priority planning and hierarchical learning:
h1 learning: setting a threshold parameter th ∈ [0,1 ]]Then filtering out the velocity V(g,r)The middle probability is larger than th, and the positions of the values are resources with first priority; obtaining an explicit set of the first priority resource according to the position of the first priority resource and by means of a 0 or 1 value randomly generated by the probability of the position in the velocity matrix, wherein the explicit set is used for constructing NEW _ X(g,r)The update rule is:
wherein ,
representing the matrix Cut _ V(g,r)The element in (1), S (-) is sigmoid function, tau represents filtering threshold, phi (-) is binary function, and the expression is:
wherein NEW _ X(g,r)Representing the new location, cost (-) represents a cost function, C represents a cost constraint, elem represents an element of the resource allocation matrix,
represents an add resource operation, i.e., changing a given element in a vector to 1;
h2 learning: from Pbest(g,r)Voting Gbest to obtain second priority resource and selection probability thereof, and passing through second priority resource Cut _ X(g,r)To construct NEW _ X(g,r)。Cut_X(g,r) and NEW_X(g,r)The update formula of (2) is as follows:
wherein 
Denotes Cut _ X(g,r)The element in (1), rand (0,1) represents a random number between 0 and 1, Sigmoid (·) represents a Sigmoid function,
representing Gbest and Pbest(g,r)The elements of the matrix obtained after the voting,
representing the matrix NEW _ X(g,r)The elements of (1);
h3 learning: let vector Other _ X(g,r)Represents NEW _ X(g,r)The opposite vector is that the 0 value and the 1 value at each position in the vector are interchanged to obtain the third priority resource; add not in NEW _ X without violating constraints(g,r)To construct NEW _ X from random resources in (2)(g,r)，NEW_X(g,r)The update formula is:
wherein ,
is the matrix Other _ X(g,r)The elements (A) and (B) in (B),
representing the matrix NEW _ X(g,r)Of (1).
Further, still include: updating individual best Pbest(g,r)And global optimal solution Gbest of the whole population; the method specifically comprises the following steps:
and calculating the fitness values corresponding to all the individual positions, selecting the optimal one of the fitness values, and comparing the selected optimal one with the fitness value of the old global optimal position Gbest. If the former is better than the latter, Gbest is updated, otherwise it remains unchanged.
The second purpose of the invention is realized by the following technical scheme:
a system of suppressing the propagation of fraud information, comprising:
the network information analysis module is used for analyzing the network information and acquiring fraud information;
the model construction module is used for constructing a fraud information propagation network model according to the fraud information;
the anti-fraud social resource acquisition module is used for acquiring anti-fraud social resources, wherein the anti-fraud social resources comprise immune resources, protection resources and recovery resources;
the resource optimization module is used for optimizing the anti-fraud social resources to obtain a first priority resource, a second priority resource and a third priority resource;
and the node notification module is used for notifying the first priority resource, the second priority resource and the third priority resource of the nodes through a fraud information propagation network model so as to further inhibit fraud information propagation.
The working process of the system is as follows:
(1) and (4) defining problems. The decision variable of the problem is the allocation matrix of the resources, the constraint conditions are limited financial budget and human resources (collectively referred to as resource cost), and the optimization objective function is the minimization of the propagation rate of the fraud information.
(2) Parameters and populations are initialized. The parameters that need to be initialized include: the population size NP, the number of packets NG, and the dimension D of the solution. And randomly generating a candidate population, and validating the solution provided by the candidate population by means of a repair function to obtain a formal first generation population. The criterion for a legal solution is that "the total cost of the solution does not exceed a given financial budget". Initializing individual optimal Pbest from first generation population(g,r)And global optimal solution Gbest for the entire population.
(3) And (6) updating the speed. Firstly, according to fitness evaluation indexes, sequencing individuals in a population from high to low, then grouping the individuals, wherein the individuals in a low-level group need to learn from any two individuals in a high-level group, and updating the speed of the individuals.
(4) And constructing a candidate solution. The candidate solution is used for assisting in updating the position of the individual, initializing
Is a zero vector.
