False information tracing method for time-varying cross-platform coupling social network

Abstract

The invention relates to the technical field of information identification, in particular to a false information tracing method for a time-varying cross-platform coupling social network, which deeply studies the tracing problem of false information in the time-varying coupling network, and proposes a targeted time-varying coupling network false information tracing algorithm-RIDTCTem on the basis of a static tracing algorithm RIDTC. According to the algorithm, the position information of the time-varying network snapshots is creatively fused into the traceability process, the contribution rate of different snapshots to the traceability result is fully considered, and a large number of experimental verifications are carried out on a series of time-varying coupling networks with different scales. Results show that the RIDTCTem algorithm shows remarkable advantages in the aspects of traceability accuracy, reduction of source node error distance and the like, and compared with a comparison algorithm, the RIDTCTem algorithm is more prominent in adaptability and stability, so that the problem that the difficulty of traceability of false information in an existing traceability method is large is solved.

Description

Technical Field

[0001] The present invention relates to the field of information identification technology, and in particular to a method for tracing the source of false information in a time-varying, cross-platform coupled social network. Background Art

[0002] Time-varying topology snapshots refer to the topological structure data of social networks obtained at different time points, which can reflect the dynamic changes of network structure. Compared with static topology snapshots, time-varying topology snapshots are closer to the network status in the real world. Therefore, studying the problem of false information tracing under time-varying topology snapshots has important practical significance and application value.

[0003] In the study of false information tracing, the characteristics of time-varying networks make the problem more complicated. The false information tracing of time-varying networks is to try to find the original source of false information from these observation snapshots. However, due to the dynamic variability of the network, although there is an inherent evolutionary relationship between snapshots at different observation times, it is often difficult to obtain the true linkage process between each observation snapshot, which increases the difficulty of tracing. Summary of the invention

[0004] The purpose of the present invention is to provide a method for tracing the source of false information for time-varying cross-platform coupled social networks, aiming to solve the problem that it is difficult to trace the source of false information using existing tracing methods.

[0005] To achieve the above object, the present invention provides a method for tracing the source of false information in a time-varying cross-platform coupled social network, comprising the following steps:

[0006] Calculate the score value of each node being the source of propagation in each observation snapshot;

[0007] Apply the contribution rate of the observed snapshot to the score value of each node;

[0008] The comprehensive calculation node is the final score value of the propagation source;

[0009] Consider coupling strength and anonymity to determine the platform where the false information initially appeared.

[0010] The step of calculating the score value of each node being a propagation source in each observation snapshot includes:

[0011] For each platform, the static tracing algorithm is applied to each snapshot of the platform, and finally the score value of each node in the snapshot is obtained.

[0012] The step of applying the contribution rate of the observed snapshot to the score value of each node includes:

[0013] Calculating a contribution rate of each of the snapshots;

[0014] The contribution rate is applied to the score value of each of the nodes.

[0015] The comprehensive calculation node is the final score value of the propagation source, including:

[0016] Get the final score value based on the score vector;

[0017] For each platform, the node with the highest final score value is selected as the source of false information.

[0018] The consideration of coupling strength and anonymity to determine the platform where the false information first appeared includes:

[0019] Calculate the effect of coupling strength and anonymity value on node scores;

[0020] Updating the score value of the node;

[0021] The platform where the node with the highest score is located is selected as the platform where the false information initially appears.

[0022] The present invention discloses a false information tracing method for a time-varying cross-platform coupled social network, comprising the following steps: calculating the score value of each node as a propagation source in each observation snapshot; applying the contribution rate of the observation snapshot to the score value of each node; comprehensively calculating the final score value of the node as a propagation source; and considering the coupling strength and anonymity to determine the platform where the false information first appears. This method deeply studies the tracing problem of false information in a time-varying coupled network, and proposes a targeted time-varying coupled network false information tracing algorithm-RI_DTC_Tem on the basis of the static tracing algorithm RI_DTC. This algorithm is not only a simple extension of the static coupling network tracing algorithm, but also innovatively integrates the location information of the time-varying network snapshot into the tracing process, and fully considers the contribution rate of different snapshots to the tracing results. Through a large number of experimental verifications on a series of time-varying coupled networks of different sizes, the results show that the RI_DTC_Tem algorithm shows significant advantages in tracing accuracy and reducing the error distance of the source node. Compared with the comparison algorithm, its adaptability and stability are more prominent, thereby solving the problem that the existing tracing method is difficult to trace false information. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0024] Figure 1It is a flow chart of a false information tracing method for a time-varying cross-platform coupled social network provided by the present invention.

