An Open-Source Intelligence Trusted Analysis Method Based on Bayesian Network and Evidence Theory

Abstract

The present invention provides an open-source intelligence credibility analysis method based on Bayesian network and evidence theory, including: (1) Acquisition, identification and extraction of open-source intelligence, continuously collecting the latest open-source intelligence from multiple intelligence sources; (2) Fusion of open-source intelligence, integrating different forms of intelligence to form a data set with a unified format; (3) Conducting credibility analysis on open-source intelligence based on Bayesian network and evidence theory. First, comprehensively considering various credibility influencing factors, including the Bayesian network prediction probability of the occurrence of a situation event related to the open-source intelligence, calculating the credibility of the intelligence, and using the membership function to convert the credibility value of the intelligence into a trust evaluation triple. Finally, using the D-S evidence theory to fuse multiple evidences to obtain the judgment result of a certain situation event, and accordingly generating an evaluation of the intelligence related to the event. The present invention can effectively screen and sort out open-source intelligence quickly, screen out highly credible open-source intelligence, and greatly reduce the time consumed by manual screening.

Description

Technical Field

[0001] The present invention relates to the field of computing the credibility of intelligence, and particularly to an open-source intelligence credibility analysis method based on Bayesian network and evidence theory Background Art

[0002] Currently, various open-source intelligence systems have been developed at home and abroad. However, most of them focus on intelligence collection and aggregation. There are relatively few research results on the perception of open-source intelligence credibility, lacking theoretical guidance on trust management, and it is difficult to fully utilize the value of open-source intelligence. The credibility analysis of open-source intelligence faces a series of challenges. The current credibility discrimination technology mainly relies on neural network algorithms, especially the deep belief network DBN and the recurrent neural network RNN. Most of the past research is based on supervised learning methods. However, the data in actual open-source intelligence is unlabeled, and it is obviously unrealistic to manually label this intelligence Summary of the Invention

[0003] Object of the Invention: The technical problem to be solved by the present invention is to provide an open-source intelligence credibility analysis method based on Bayesian network and evidence theory for the deficiencies of the prior art, including the following steps

[0004] Step 1: Obtain open-source intelligence and perform identification and extraction

[0005] Step 2: Perform open-source intelligence fusion, integrate different forms of intelligence to form a data set with a unified format, for example, convert picture and multimedia information into pure text content through corresponding technologies and store it in a database

[0006] Step 3: Conduct credibility analysis on open-source intelligence based on Bayesian network and evidence theory. First, comprehensively consider various credibility influencing factors, including the Bayesian network prediction probability of the occurrence of a situation event related to the open-source intelligence, calculate the credibility of the intelligence, and use the membership function to convert the credibility value of the intelligence into a trust evaluation triple. Finally, use the D-S evidence theory to fuse multiple evidences to obtain a judgment result for a certain situation event, and accordingly evaluate the intelligence related to the event. In order to accurately calculate the credibility of open-source intelligence, the present invention proposes four key factors affecting the credibility of open-source intelligence, namely the credibility of the intelligence source, timeliness, recognition degree, and battlefield situation. As an influencing factor of the intelligence credibility proposed by the present invention, the battlefield situation combines historical experience and uses the Bayesian network to predict the probability of the occurrence of the situation event e j The probability of occurrence, so as to confirm the credibility of the open-source intelligence related to the situation event. According to the above factors, the present invention proposes a new calculation formula for the credibility of open-source intelligence. Using this formula, the credibility of the intelligence released by the intelligence source k about the event e jThe credibility of the intelligence, and the membership function is used to convert the credibility of the intelligence into a trust evaluation triple. Finally, the D-S evidence theory is comprehensively used for the events e released by multiple intelligence sources j The trust evaluation triples of the intelligence are fused to obtain the fused trust evaluation triple of the event e j and judge whether to believe the event e according to the evaluation result j and judge accordingly whether the intelligence related to the event e j is credible.

[0007] Furthermore, in step 1, a web crawler program is used to collect open-source intelligence from important domestic and foreign military news websites, national official websites, large well-known commercial websites, etc. Since the web page format of the same website is usually carried by a very small number of fixed templates and each page has a similar structure, the template can be extracted from the specific web page configuration. For example, different pages of the same website can be crawled according to the URL in the web page configuration, and the template used by the website can be extracted, the information to be extracted is located in the template, and finally the information is stored in the database.

[0008] The open-source intelligence fusion described in step 2 refers to the internal fusion of open-source intelligence. The internal fusion means integrating the information with different data structures obtained through different collection channels to form a data set with a unified format. For example, the open-source intelligence with different data formats collected is integrated into a data set with a unified format through methods such as title extraction, abstract extraction, keyword extraction, full-text extraction, related term extraction, and rich document recognition, and finally a structured database is constructed.

[0009] Step 3 includes:

[0010] Step 3-1, determine the following factors that affect the credibility of open-source intelligence: [[ID=2I]]

[0011] 1) The credibility of the intelligence source

[0012] Generally, it is considered that the higher the trust value of the source of the intelligence, the more credible the intelligence sent by this intelligence source.

[0013] 2) Timeliness

[0014] Timeliness mainly considers the time elapsed from the generation of the intelligence to its use. Because the actual situation changes rapidly, the newer the intelligence, the greater the reference value.

[0015] 3) Recognition degree

[0016] The recognition degree mainly considers how many times the intelligence has been forwarded from generation to reception. According to the forwarding strategy of the intelligence, it is generally considered that the more times a piece of intelligence is forwarded, the more credible the intelligence is. [[ID:40]]

[0017] 4) Battlefield Situation

[0018] The battlefield situation mainly considers predicting the probability of an event occurring based on historical experience. According to the characteristics of the event, the factors affecting the occurrence of the event are extracted to determine the variable set of the Bayesian network. Then, based on expert knowledge, etc., structure learning is carried out to determine the causal relationship between variables and form the structure of the Bayesian network. Each node in the Bayesian network represents a variable, and the connection lines between nodes represent the causal relationship between variables. No connection line indicates that the variables are conditionally independent.