(5) And planning the priority of the resources. A resource here refers to a non-zero variable in the position vector. The priority of these elements is divided into three levels. The first priority being from speed
Is collected in a definite set, the threshold parameter th has a value range of [0, 1%]For adjusting Cut _ X(g,r)The number of resources of the explicit set of (2). Resources of the second priority are represented by Pbest(g,r)And the Gbest voting, firstly estimating the selection probability of each resource, and then putting the resources into a vector according to the probability
The remaining resources are the third priority resources.
(6) The candidate solutions are updated through a layered learning mechanism. Updating candidate solution NEW _ X by using resources of first priority and second priority in sequence without violating constraint(g,r)And finally randomly selecting a part of resources with the third priority according to the residual cost budget to perfect NEW _ X(g,r)。
(7) And (4) updating the position. Assigning the obtained legal candidate solution to a position vector, i.e. X(g,r)＝NEW_X(g,r)。
(8) Updating individual best Pbest(g,r)And global optimal solution Gbest for the entire population. And calculating the fitness values corresponding to all the individual positions, selecting the optimal one of the fitness values, and comparing the selected optimal one with the fitness value of the old global optimal position Gbest. If the former is better than the latter, then update Gbest, otherwise keepAnd is not changed.
If the end condition is reached, the optimization procedure is ended, otherwise, the step (3) is returned to.
Compared with the prior art, the invention has the following advantages and beneficial effects:
the invention introduces the concept of resource priority to preprocess the solution space, reduces the retrieval range and improves the retrieval efficiency; by means of hierarchical learning, diversity in spatial search is improved, premature trapping is avoided, performance of a group intelligent algorithm is improved, and the problem of phishing information propagation containment is solved more efficiently.
Drawings
FIG. 1 is a flow chart of a method of suppressing the propagation of fraud information in the present invention;
FIG. 2 is a schematic structural diagram of a fraud information propagation network model according to embodiment 1 of the present invention;
FIG. 3 is a schematic diagram of the anti-fraud social resource role of embodiment 1 of the present invention;
FIG. 4 is a flow chart of social resource optimization by the swarm intelligence algorithm according to embodiment 1 of the invention;
FIG. 5 is a block diagram of a system architecture for suppressing the propagation of fraud information according to the present invention.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited thereto.
Example 1
A method of suppressing the spread of fraud information, as shown in fig. 1, comprising the steps of:
analyzing the network information to acquire fraud information; the method specifically comprises the following steps: and screening the network information through a fraud analysis system to acquire fraud information.
Constructing a fraud information propagation network model according to the fraud information, wherein the fraud information propagation network model is shown in FIG. 2; the method specifically comprises the following steps: suppose there is a crowd network G containing N nodesN. For any one of the nodes i (i ═ 1, 2.., N), it may be in any one of four states: susceptible State (Susceptible, S for short)The node is converted in four states, namely an Exposed state (E), an cheated state (I) and a Recovered state (R); the target personnel in the susceptible state S can convert the fraud information into an immune state with a first probability after receiving the fraud information, wherein the first probability is an immunity rate; if the person in the susceptible state does not generate immunity for the first time, the person is converted into the exposure transition state with a second probability, and the second probability is the infection rate; after receiving the fraud information, the person in the exposure state enters a fraud state according to a third probability, wherein the third probability is a conversion rate; after the person in the cheated state is identified to be cheated, stopping loss according to a fourth probability, and entering a recovery state, wherein the fourth probability is a recovery rate; persons in a recovering state are immune to fraud information;
the probability vectors for node states are as follows:
where t denotes the time, i denotes the user number,
respectively, the probabilities that the user i is in S, E, I, R four states at this time point, and the sum of the four probabilities is 1.