[0025] Figure 2 It is a time-varying snapshot tracing process diagram based on the RI_DTC_Tem algorithm.

[0026] Figure 3 It is a schematic diagram of the coupling network structure.

[0027] Figure 4 This is a comparison chart of the average error distance between the time-varying coupled network tracing RI_DTC_Tem algorithm and two comparison algorithms (LP, PSLM_Tem) in three real time-varying coupled networks.

[0028] Figure 5 It is a comparison graph of the average error distance surface of the time-varying coupled network tracing RI_DTC_em algorithm and the comparison algorithm (LP, PSLM_Tem) in the coupled network "BA1200-WS1000".

[0029] Figure 6 This is a graph showing the changing trends of the traceability accuracy of the three algorithms under different network change rates and coupling strengths in the "BA1200-WS1000" coupling network.

[0030] Figure 7 Under the parameter combination of coupling strength λ = 0.5 and network change rate μ = 10%, the traceability effect of each snapshot is statistically analyzed.

[0031] Figure 8 This is a graph showing the changes in average error distance and traceability accuracy of the source platform under different numbers of observation snapshots and coupling strengths when the network change rate μ=10%.

[0032] Fig. 9 and Fig.10 The accuracy of the “BA1200-WS1000” coupled network and the source platform estimation accuracy of the real time-varying coupled network are respectively counted. DETAILED DESCRIPTION

[0033] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present invention, and should not be construed as limiting the present invention.

[0034] See also Figures 1 to 10 The present invention provides a method for tracing the source of false information in a time-varying cross-platform coupled social network, comprising the following steps:

[0035] S1 calculates the score value of each node being a propagation source in each observation snapshot;

[0036] S11 applies the static tracing algorithm to each snapshot of each platform, and finally obtains the score value of each node in the snapshot;

[0037] Specifically, platform 1 counts the topology of t snapshots to detect the source of false information on the platform. First, the RI_DTC (static tracing) algorithm is applied to each snapshot to detect the information source of the topology snapshot in each statistical time period. After applying the RI_DTC algorithm in each observation snapshot, each node in its topology calculates a score value, which reflects the probability of the node being the source point in the observation snapshot.

[0038] RI_DTC (static tracing) algorithm process:

[0039] The algorithm uses a divide-and-conquer strategy, first separating the coupled network into multiple single-platform networks, and then tracing each single-platform network separately.

[0040] For each single-platform network, monitoring nodes are set up, and the symbol α can be used to represent the proportion of monitoring nodes to the entire network nodes. These monitoring nodes will record the time when they forward or publish false information during the spread of false information (this time is called "absolute I state time"). The absolute I state time of any monitoring node v can be represented by the symbol t v If there are k monitoring nodes, the absolute I-state time of k monitoring nodes can be expressed as T = {t1, t2, ..., t k In the process of tracing the source of each single-platform network, we only know the time when these monitoring nodes forwarded / published false information and the topology of the entire network. First, we need to convert the absolute I state time t v Convert to "relative I-state time" (symbol t' v The process is implemented by the following formula:

[0041] t′ v =max(T)-t v +1

[0042] Next, initialize the label information of each monitoring node v, which are three variable elements - id, t′ v and d v Then the label information of each monitoring node v can be represented by a triple O v ={id,t′ v ,d v}; where id represents the number of the monitoring node itself, t′ v Indicates the relative I state time of the monitoring node, dv Indicates the number of edges that the monitoring node passes through during the tracing process.

