CN110851660B - Immune retrospective rumor-refuting method based on rumor propagation model in social network - Google Patents

Abstract

The invention relates to an immune backtracking and rumor splitting method based on a rumor propagation model in a social network, which comprises the following steps: establishing a node propagation model comprising S, C, D, I states and R states, and performing state conversion on each state under certain conversion factors; an excitation mechanism is introduced in the state conversion process, and rumor propagation is inhibited; the conversion factors comprise personal factors, neighbor factors, content factors and time delay factors; wherein S represents untouched information, C represents dissemination of the information, D represents suspicion of the information, I represents no interest in the information and no dissemination, and R represents a public rumor issued by a dissemination authority. According to the method, personal factors, neighbor factors, content factors and time delay factors of each user are considered in the state transition probability calculation, and the model is closer to the reality; the introduction of the incentive mechanism enables users to send correct rumor-seeking information, so that rumors can be discovered and suppressed earlier.

Description

Immune retrospective rumor-refuting method based on rumor propagation model in social network

technical field

The invention belongs to the field of information data security in a large-scale online social network, and particularly relates to an immune backtracking and rumor-refuting method based on a rumor propagation model in a social network.

Background technique

With the development of society and technology, social network users are getting more and more and younger. The amount of information transmitted every day on social networks is huge, and among them, there are many rumors. It is easy for young users to believe rumours. Therefore, the harm caused by the spread of rumors in the network is also increasing.

At present, the dynamic characteristics of rumor propagation are mainly studied by the use of infectious disease models. Based on this, researchers have proposed many improved SIR (Susceptible-Infected-Recovery) models. For example: susceptible-hesitated-infected-removed (SHIR), susceptible-known-infected-removed (SKIR), etc. These methods utilize mean-field methods to solve differential equations to obtain thresholds for rumor propagation. However, these methods do not take into account the differences between users in the same category when classifying users.

In the face of the massive spread of rumors, the current control methods are divided into three categories. One is to control the influential nodes in the network. However, when rumors break out, accurate and rapid control of influential nodes that are about to be infected can effectively suppress the spread of rumors. The second method is to control the path of rumors spreading. This method needs to quickly locate the nodes where rumors spread, and then cut off the way for these nodes to spread rumors. However, in the large-scale spread of rumors on social networks, the spread of rumors is very fast and not easy to be discovered. Even if it is discovered, it has already spread to a certain scale. At this time, the first and second methods to control the rumors are too expensive. The third method is to spread rumors. Once users know the truth, they are immune to rumors. However, the current method is too long in response time, passive, and cannot quickly transmit rumor-defying information to infected users.

SUMMARY OF THE INVENTION

The purpose of the present invention is to overcome the deficiencies of the prior art, provide an immune retrospective refutation method based on a rumor propagation model in a social network, and propose an infectious disease model of a rumor propagation mechanism, which comprehensively considers personal factors, neighbor factors, and content factors. and delay factors; meanwhile, the rumors are controlled by an incentive-based immune backtracking and refuting method; the method of the invention can be applied to the information security and public opinion control mechanism in large-scale online networks.

In order to achieve the above object, the technical scheme of the present invention is:

An immune backtracking and rumor-refuting method based on a rumor propagation model in a social network, comprising: establishing a node propagation model including five states of S, C, D, I and R, and each state performs state transformation under certain transformation factors; state transformation In the process, an incentive mechanism is introduced to suppress the spread of rumors; the conversion factors include personal factors, neighbor factors, content factors and delay factors; among them, S means untouched information, C means spreading this information, D means doubting this information, I means If you are not interested in the information, it will not be disseminated. R means dissemination of rumors issued by authorities.

Preferably, an incentive mechanism is introduced in the state transition process to suppress the spread of rumors, including:

When there is a node that moves the R state during the state transition process, add 1 to the node propagation time;

When the node propagation time is greater than the time delay of the node, the source of the rumor is searched according to the link between the nodes; if there are multiple rumor sources, the node with the smallest delay among the multiple rumor sources is selected as the rumor source node; and the rumor source node add 1 to the propagation time;

When the propagation time of the rumor source node is greater than the time delay of the rumor source node, the node in the moving R state will pass the rumor refutation message to the rumor source node; at the same time, the state of the node in the moving R state is changed to R; The state changes to move the R state node.