The transformation between the four states is described as follows:
s → E. The susceptible person in the S state is a target group of criminals, and after receiving the fraud information, the susceptible person is converted into an exposure state (i.e. a high-risk state that has been locked by the criminals and is close to a fraudulent state) with a certain probability. This probability, called the infection rate, is denoted as ui|0＜ui＜1}。
E → I. The exposer in the state E can be cheated by criminals with a certain probability due to factors such as insufficient anti-cheating awareness of the exposer or personal information theft, enter a cheated state I, and suffer property or other losses. The corresponding probability is referred to as the conversion,denoted as { ξi|0＜ξi＜1}。
I → R. When the cheated person in the I state is aware of the possibility of being cheated, measures such as freezing the bank card, contacting customer service, alarming and the like are taken, and the measures are prevented with a certain probability. This probability, called the recovery rate, is denoted as { δ }i|0＜δi＜1}。
S → R, the susceptible person in S state may have been familiar with the specific fraud loop in advance, and immunize the received fraud information so as to directly enter the recovery state R without passing through the states E and I, and the corresponding immunity rate is expressed as { theta }i|0＜θi＜1}。
Node i may be in S, E, I, R one of four states. Converting in four states, specifically; the node is converted in four states, and the state conversion process is as follows:
wherein ,θiFor the immunological rate, uiIs the infection rate, xiiIn order to achieve a high conversion rate,
for the recovery rate, t represents the time, i represents the user number,
respectively representing the probability that the user i is in S, E, I, R four states at the moment;
in the model, the linear upper bound corresponding to the change of the state E and the state I can be represented by a probability matrix L, and the specific expression is as follows:
wherein ,
denotes an identity matrix, T ═ diag ([ θ [ ])1,...,θN]) Matrix representing diagonal element as node immunity rate, U ═ diag ([ U ] s)1,...,uN]) Matrix representing off-diagonal element as node infection rate, F ═ diag ([ ξ [ ]1,...,ξN]) A matrix representing diagonal elements as nodal conversion, D ═ diag ([ δ [ ]1,...,δN]) The diagonal elements are elements representing the node recovery rate;
the real part of the maximum eigenvalue of the matrix L represents the exponential growth rate of the probability of the population in these two states, denoted as λ (L); the goal to be optimized is to minimize the growth rate of the population in the E and I states, i.e., min (λ (L)).
Obtaining anti-fraud social resources, wherein the anti-fraud social resources comprise immune resources, protection resources and recovery resources; the method specifically comprises the following steps: immune resource R1For increasing the immune rate thetaiProtection resource R2For reducing the infection rate uiAnd recovering the resource R3For increasing the recovery rate deltai(ii) a The anti-fraud social resource role is schematically illustrated in FIG. 3.
By R ═ τri]Representing resource allocation matrices by τriWhen the resource matrix indicates that the r-th resource is allocated to the i-th node, the format of the resource matrix is:
wherein τriIndicating that the r-th resource is allocated to the i-th node.
Different resources have different costs, using c1(·)、c2(·)、c3(. represents resources R, respectively)1、R2、R3Unit cost of (2); the total cost is the sum of the costs of the various types of resources, expressed as:
wherein, i is 1,2, and N represents a node number, and c is used as1(·)、c2(·)、c3(. represents resources R, respectively)1、R2、R3Unit cost of (2).
Optimizing the anti-fraud social resources to obtain a first priority resource, a second priority resource and a third priority resource;
the flow chart of the social resource optimization by the group intelligence algorithm is shown in fig. 4, and specifically includes:
after initializing parameters and a population, dividing the sorted individuals into NG groups, wherein the higher the fitness value of the individual is, the higher the group is, namely the smaller the serial number of the group is, and otherwise, the lower the group is, namely the larger the serial number of the group is; the initialized parameters include: the size NP of the population, the number NG of the groups and the dimension D of the solution;
by NPgIndicates the total number of individuals in group g, in the pre-NG-1 group, NPgFloor (NP/NG); in the last group, NPg＝NP-(NG-1)×floor(NP/NG)；
The position and velocity of the r-th particle in the g-th group are respectively expressed as
And
wherein g 1,2, NG denotes the group in which the particles are located, and r 1,2, NPgThe number of the particle in the group is represented, D is 1,2, and D represents the dimension of the problem;
the individuals of the low-level group need to learn from two listed individuals randomly drawn from the high-level group and update the velocity Vg,r。
Further, the update speed is specifically as follows:
assuming two sample individuals and located in the g ' and g ' groups, respectively, and (1 ≦ g ' ≦ g-1), the velocity update formula is:
wherein ,
which is indicative of the velocity of the particles,
denotes the position of the particle, r1 and r2Is two value ranges of [0,1 ]]W represents the influence degree of the past speed on the current speed, c is a constant, NG represents the number of particle groups, g ' and g ' represent groups where two sample particles are located, and r ' represent intra-group numbers of the two sample particles;