[0043] Next, set up an iterative process and let the monitoring node start broadcasting its label information to neighboring nodes. Use the variable iter to represent the iteration round, and iter is initialized to 1. In each iteration iter, let the relative I state time t v The monitoring node v whose ′ is equal to iter broadcasts label information O to its neighbors v ; If the neighbor node receives the label information for the first time, save it and update the distance variable d of the monitoring node v =d v +1, indicating that the monitoring node v reaches itself through an edge; in the next iteration, the relative I state time t v The monitoring node v whose ′ is equal to iter and the node that received the label information in the previous iteration simultaneously broadcast the label information they own to the surrounding neighbor nodes.

[0044] In each iteration iter, if there is a node v that receives the label information of all monitoring nodes, then the distance vector from the node v to the k monitoring nodes can be expressed as D = {d1, d2, ..., d k}. Finally, the comprehensive score formula for calculating whether the node v is a source of false information is:

[0045]

[0046] In the formula, s v represents the comprehensive score of a node v, θ represents the formula balance factor, and Pearson (D, T) is used to calculate the similarity between vector D and vector T. The calculation formula is:

[0047]

[0048] In the formula, and are the average values ​​of D and T respectively, and k represents the number of monitoring nodes.

[0049] In the t-th observation snapshot of each single platform, after the above RI_DTC (static traceability) algorithm, the score value s of each node v in the t-th observation snapshot of the platform will be obtained. v (t). For example, in the t-1th observation snapshot, the score value of any node v is represented by s v (t-1), and so on for the score values ​​in other snapshots.

[0050] S2 applies the contribution rate of the observed snapshot to the score value of each node;

[0051] S21 calculates the contribution rate of each snapshot;

[0052] Specifically, the importance of a snapshot is determined by the time it appears, i.e., the later the snapshot, the higher its contribution to the traceability result. This solution comprehensively considers the information of different snapshots to improve the accuracy and reliability of traceability.

[0053] The formula for calculating the weight factor of the observation snapshot at time t is as follows:

[0054]

[0055] The above formula is used to automatically assign weights to each observation snapshot. From this formula, we can see that the larger the observation time t, the larger the corresponding weight factor θ t That is, the later the snapshot is, the greater its contribution to the detection of the source point.

[0056] S22 applies the contribution rate to the score value of each of the nodes.

[0057] Specifically, the weight factor of the snapshot observed at that moment is applied to the traceability score of each node, which can be expressed as follows:

[0058] s v (i)←θ i *s v (i),(i=t,t-1,...,1)

[0059] In the formula, θ i represents the weight of the i-th snapshot, s v (i) represents the score value of node v in the i-th observation snapshot.

[0060] The S3 comprehensive calculation node is the final score value of the propagation source;

[0061] S31 obtains a final score value based on the score vector;

[0062] Specifically, after tracing back t observation snapshots, a node v will have a t-dimensional score vector s v ={s v (t),s v (t-1),...,s v (1)}. It should be noted that each node does not necessarily appear in all observation snapshots. For example, node v appears in observation snapshot t3, but in observation snapshot t2, the node may not exist in the topology. Therefore, the score vector s of node v is v The s v(2) There is no value. For this situation, this method first fills it with 0 value, and then replaces it according to the global average probability. This strategy can be understood as follows: if a node is considered as a source node in multiple snapshots, even if it is offline in some snapshots, the probability that it is the final source point will be greater. For the t-dimensional score vector of node v, its average value is finally used as its final score value:

[0063] s v =average(s v (t),s v (t-1),...,s v (1))

[0064] It should be noted that among the t observed snapshots obtained, the nodes in the I state in the first snapshot topology are not the only source of false information. Instead, it is the snapshot topology obtained by the first statistical analysis after the false information has been propagated for a certain period of time.

[0065] S32 selects, for each platform, the node with the highest final score value as the source of false information.

[0066] Specifically, due to the differences in network topology between different snapshots in a time-varying network, the estimated source of each snapshot can only be regarded as a candidate set of the true source points of the platform. Therefore, the nodes of each platform are sorted from large to small according to the score size, and finally the node with the largest score value is selected as the estimated value of the false information source of the platform:

[0067]

[0068] In the formula, agrmax() refers to selecting the position index corresponding to the maximum value in the brackets, that is, the node number; n refers to the maximum node number value.

[0069] S4 considers coupling strength and anonymity to determine the platform where false information initially appears.