Preferably, the time delay is based on a time delay factor, representing the time from receiving information to processing information for different users; specifically, the computer generates data within 0 to 10 that obeys a normal distribution.

Preferably, the transition probability of each state is as follows:

分别表示i节点的状态由S变成C、I和D的概率；

表示i节点收到辟谣消息后的状态由S变成I的概率；

分别表示i节点的状态由C变成R和I的概率；

分别表示i节点的状态由D变成R、C和I的概率；其中，

和

是激励导致的转态转化；ξ为激励系数，表示激励的代价程度；β是一个参数，属于0到1之间；

表示i节点收到谣言的个人因素；

表示i节点收到辟谣的个人因素；

表示邻居因素；

表示内容因素；d_i表示i节点的节点度；d_avg表示网络平均节点度；d_max表示网络最大节点度；

是时延因素，表示不同用户从收到信息到处理信息的不同时间。in,

Represent the probability that the state of the i node changes from S to C, I and D, respectively;

Indicates the probability that the state of the i-node after receiving the rumor-refuting message changes from S to I;

Represent the probability that the state of the i node changes from C to R and I, respectively;

respectively represent the probability that the state of the i node changes from D to R, C and I; among them,

and

is the transformation caused by excitation; ξ is the excitation coefficient, indicating the cost of the excitation; β is a parameter, belonging to between 0 and 1;

Indicates the personal factor of the i-node receiving the rumor;

Indicates the personal factor that i-node received to refute rumors;

represents the neighbor factor;

Represents the content factor; d _i represents the node degree of the i node; d _avg represents the average node degree of the network; d _max represents the maximum node degree of the network;

is the delay factor, which represents the different time from receiving information to processing information for different users.

Preferably, the personal factors of i-node receiving rumors are expressed as follows:

Among them, i∈[1,N], N represents the number of nodes;

The personal factors that i-node received for refuting rumors are expressed as follows:

Among them, α is a parameter preset according to the network structure;

The neighbor factor is expressed as follows:

是邻居因素，表示邻节点对用户的影响；avg(·)表示求平均值；max(·)表示求最大值，k表示用户的邻节点中传谣的序号；d_j表示k节点的节点度。in,

is the neighbor factor, which represents the influence of neighbor nodes on the user; avg( ) represents the average value; max( ) represents the maximum value, k represents the sequence number of rumors in the user's neighbor nodes; d _j represents the node degree of the k node .

The content factors are expressed as follows:

表示内容因素，不同的用户会对不同的信息感兴趣。in,

Indicates content factors, different users will be interested in different information.

After adopting the above scheme, the beneficial effects of the present invention are:

(1) An immune retrospective rumor refuting method based on a rumor propagation model in a social network of the present invention proposes a rumor propagation model, which comprehensively considers each user's personal factors, neighbor factors, content factors and delay factors, and accordingly Calculate the user state transition probability; for the method of disseminating rumor-refuting information to control rumors, an incentive mechanism is introduced to enable users to actively participate in the rumor-refuting process, and an immune retrospective rumor-refuting method is proposed to effectively spread the rumor-refuting information to users who believe in rumors; through the method of the present invention, Once users are exposed to rumors and accept incentives, they can start to control the rumors, and the information sent to refute rumors can also be effectively spread on the Internet, so that users who have not received rumors can be immune to rumors, and users who have spread and believed rumors will wake up ;

(2) The method of the present invention can solve the situation of multiple rumors spreading. Simulation results show that our method results in fewer people spreading rumors and more people immune to it.

The present invention will be further described in detail below with reference to the accompanying drawings and embodiments, but the method for immune backtracking and refuting rumors based on a rumor propagation model in a social network of the present invention is not limited to the embodiments.

Description of drawings

FIG. 1 is a transformation relationship diagram of nodes between different transition states according to an embodiment of the present invention;

Fig. 2 is the situation of different node degrees of rumor sources according to the embodiment of the present invention; wherein, (a) indicates that the node degree is 1; (b) indicates that the node degree is an average degree of 11; (c) indicates that the node degree is a maximum degree of 223;

Fig. 3 is a situation where the number of rumors is different in the embodiment of the present invention; wherein (a) represents the number of rumors of two; (b) represents the number of rumors of five;

Figure 4 is a comparison of the immunity backtracking method according to the embodiment of the present invention, acquaintance immunity and target immunity; wherein (a) represents the number of infected nodes; (b) represents the number of immune nodes.