further, still include: constructing a candidate solution for assisting in updating the location of the individual; the method comprises the following specific steps:
updating individual locations by means of priority planning and hierarchical learning:
h1 learning: setting a threshold parameter th ∈ [0,1 ]]Then filtering out the velocity V(g,r)The middle probability is larger than th, and the positions of the values are the resources with the first priority; obtaining an explicit set of the first priority resource according to the position of the first priority resource and by means of a 0 or 1 value randomly generated by the probability of the position in the velocity matrix, wherein the explicit set is used for constructing NEW _ X(g,r)The update rule is:
wherein 
Representing the matrix Cut _ V(g,r)The element in (1), S (-) is sigmoid function, tau represents filtering threshold, phi (-) is binary function, and the expression is:
NEW _ X in the formula(g,r)Representing new position, cost (-) representing cost function, C representing costA constraint, elem, denotes an element of the resource allocation matrix,
represents an add resource operation, i.e., changing a given element in a vector to 1;
h2 learning: from Pbest(g,r)And the Gbest votes to obtain a second priority resource and the selection probability thereof, and constructs NEW _ X through the second priority resource(g,r)；Cut_X(g,r)The update formula of (2) is as follows:
wherein 
Denotes Cut _ X(g,r)The element in (1), rand (0,1) represents a random number between 0 and 1, Sigmoid (·) represents a Sigmoid function,
representing Gbest and Pbest(g,r)The elements of the matrix obtained after the voting,
representing the matrix NEW _ X(g,r)Of (1).
H3 learning: let vector Other _ X(g,r)Represents NEW _ X(g,r)The opposite vector is that the 0 value and the 1 value at each position in the vector are interchanged to obtain the third priority resource; add not in NEW _ X without violating constraints(g,r)To construct NEW _ X from random resources in (2)(g,r)，NEW_X(g,r)The update formula is:
wherein ,
is the matrix Other _ X(g,r)The elements (A) and (B) in (B),
representing the matrix NEW _ X(g,r)Of (1).
Further, still include: updating individual best Pbest(g,r)And global optimal solution Gbest of the whole population; the method specifically comprises the following steps:
and calculating the fitness values corresponding to all the individual positions, selecting the optimal one of the fitness values, and comparing the selected optimal one with the fitness value of the old global optimal position Gbest. If the former is better than the latter, Gbest is updated, otherwise it remains unchanged.
And carrying out node notification on the first priority resource, the second priority resource and the third priority resource through a fraud information propagation network model so as to suppress fraud information propagation.
In order to prove the effectiveness of the group intelligent algorithm adopted in the method, the genetic algorithm, the classical discrete particle swarm algorithm, the differential evolution algorithm, the random method and the like are used for carrying out optimization design on similar problems. The algorithm comparison result shows that the adopted group intelligent method has the leading performance in optimizing the network transmission resource scheduling problem. This proves that the phishing information propagation containment method invented by the group intelligent method is also very promising.
Example 2
A system for suppressing the spread of fraud information, as shown in fig. 5, comprising:
the network information analysis module is used for analyzing the network information and acquiring fraud information;
the model construction module is used for constructing a fraud information propagation network model according to the fraud information;
the anti-fraud social resource acquisition module is used for acquiring anti-fraud social resources, wherein the anti-fraud social resources comprise immune resources, protection resources and recovery resources;
the resource optimization module is used for optimizing the anti-fraud social resources to obtain a first priority resource, a second priority resource and a third priority resource;
and the node notification module is used for notifying the first priority resource, the second priority resource and the third priority resource of the nodes through a fraud information propagation network model so as to further inhibit fraud information propagation.
The working process of the system is as follows:
(9) and (4) defining problems. The decision variable of the problem is the allocation matrix of the resources, the constraint conditions are limited financial budget and human resources (collectively referred to as resource cost), and the optimization objective function is the minimization of the propagation rate of the fraud information.
(10) Parameters and populations are initialized. The parameters that need to be initialized include: the population size NP, the number of packets NG, and the dimension D of the solution. And randomly generating a candidate population, and validating the solution provided by the candidate population by means of a repair function to obtain a formal first generation population. The criterion for a legal solution is that "the total cost of the solution does not exceed a given financial budget". Initializing individual optimal Pbest from first generation population(g,r)And global optimal solution Gbest for the entire population.