[0070] S41 calculates the effect of coupling strength and anonymity value on node scores;

[0071] Specifically, the coupling strength λ between platforms is set, which represents the degree of interaction between platforms; an anonymity value γ is set for each coupling edge, which represents the visibility between the node pairs of two platforms. The higher the degree of interaction between platforms, the faster the false information spreads between multiple platforms. The higher the node anonymity, the more likely the user is to publish false information on another platform. In the cross-platform coupled social network false information propagation model, whether a node publishes false information on another platform, this method considers the influence of coupling strength and anonymity value at the same time. It is believed that if the coupling strength and anonymity value are higher, it will be more helpful for the node to publish false information on another platform. Based on the same idea, after calculating the score value of a node belonging to the source of false information, the coupling strength (λ) and anonymity value (γ) are reacted to the score value and added to the score of its mapping node on another platform, thereby effectively combining the two platforms in the traceability process.

[0072] The coupling strength (λ) and the anonymity value (γ) are expressed by β, and the calculation process is:

[0073] β=1-(1-r(0,λ))(1-r(0,γ))

[0074] In the formula, r(a,b) represents a random value selected from the uniform distribution from a to b. Since the coupling strength and node anonymity between real social platforms are difficult to determine, this method adopts a sampling method within a range to quantify these two values. If the coupling strength between platforms is stronger and the node anonymity is higher, the sampling result will be larger accordingly.

[0075] S42 updates the score value of the node;

[0076] Specifically, after the coupling strength (λ) and the anonymity value (γ), the score value of node v is updated as follows:

[0077] s v ←θ*β*s u +(1-θ)s v

[0078] In the formula, θ is the balance factor, which is set to 0.5. This formula means that for node u, if it has a corresponding mapping node v on another platform, its score value will be applied to the node v of the other platform according to the anonymous value.

[0079] S43 selects the platform where the node with the highest score value is located as the platform where the false information initially appears.

[0080] Specifically, each node of each platform will eventually have a score value. These nodes will be sorted from large to small according to the score value, and finally the platform where the node with the largest score value is located will be selected as the estimated value of the platform where the false information first appeared:

[0081] plat s = argmax(s1,s2,...,s n )

[0082] In the formula, agrmax() refers to selecting the position index corresponding to the maximum value in the brackets, that is, the node number; n refers to the maximum node number value.

[0083] The following is the verification of this application:

[0084] It is particularly noted that the experimental data used in this embodiment is only used for the purpose of demonstrating the results of the present invention and is not considered to be a limitation of the present invention. According to the specific settings required by the actual scenario, no specific limitations are made in this application.

[0085] In this embodiment, 8 subnetworks are used to form 4 groups of coupled networks, as shown in Table 1. Among them, the minimum number of coupled network nodes and edges is (306, 4110), and the maximum number of coupled network nodes and edges is (9135, 37957). The coupled network "BA1200-WS1000" is two synthetic networks. This embodiment sets a set of network change parameters for the coupled network, represented by the symbol μ, and the parameter setting range is μ = [10%, 15%, 20%, 25%, 30%]. The HG_2011 and HG_2012 data sets are contact time networks between high school students. HG_2011 gives the contact records of students in 3 classes within 4 days in December 2011, and HG_2012 gives the contact records of students in 5 classes within 7 days in November 2012. The Email_Eu data set is the information of sending and receiving emails between members of the research institution. The SFHH dataset is the interactive information of 405 participants in the conference (June 4-5, 2009). The ACMH2009 dataset is a dynamic network of about 110 conference participants collected during the conference for about 2.5 days. The BitcoinA dataset is a trust network composed of people who trade on the platform. This embodiment sets the anonymity of the highly identifiable edge γ = 0.2, and the anonymity of other edges γ = 0.8; the monitoring node ratio α = 30%; the coupling strength of the coupling network λ = [0, 0.2, 0.5, 0.8, 1.0]; the algorithms compared are the LP algorithm and the PSLM_Tem algorithm; the evaluation indicators of this embodiment are the average error distance and the traceability accuracy.