Detailed ways

The technical solutions in the embodiments of the present invention will be described and discussed in detail below with reference to the accompanying drawings of the present invention. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

The present invention is an immune backtracking and rumor-refuting method based on a rumor propagation model in a social network, which divides users into five categories: susceptible, infected, suspected, immune, and recovered. A Susceptible-Contagious-Doubt-Immune-Recoverable (SCDIR) model is established to study the characteristics of rumor propagation. However, these counter-rumors are not all correct. The counter-rumors sent by users are regarded as rumors, and the counter-rumors released by the government or authorities are regarded as rumor-refuting information. Then, an incentive mechanism is introduced to provide rewards for immune users and suspicious users to contact the official to become a recovery node when they receive rumors, and then spread counter-rumors. Due to the high interest of suspected users, a special recovery node with mobility will be generated, and then under the guidance of the immune backtracking method to refute rumors, it will carry out mobile propagation of counter-rumors.

Specifically, the methods include:

Step 1): Divide users into 5 types according to their different responses to information, and establish a SCDIR propagation model. Among the 5 states, S means no contact with the information, C means dissemination of the information, D means to doubt the information, I means not interested in the information and will not spread it, and R means the refutation of rumors issued by the dissemination authority. The transformation relationship between different states is shown in Figure 1.

Step 2): State transition factors. According to Fig. 1, each arrow represents a transformation relationship, indicating that there is a transformation probability. According to graph theory, N users in a social network are regarded as node sets V={v ₁ ,...,v _N }, i∈[1,N], the value of v _i represents the state, and the initial state is S. The friend relationship between nodes is regarded as an edge matrix E, which constitutes a graph G={V,E}. The factors of node transformation are divided into personal factors, neighbor factors, content factors and delay factors. They are defined as follows:

是i节点收到谣言的个人因素，

是i节点收到辟谣的个人因素，对于影响力大的节点，更倾向于发布辟谣而不是谣言，因此个人因素分成两种，表示节点影响力对信息传播的影响。avg(·)和max(·)分别表示求平均值和求最大值，k是用户的邻节点中传谣的序号。

是邻居因素，表示邻节点对用户的影响。d_avg是网络平均度。α是一个参数，根据网络结构调整，使得(d_avg/d_max)α的值接近0.5。f_c ⁱ是内容因素，不同的用户会对不同的信息感兴趣。

是时延因素，表示不同用户从收到信息到处理信息的不同时间。具体值通过计算机生成0到10内，服从正态分布的数据。where d _i is the node degree of node i, d _max is the maximum node degree of the network,

is the personal factor for the i-node to receive rumors,

It is the personal factor of the i-node receiving rumors. For nodes with great influence, it is more inclined to publish rumors rather than rumors. Therefore, personal factors are divided into two types, indicating the influence of node influence on information dissemination. avg( ) and max( ) represent the average and maximum values, respectively, and k is the sequence number of rumors in the user's neighbors.

is the neighbor factor, which represents the influence of neighbor nodes on the user. d _avg is the network average degree. α is a parameter adjusted according to the network structure such that the value of (d _avg /d _max )α is close to 0.5. _fci ^is the content factor, different users will be interested in different information.

is the delay factor, which indicates the different time from receiving the information to processing the information for different users. The specific value is generated by the computer from 0 to 10, and the data obeys the normal distribution.

Step 3): The mutual transformation process of the 5 states of the SCDIR model. According to the relation of accompanying drawing 1 and the conversion factor of step 2), calculate the state transition probability as follows:

表示i节点的状态由S变成C的概率。I′表示收到辟谣转化成I状态。

和

是激励导致的转态转化，其中ξ是激励系数，表示激励的代价程度。β是一个参数，属于0到1之间。SCDIR模型的算法描述如下：in,

Indicates the probability that the state of the i-node changes from S to C. I' means that the refuted rumor is received and transformed into the I state.

and

is the transformation caused by the incentive, where ξ is the incentive coefficient, which represents the cost of the incentive. β is a parameter that falls between 0 and 1. The algorithm of the SCDIR model is described as follows:

Step 4): When a node that moves the R state appears in the state transition process, it indicates that the node with the D state is excited. Node D knows who the source of the news is because of doubts about the received information. Therefore, the motivated mobile R node needs to pass the rumor-refuting message to the person who spread the rumor, and force its state to be changed, thereby suppressing the spread of rumors. In the same way, the person who spread the rumor also knows the source of the rumor, and so on, backtracking until the source receives the rumor. Algorithm 2 describes this process.