(11) And (6) updating the speed. Firstly, according to fitness evaluation indexes, sequencing individuals in a population from high to low, then grouping the individuals, wherein the individuals in a low-level group need to learn from any two individuals in a high-level group, and updating the speed of the individuals.
(12) And constructing a candidate solution. The candidate solution is used for assisting in updating the position of the individual, initializing
Is a zero vector.
(13) And planning the priority of the resources. A resource here refers to a non-zero variable in the position vector. The priority of these elements is divided into three levels. The first priority being from speed
Is collected in a definite set, the threshold parameter th has a value range of [0, 1%]For adjusting Cut _ X(g,r)The number of resources of the explicit set of (2). Resource of the second priority is composed ofPbest(g,r)And the Gbest voting, firstly estimating the selection probability of each resource, and then putting the resources into a vector according to the probability
The remaining resources are the third priority resources.
(14) The candidate solutions are updated through a layered learning mechanism. Updating candidate solution NEW _ X by using resources of first priority and second priority in sequence without violating constraint(g,r)And finally randomly selecting a part of resources with the third priority according to the residual cost budget to perfect NEW _ X(g,r)。
(15) And (4) updating the position. Assigning the obtained legal candidate solution to a position vector, i.e. X(g,r)＝NEW_X(g,r)。
(16) Updating individual best Pbest(g,r)And global optimal solution Gbest for the entire population. And calculating the fitness values corresponding to all the individual positions, selecting the optimal one of the fitness values, and comparing the selected optimal one with the fitness value of the old global optimal position Gbest. If the former is better than the latter, Gbest is updated, otherwise it remains unchanged.
If the end condition is reached, the optimization procedure is ended, otherwise, the step (3) is returned to.
The above embodiments are preferred embodiments of the present invention, but the present invention is not limited to the above embodiments, and any other changes, modifications, substitutions, combinations, and simplifications which do not depart from the spirit and principle of the present invention should be construed as equivalents thereof, and all such changes, modifications, substitutions, combinations, and simplifications are intended to be included in the scope of the present invention.
Claims (10)
1. A method of suppressing the spread of fraud information, comprising the steps of:
analyzing the network information to acquire fraud information;
constructing a fraud information propagation network model according to the fraud information;
obtaining anti-fraud social resources, wherein the anti-fraud social resources comprise immune resources, protection resources and recovery resources;
optimizing the anti-fraud social resources to obtain a first priority resource, a second priority resource and a third priority resource;
and carrying out node notification on the first priority resource, the second priority resource and the third priority resource through a fraud information propagation network model so as to suppress fraud information propagation.
2. The method for suppressing the propagation of fraud information of claim 1, wherein said analyzing the communication information to obtain fraud information comprises: and screening the network information through a fraud analysis system to acquire fraud information.
3. The method of suppressing fraud information propagation according to claim 1, wherein said building a fraud information propagation network model from fraud information is specifically: constructing a state space of the user according to the phishing information, and assuming that the users in the network are nodes, each node i will be in a conversion of four states, including: susceptible state S, exposed state E, cheated state I, and recovered state R;
the probability vectors corresponding to the different states are represented as:
where t denotes the time, i denotes the user number,
respectively, the probabilities that the user i is in S, E, I, R four states at this time point, and the sum of the four probabilities is 1.