[0086] The calculation formula of the average error distance evaluation method is as follows:

[0087]

[0088] Where Ed represents the average error distance evaluation score, M represents the number of simulations of the embodiment, and d uv Represents the shortest path distance between node u and node v. The traceability accuracy evaluation method is to statistically average the proportion of the total number of simulations in which the suspected source node estimated by this method is the same as the real source node in M ​​rounds of simulation.

[0089] In this embodiment, the platform where the false information first appears is named the “source platform” and is symbolized by s Another platform that did not have any false information at the initial moment was named "non-source platform" and was symbolized by plat s* express.

[0090] The verification results of this embodiment are as follows Figures 5 to 10 As shown:

[0091] Figure 5 The average error distance comparison between the proposed time-varying coupled network tracing RI_DTC_Tem algorithm and two comparison algorithms (LP, PSLM_Tem) in three real time-varying coupled networks is shown. As can be seen from the figure, the algorithm of the present invention has obvious advantages over the comparison algorithms, especially in the source platform (plat s ) on non-source platforms. In the figure, although the average error distances of the three algorithms are relatively close, the algorithm of the present invention still obtains a relatively small average error distance.

[0092] Figure 6 The figure shows the average error distance surface comparison of the time-varying coupled network tracing RI_DTC_em algorithm proposed in the present invention and the comparison algorithm (LP, PSLM_Tem) in the coupled network "BA1200-WS1000". It can be seen from the figure that the RI_DTC_Tem algorithm proposed in the present invention shows obvious advantages in average error distance. In the worst case, when the network change rate μ=30% on a non-source platform, the average error value of the RI_DTC_Tem algorithm can also achieve an effect of less than 2 hops. In addition, it can be clearly seen that the network change rate has a significant impact on tracing. The higher the network change rate, the more difficult it is to observe or restore the information propagation process, and even the phenomenon of "fault" in information propagation may occur. The incompleteness of this network deepens the difficulty of tracing the current snapshot, making the average error distance on the network change rate axis change more strongly than the coupling strength axis.

[0093] Figure 7The variation trend of the traceability accuracy of the three algorithms under different network change rates and coupling strengths in the "BA1200-WS1000" coupling network is shown in detail. As can be seen from the figure, the traceability accuracy of the algorithm proposed in the present invention is higher than that of the two comparison algorithms in both the source platform and the non-source platform.

[0094] Figure 8 It means that under the parameter combination of coupling strength λ=0.5 and network change rate μ=10%, the traceability effect of each snapshot is counted separately, and the average error distance and average traceability accuracy are used as measurement indicators to statistically observe the impact of snapshots on the traceability results. As can be seen from the figure, whether on the source platform or on the non-source platform, the average error distance shows a downward trend, while the average traceability accuracy shows an upward trend. The experimental results show that the method of assigning weights to different observation snapshots proposed in the present invention is effective and feasible, proving that the later the observation snapshot, the greater its contribution to traceability.

[0095] Fig. 9 The figure shows the changes in the average error distance and traceability accuracy of the source platform under different numbers of observation snapshots and coupling strengths when the network change rate μ=10%. As can be seen from the figure, as the number of snapshots increases, the average error distance shows an overall downward trend. Correspondingly, the traceability accuracy has an upward trend. This shows that by increasing the number of observation snapshots, the traceability effect can be effectively increased. The reason is that due to the dynamic characteristics of the time-varying network, a single observation snapshot may not provide enough information about the propagation process, while multiple snapshots can provide the required data for traceability from different angles.

[0096] Fig. 9 as well as Fig.10 The source platform estimation accuracy of the "BA1200-WS1000" coupling network and the real time-varying coupling network is counted respectively, that is, step S4 is performed to consider the coupling strength and anonymity to determine the platform where the false information first appeared. Fig. 9 In the coupled network "BA1200-WS1000" synthesized in the paper, the accuracy of source platform estimation decreases as the network change rate increases. In terms of coupling strength, low coupling strength means that the synchronization speed of nodes between platforms is slower, and the contribution rate of nodes in one platform to whether an account on another platform publishes false information is reduced, which makes the two platforms relatively more independent. When the coupling strength is high, the state changes between the node pairs of the two platforms tend to be consistent, and the propagation of false information is similar to the propagation process of a single platform. Fig.10In the real time-varying coupled network, as the coupling strength increases, the source platform estimation accuracy of each time-varying coupled network also shows an upward trend. Unlike the synthesized "BA1200-WS1000" coupled network, due to the irregularity of the evolution of the real time-varying network, its source platform estimation accuracy does not strictly increase with the change of coupling strength, and there is a certain volatility. However, judging from the statistical results of the synthesized coupling network and the real time-varying network, step S4 proposed in the present invention has certain feasibility.