Step 5): Referring to Figures 2 to 4, simulations were performed on the AS-level Internet dataset and compared with target immunization (TI) and acquaintance immunization (AI). In Figures 2 and 3, S is a susceptible node (users who have not been exposed to information), I is an immune node (users who are not interested in information and will not spread it), and R is a rumor-refuting node (a rumor-refuting node published by an authority) user), C is a rumor node (user who spread rumors), D is a suspect node (doubt about information), and M is a mobile node (a mobile rumor-refuting node). The information here refers to rumors and rumors. In Figure 4, NI refers to no immunization method, TI refers to target immunization, AI refers to acquaintance immunization, 0.1, 0.3, 0.5, 0.7, and 0.9 refer to the proposed retrospective immunization method, and the incentive coefficients are Simulation plots of 0.1, 0.3, 0.5, 0.7, 0.9.

The above is only a preferred embodiment in the example of the present invention. However, the present invention is not limited to the above-mentioned embodiments, and any equivalent changes and modifications made according to the present invention, when the resulting functional effects do not exceed the scope of this scheme, all belong to the protection scope of the present invention.

Claims (3)

1. An immune backtracking and rumor-refuting method based on a rumor-propagation model in a social network, comprising: establishing a node propagation model comprising five states of S, C, D, I and R, and each state is under certain transformation factors. Carry out state transformation; introduce an incentive mechanism in the process of state transformation to suppress the spread of rumors; the transformation factors include personal factors, neighbor factors, content factors and delay factors; where S means untouched information, C means dissemination of this information, D means If you suspect this information, I means that you are not interested in the information and will not spread it, and R means to spread the rumors issued by the authority; The transition probabilities for each state are as follows:

in,

Represent the probability that the state of the i node changes from S to C, I and D, respectively;

Indicates the probability that the state of the i-node after receiving the rumor-refuting message changes from S to I;

Represent the probability that the state of the i node changes from C to R and I, respectively;

respectively represent the probability that the state of the i-node changes from D to R, C and I; among them,

and

is the transformation caused by excitation; ξ is the excitation coefficient, indicating the cost of the excitation; β is a parameter, belonging to between 0 and 1;

Indicates the personal factor of the i-node receiving the rumor;

Indicates the personal factor that i-node received to refute rumors;

represents the neighbor factor;

Represents the content factor; d _i represents the node degree of the i node; d _avg represents the average node degree of the network; d _max represents the maximum node degree of the network;

It is a delay factor, which means that different users have different time from receiving information to processing information; The personal factors of i-node receiving rumors are expressed as follows:

Among them, i∈[1,N], N represents the number of nodes; The personal factors that i-node received for refuting rumors are expressed as follows:

Among them, α is a parameter preset according to the network structure; The neighbor factor is expressed as follows:

in,

is the neighbor factor, which represents the influence of neighbor nodes on the user; avg(•) represents the average value; max(•) represents the maximum value, k represents the sequence number of rumors in the user's neighbor nodes; d _j represents the node degree of the k node ; The content factors are expressed as follows:

in,

Indicates content factors, different users will be interested in different information. 2. The immune retrospective refuting rumor method based on the rumor propagation model in the social network according to claim 1, is characterized in that, introducing incentive mechanism in the state transition process, suppressing rumor propagation, specifically comprises: When there is a node that moves the R state during the state transition process, add 1 to the node propagation time; When the node propagation time is greater than the time delay of the node, the source of the rumor is searched according to the link between the nodes; if there are multiple rumor sources, the node with the smallest delay among the multiple rumor sources is selected as the rumor source node; and the rumor source node add 1 to the propagation time; When the propagation time of the rumor source node is greater than the time delay of the rumor source node, the node in the moving R state will pass the rumor refutation message to the rumor source node; at the same time, the state of the node in the moving R state is changed to R; The state changes to move the R state node. 3. The immune retrospective refutation method based on a rumor propagation model in a social network according to claim 2, wherein the time delay is based on a time delay factor, representing the time from receiving information to processing information by different users; The computer generates data from 0 to 10 that obey a normal distribution.

Images