4. The method of suppressing the propagation of fraud information of claim 3, wherein there are defined conditions for said nodes to make a transition in four states, in particular: human in susceptible stateA fraud information target person, wherein the fraud information target person generates immunity to fraud information with a first probability after receiving fraud information, and the first probability is an immunity rate thetai(ii) a If the person in susceptible state does not generate immunity for the first time, the person is converted into exposure state with a second probability, wherein the second probability is infection rate ui(ii) a After the person in the exposure state receives the fraud information, the person enters the fraud state with a third probability, wherein the third probability is the conversion rate xii(ii) a After the person in the cheated state is identified as being cheated, stopping loss according to a fourth probability, and entering a recovery state, wherein the fourth probability is the recovery rate
Persons in a recovering state are immune to fraud information;
the state transition process corresponding to the four states is as follows:
wherein ,θiFor the immunological rate, uiIs the infection rate, xiiIn order to achieve a high conversion rate,
for the recovery rate, t represents the time, i represents the user number,
respectively representing the probability that the user i is in S, E, I, R four states at the moment;
in the model, the linear upper bound corresponding to the change of the state E and the state I can be represented by a probability matrix L, and the specific expression is as follows:
wherein ,
denotes an identity matrix, T ═ diag ([ θ [ ])1,...,θN]) Matrix representing diagonal element as node immunity rate, U ═ diag ([ U ] s)1,...,uN]) Matrix representing off-diagonal element as node infection rate, F ═ diag ([ ξ [ ]1,...,ξN]) A matrix representing diagonal elements as nodal conversion, D ═ diag ([ δ [ ]1,...,δN]) The diagonal elements are elements representing the node recovery rate;
the real part of the maximum eigenvalue of the matrix L represents the exponential growth rate of the probability of the population in these two states, denoted as λ (L); the goal to be optimized is to minimize the growth rate of the population in the E and I states, i.e., min (λ (L)).
5. The method of suppressing the propagation of fraud information of claim 1, wherein said anti-fraud social resources are obtained, said anti-fraud social resources comprising immune resources, protection resources, recovery resources, in particular: immune resource R1For increasing the immune rate thetaiProtection resource R2For reducing the infection rate uiAnd recovering the resource R3For increasing the recovery rate deltai；
By R ═ τri]Representing resource allocation matrices by τriWhen the resource matrix indicates that the r-th resource is allocated to the i-th node, the format of the resource matrix is:
wherein ,τriIndicating that the r resource is allocated to the ith node;
different resources have different costs, using c1(·)、c2(·)、c3(. represents resources R, respectively)1、R2、R3Unit cost of (2); the total cost is the sum of the costs of the various types of resources, expressed as:
wherein, i is 1,2, and N represents a node number, and c is used as1(·)、c2(·)、c3(. represents resources R, respectively)1、R2、R3Unit cost of (2).
6. The method of suppressing the propagation of fraud information of claim 5, wherein said optimizing said anti-fraud social resource results in a first priority resource, a second priority resource, and a third priority resource, specifically:
after initializing parameters and a population, dividing the sorted individuals into NG groups, wherein the higher the fitness value of the individual is, the higher the group is, namely the smaller the serial number of the group is, and otherwise, the lower the group is, namely the larger the serial number of the group is; the initialized parameters include: the size NP of the population, the number NG of the groups and the dimension D of the solution;
by NPgIndicates the total number of individuals in group g, in the pre-NG-1 group, NPgFloor (NP/NG); in the last group, NPg＝NP-(NG-1)×floor(NP/NG)；
The position and velocity of the r-th particle in the g-th group are respectively expressed as
And
wherein g 1,2, NG denotes the group in which the particles are located, and r 1,2, NPgThe number of the particle in the group is represented, D is 1,2, and D represents the dimension of the problem;
the individuals of the low-level group need to learn from two listed individuals randomly drawn from the high-level group and update the velocity Vg,r。
7. The method of suppressing the propagation of fraud information of claim 6, wherein said updating speed is specifically as follows:
assuming two sample individuals and located in the g ' and g ' groups, respectively, and (1 ≦ g ' ≦ g-1), the velocity update formula is:
wherein ,
which is indicative of the velocity of the particles,
denotes the position of the particle, r1 and r2Is two value ranges of [0,1 ]]W represents the degree of influence of the past speed on the current speed, c is a constant, NG represents the number of particle groups, g 'and g "represent the groups in which two sample particles are located, and r' and r" represent the intra-group numbers of the two sample particles.