[0097] Table 1 Network data parameters used in the embodiment of the present invention

[0098]

[0099]

[0100] This method deeply studies the problem of false information tracing in time-varying coupled networks, and proposes a targeted false information tracing algorithm for time-varying coupled networks, RI_DTC_Tem, based on the static tracing algorithm RI_DTC. This algorithm is not only a simple extension of the static coupled network tracing algorithm, but also innovatively integrates the location information of the time-varying network snapshot into the tracing process, and fully considers the contribution rate of different snapshots to the tracing results. Through a large number of experimental verifications on a series of time-varying coupled networks of different sizes, the results show that the RI_DTC_Tem algorithm has significant advantages in tracing accuracy and reducing the error distance of the source node. Compared with the comparison algorithm, its adaptability and stability are more prominent. In addition, the experiment also explores the impact of the number of tracing snapshots on the tracing results, and finds that appropriately increasing the number of snapshots can improve the tracing effect to a certain extent. In order to further understand the impact of network structure changes on tracing performance, the experiment uses a synthetic coupled network and observes the tracing results under different network change rates. The experiment finds that with the increase of the network change rate, the false information propagation process becomes more complicated and chaotic, resulting in a significant decrease in the tracing effect. On this basis, this method further expands the tracing task by introducing network coupling strength and node anonymity as estimation factors of node probability, trying to determine the platform where the false information first appeared. The feasibility and effectiveness of this extended task are demonstrated through experimental verification in synthetic and real time-varying networks. This extension not only enriches the research content of false information tracing, but also provides a useful reference for tracing the source of false information in practical applications.

[0101] In summary, considering the complex and changeable characteristics of time-varying networks, the research of this method can provide the following valuable reference suggestions for the false information tracing work in time-varying coupled social networks:

[0102] First, snapshots at different observation times have different contribution rates to traceability, and this contribution rate is closely related to the sampling time of the snapshot. Specifically, the wider the spread of false information, the more traceability information it provides, and the greater its influence on traceability. The automatic snapshot weight allocation method proposed in this method provides researchers with a new perspective, which helps to further explore the differences in contribution rates between snapshots and thus optimize the performance of the traceability algorithm.

[0103] Secondly, the number of traceability snapshots also has a significant impact on the traceability results. By increasing the number of traceability snapshots, we can more comprehensively understand the dynamic changes of the network structure, thereby improving the accuracy of the traceability results to a certain extent. Therefore, in practical applications, the number of traceability snapshots should be reasonably adjusted according to the specific situation to obtain the best traceability effect.

[0104] In addition, the rate of change between snapshots is also a key factor affecting the traceability effect. As the rate of change of the network increases, the spread of false information becomes more complicated and chaotic, bringing greater challenges to the traceability work. Therefore, when designing and implementing the traceability algorithm, the impact of the network change rate should be fully considered to improve the adaptability and stability of the traceability algorithm.

[0105] Finally, the idea of ​​applying network coupling strength and node anonymity to node scores proposed in this method provides a feasible reference solution to solve the problem of being unable to obtain the coupling relationship between two platforms in reality. The effectiveness of this solution has been verified through experiments. In actual operation, the weights of network coupling strength and node anonymity can be adjusted according to specific circumstances to better meet the traceability needs in different scenarios.