8. The method of suppressing the propagation of fraud information of claim 7, further comprising: constructing a candidate solution for assisting in updating the location of the individual; the method comprises the following specific steps:
updating individual locations by means of priority planning and hierarchical learning:
h1 learning: setting a threshold parameter th ∈ [0,1 ]]Then filtering out the velocity V(g,r)The middle probability is larger than th, and the positions of the values are resources with first priority; obtaining an explicit set of the first priority resource according to the position of the first priority resource and by means of a 0 or 1 value randomly generated by the probability of the position in the velocity matrix, wherein the explicit set is used for constructing NEW _ X(g,r)The update rule is:
wherein ,
representing the matrix Cut _ V(g,r)The element in (1), S (-) is sigmoid function, tau represents filtering threshold, phi (-) is binary function, and the expression is:
wherein NEW _ X(g,r)Representing the new location, cost (-) represents a cost function, C represents a cost constraint, elem represents an element of the resource allocation matrix,
represents an add resource operation, i.e., changing a given element in a vector to 1;
h2 learning: from Pbest(g,r)Voting Gbest to obtain second priority resource and selection probability thereof, and passing through second priority resource Cut _ X(g,r)To construct NEW _ X(g,r)；Cut_X(g,r) and NEW_X(g,r)The update formula of (2) is as follows:
wherein 
Denotes Cut _ X(g,r)The element in (1), rand (0,1) represents a random number between 0 and 1, Sigmoid (·) represents a Sigmoid function,
representing Gbest and Pbest(g,r)The elements of the matrix obtained after the voting,
representing the matrix NEW _ X(g,r)The elements of (1);
h3 learning: let vector Other _ X(g,r)Represents NEW _ X(g,r)The opposite vector is that the 0 value and the 1 value at each position in the vector are interchanged to obtain the third priority resource; add not in NEW _ X without violating constraints(g,r)To construct NEW _ X from random resources in (2)(g,r)，NEW_X(g,r)The update formula is:
wherein ,
is the matrix Other _ X(g,r)The elements (A) and (B) in (B),
representing the matrix NEW _ X(g,r)Of (1).
9. The method of suppressing the propagation of fraud information of claim 8, further comprising: updating individual best Pbest(g,r)And global optimal solution Gbest of the whole population; the method specifically comprises the following steps:
calculating the fitness values corresponding to all the individual positions, selecting the optimal one of the fitness values, and comparing the selected optimal one with the fitness value of the old global optimal position Gbest; if the former is better than the latter, Gbest is updated, otherwise it remains unchanged.
10. A system for suppressing the propagation of fraud information, comprising:
the network information analysis module is used for analyzing the network information and acquiring fraud information;
the model construction module is used for constructing a fraud information propagation network model according to the fraud information;
the anti-fraud social resource acquisition module is used for acquiring anti-fraud social resources, wherein the anti-fraud social resources comprise immune resources, protection resources and recovery resources;
the resource optimization module is used for optimizing the anti-fraud social resources to obtain a first priority resource, a second priority resource and a third priority resource;
and the node notification module is used for notifying the first priority resource, the second priority resource and the third priority resource of the nodes through a fraud information propagation network model so as to further inhibit fraud information propagation.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111658983.4A CN114363464B (en) | 2021-12-30 | 2021-12-30 | Method and system for preventing fraud information from spreading |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111658983.4A CN114363464B (en) | 2021-12-30 | 2021-12-30 | Method and system for preventing fraud information from spreading |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114363464A true CN114363464A (en) | 2022-04-15 |
| CN114363464B CN114363464B (en) | 2023-04-21 |
Family
ID=81104835
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111658983.4A Active CN114363464B (en) | 2021-12-30 | 2021-12-30 | Method and system for preventing fraud information from spreading |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114363464B (en) |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070299777A1 (en) * | 2004-05-02 | 2007-12-27 | Markmonitor, Inc. | Online fraud solution |
| US20130138428A1 (en) * | 2010-01-07 | 2013-05-30 | The Trustees Of The Stevens Institute Of Technology | Systems and methods for automatically detecting deception in human communications expressed in digital form |