[0106] Figure 2 Illustration of the time-varying snapshot tracing process based on the RI_DTC_Tem algorithm:

[0107] The figure shows the observed topological snapshots of a single-platform network in different time periods during the false information tracing process. If the node status and topological structure of the network in the time period are counted in "hours", the observation time t1 represents the statistical information of the first hour, the observation time t2 represents the statistical information of the second hour, and so on. The figure only shows the network information of three statistical time periods. In each statistical time period, the state of each node (including monitoring nodes), such as S state, I state or R state, and the topological structure of the network in the current statistical time period can be known. Moreover, the node status and network topological structure may change in different statistical time periods. In the tracing process, reverse tracing starts from the last statistical time period (such as the "t3 observation time" in the figure), and the RI_DTC (static tracing) algorithm is executed in each statistical time period to obtain the score value of each node in each statistical time period. As shown in the figure, after tracing back to the node Y for three observation times, three score values ​​are obtained, namely s, t3, t4, t5, t6, t7, t8, t9, t10, t11, t12, t13, t14, t15, t16, t17, t18, t19, t20, t21, t22, t23, t24, t25, t26, t27, t28, t29, t30, t31, t32, t33, t34, t35, t36, t37, t38, t39, t31, t39, t31, t33, t39, t34, t35, t36, t38, t39, t31, t33, t39, t34, t36Y (3)s Y (2) s Y (1). Next, calculate the contribution rate of each observation moment to the entire traceability process, and apply this contribution rate to the traceability score of the node at that observation moment. Finally, comprehensively calculate the traceability score of each node, and select the node with the largest score as the source of false information.

[0108] This process is based on the RI_DTC (static tracing) algorithm and combines statistical information from different time periods of the false information propagation process, so it is named RI_DTC_Tem algorithm ("Tem" is the abbreviation of the word "Temporal").

[0109] Figure 3 Description of coupling network structure diagram:

[0110] In the coupled social network model constructed, two subgraphs are defined, where G1 can represent the online social network and G2 represents the offline social network. c Represents a coupled network connecting G1 and G2. Some nodes in network G2 (such as node v5) maintain two accounts in network G1 (such as node v 10 、v 14 ). Some of the connections between these accounts are known and observable (such as node v 10 ), represented by a red dotted line; other connections may not be directly observable (such as node v 14 ), which has a certain degree of anonymity. In order to quantify this anonymity, this paper introduces the concept of anonymity, which is represented by the symbol (γ). The higher the anonymity, the lower the identifiability of the edge relationship and the higher the degree of user privacy protection. The intensity of information flow between the two platforms is represented by the coupling strength (represented by the symbol λ).

[0111] What is disclosed above is only a preferred embodiment of the method for tracing the source of false information in a time-varying, cross-platform coupled social network of the present invention. Of course, it cannot be used to limit the scope of rights of the present invention. Ordinary technicians in this field can understand that all or part of the processes of the above embodiments and equivalent changes made according to the claims of the present invention still fall within the scope of the invention.

Claims

1. A method for tracing the source of false information in time-varying cross-platform coupled social networks, characterized in that: The following steps are involved: Calculate the score value of each node being the source of propagation in each observation snapshot; Apply the contribution rate of the observed snapshot to the score value of each node; The comprehensive calculation node is the final score value of the propagation source; Consider coupling strength and anonymity to determine the platform where the false information initially appeared.

2. The false information tracing method for time-varying cross-platform coupled social networks as claimed in claim 1, characterized in that: The calculation of the score value of each node being a propagation source in each observation snapshot includes: For each platform, the static tracing algorithm is applied to each snapshot of the platform, and finally the score value of each node in the snapshot is obtained.

3. The method for tracing the source of false information for time-varying cross-platform coupled social networks as claimed in claim 2, characterized in that: The step of applying the contribution rate of the observed snapshot to the score value of each node includes: Calculating a contribution rate of each of the snapshots; The contribution rate is applied to the score value of each of the nodes.

4. The method for tracing the source of false information for time-varying cross-platform coupled social networks as claimed in claim 3, characterized in that: The comprehensive calculation node is the final score value of the propagation source, including: Get the final score value based on the score vector; For each platform, the node with the highest final score value is selected as the source of false information.

5. The method for tracing the source of false information for time-varying cross-platform coupled social networks as claimed in claim 4, characterized in that: The considerations for coupling strength and anonymity to determine the platform where the false information first appeared include: Calculate the effect of coupling strength and anonymity value on node scores; Updating the score value of the node; The platform where the node with the highest score is located is selected as the platform where the false information initially appears.

Images