| CN107665256A (en) * | 2017-09-30 | 2018-02-06 | 陕西师范大学 | A kind of method that fraud information spreads in suppression mobile social networking |
| US10204374B1 (en) * | 2015-06-15 | 2019-02-12 | Amazon Technologies, Inc. | Parallel fraud check |
| FR3084946A1 (en) * | 2018-08-07 | 2020-02-14 | Amadeus S.A.S. | AUTOMATIC LEARNING METHODS AND SYSTEMS FOR ASSIGNED SEQUENCES |
| CN111737318A (en) * | 2020-06-24 | 2020-10-02 | 国家计算机网络与信息安全管理中心浙江分中心 | Screening method for phishing susceptible population |
| CN111814251A (en) * | 2020-07-08 | 2020-10-23 | 北京航空航天大学 | Multi-target multi-modal particle swarm optimization method based on Bayesian adaptive resonance |
| CN112579864A (en) * | 2020-12-25 | 2021-03-30 | 北京明朝万达科技股份有限公司 | Anti-fraud processing method and device |
| WO2021226878A1 (en) * | 2020-05-13 | 2021-11-18 | Paypal, Inc. | Using machine learning to mitigate electronic attacks |
-
2021
- 2021-12-30 CN CN202111658983.4A patent/CN114363464B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070299777A1 (en) * | 2004-05-02 | 2007-12-27 | Markmonitor, Inc. | Online fraud solution |
| US20130138428A1 (en) * | 2010-01-07 | 2013-05-30 | The Trustees Of The Stevens Institute Of Technology | Systems and methods for automatically detecting deception in human communications expressed in digital form |
| US10204374B1 (en) * | 2015-06-15 | 2019-02-12 | Amazon Technologies, Inc. | Parallel fraud check |
| CN107665256A (en) * | 2017-09-30 | 2018-02-06 | 陕西师范大学 | A kind of method that fraud information spreads in suppression mobile social networking |
| FR3084946A1 (en) * | 2018-08-07 | 2020-02-14 | Amadeus S.A.S. | AUTOMATIC LEARNING METHODS AND SYSTEMS FOR ASSIGNED SEQUENCES |
| WO2021226878A1 (en) * | 2020-05-13 | 2021-11-18 | Paypal, Inc. | Using machine learning to mitigate electronic attacks |
| CN111737318A (en) * | 2020-06-24 | 2020-10-02 | 国家计算机网络与信息安全管理中心浙江分中心 | Screening method for phishing susceptible population |
| CN111814251A (en) * | 2020-07-08 | 2020-10-23 | 北京航空航天大学 | Multi-target multi-modal particle swarm optimization method based on Bayesian adaptive resonance |
| CN112579864A (en) * | 2020-12-25 | 2021-03-30 | 北京明朝万达科技股份有限公司 | Anti-fraud processing method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114363464B (en) | 2023-04-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Liu et al. | Privacy-preserving traffic flow prediction: A federated learning approach | |
| CN112949837B (en) | Target recognition federal deep learning method based on trusted network | |
| CN108921051B (en) | Pedestrian attribute identification network and technology based on cyclic neural network attention model | |
| CN111339818B (en) | Face multi-attribute recognition system | |
| CN113283902B (en) | Multichannel blockchain phishing node detection method based on graphic neural network | |
| Yin et al. | GANs based density distribution privacy-preservation on mobility data | |
| CN112101404A (en) | Image classification method and system based on generation countermeasure network and electronic equipment | |
| CN112101403A (en) | Method and system for classification based on federate sample network model and electronic equipment | |
| CN109948649A (en) | The softward interview behavioral data character representation method of data-oriented opening and shares | |
| Zhang et al. | A hybrid transfer learning approach to migratable disaster assessment in social media sensing | |
| CN111259264B (en) | Time sequence scoring prediction method based on generation countermeasure network | |
| CN116227624A (en) | Federal knowledge distillation method and system oriented to heterogeneous model | |
| Choe et al. | Neural network schemes for detecting rare events in human genomic DNA | |
| CN115687758A (en) | User classification model training method and user detection method | |
| CN109978013B (en) | Deep clustering method for character action recognition | |
| CN112085051A (en) | Image classification method and system based on weighted voting and electronic equipment | |
| CN114363464A (en) | Method and system for suppressing fraud information propagation | |
| Mukhopadhyay et al. | Context-Aware Design of Cyber-Physical Human Systems (CPHS) | |
| CN116306969A (en) | Federal learning method and system based on self-supervision learning | |
| CN114358177B (en) | Unknown network traffic classification method and system based on multidimensional feature compact decision boundary | |
| Zou et al. | A neuro‐fuzzy crime prediction model based on video analysis | |
| CN116187469A (en) | Client member reasoning attack method based on federal distillation learning framework | |
| Chen et al. | Knowledge-inspired Subdomain Adaptation for Cross-Domain Knowledge Transfer | |
| CN115759234A (en) | Rumor source detection method based on space-time convolution neural network | |
| Wei et al. | CrimeSTC: A deep spatial-temporal-categorical network for citywide crime prediction |
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 |