[0019] First, it is necessary to determine the variable set of the Bayesian network. Taking the event of whether the Red Army launches an attack in a certain battle as an example, three influencing factors, namely weather (Weather), the strength of the Red Army (Force), and the terrain where the Red Army is located (Terrain), are selected to obtain the variable set X = {Weather, Force, Terrain, AttackAccident}. The possible values of each variable are as follows:

[0020] Weather: {good, bad};

[0021] Force: {strong, weak}

[0022] Terrain: {complex, flat}

[0023] AttackAccident: {occur, not occur}

[0024] Then, based on expert knowledge, etc., structure learning is carried out to determine the causal relationship between variables and form the structure of the Bayesian network. Each node in the Bayesian network represents a variable, and the connection lines between nodes represent the causal relationship between variables. No connection line indicates that the variables are conditionally independent.

[0025] Step 3-2: According to the determined network structure, parameter learning is carried out to establish a conditional probability table between nodes.

[0026] Step 3-3: According to the factors of the credibility of open-source intelligence in Step 3-1, calculate the intelligence credibility;

[0027] Step 3-4: At this time, the credibility values of each piece of intelligence about event e j are obtained. In order to increase the consideration of intelligence uncertainty and maintain the flexibility of judgment results, the membership function T(c) in fuzzy set theory is introduced, and the credibility value of the intelligence is input into the membership function to obtain the trust, uncertainty, and degree of trust of this intelligence.

[0028] Step 3-5: Use the D-S evidence theory to fuse the intelligence trust values represented by the triples.

[0029] Furthermore, step 3-2 includes: Bayes' rule is the basis of Bayesian learning, and the probability is calculated using the following formula:

[0030]

[0031] Where D is the training data, which comes from the historical messages in the Tangle network. P(D) represents the prior probability of the training data D. h represents the candidate hypothesis in the hypothesis space H. P(h) is called the prior probability of h, which represents the initial probability before training. P(D|h) represents the probability of observing the training data D when the hypothesis h is true. P(D|h) represents the probability of observing the training data D when the candidate hypothesis h is true, reflecting the probability that h is true after observing the training data D. P(h|D) represents the posterior probability of the candidate hypothesis h.

[0032] After obtaining the conditional probability of an event occurring under different conditions, we can predict the event e under any conditions based on the trained Bayesian network model. j The probability of occurrence p j .

[0033] In step 3-3, the following formula is used to calculate the credibility of intelligence:

[0034] in is the information released by the kth intelligence source about event e j Credibility of intelligence, Cr k Indicates the trust value of the intelligence sender, max m∈M (Cr m ) represents event e j All corresponding information M j The sender with the largest trust value, It represents the time from the generation of information to its calculation, b is the coefficient to be determined, Indicates the number of hops from the generation to the reception of intelligence. represents the probability of an event obtained through Bayesian network reasoning, e is a natural constant, α1, α2, α3, α4 are weights, and α1+α2+α3+α4=1; the information receiver calculates the probability of event e according to formula (2) j The credibility value of all intelligence is obtained from the set k ranges from 1 to M;

[0035] In step 3-4, the membership function T(c) is shown in formula (3):

[0036]

[0037] Where T(c) represents the membership, x represents the credibility value of the message, and δ represents the variance of the Gaussian distribution. Then, using evidence theory, we can predict the event ej The membership function results corresponding to the credibility values of all information are synthesized to obtain the comprehensive trust value T j :

[0038] where represents the membership function of the credibility of the information about event e released by the M-th information source j of the information;

[0039] Step 3-5 includes: As shown in Equation (3), the output of T(c) is a triple (x, y, z) for the evaluation of information trust. Among them, x represents the disbelief degree of the information, y represents the uncertainty degree of the information, and z represents the credibility degree of the information. The D-S evidence theory is used to fuse the information trust values represented by the triple, specifically including:

[0040] First, define the trust space T represents trusting the information, and vice versa, represents not trusting the information. Then define the hypothesis S = {T}, S = {T} means the information is credible, means the information is not credible; the hypothesis means it is uncertain whether the information is credible.

[0041] Define the mass function m() as follows:

[0042] m(S) = α(5) m(I) = 1 - α - β

[0043] (7) where α and β are intermediate parameters, 0 ≤ α, β ≤ 1 and α + β ≤ 1. The synthesis method of the mass function is:

[0044]

[0045]

[0046] where m1(I) and m2(I) represent the mass functions of information source 1 and information source 2 respectively, and K is the normalization constant, and the calculation method is:

[0047] Finally, obtain the triple of trust evaluation after the fusion of event e j and select whether to believe the event according to the evaluation result, and accordingly generate an evaluation of the information related to the event to judge whether the information is credible.

[0048] Different from traditional trust discrimination technologies, the present invention proposes an analysis method for the credibility of intelligence based on Bayesian networks and evidence theory.

[0049] Each node in a Bayesian network represents a variable, and the connections between nodes represent the causal relationships between variables. The absence of a connection indicates conditional independence between variables. A Bayesian network is a probabilistic inference model that can solve problems related to probabilistic inference with conditional dependencies. It can predict the probability of a certain situation event occurring through a Bayesian network, thereby corroborating the trustworthiness of a piece of intelligence. This is why a Bayesian network can be used in a trust analysis system.

[0050] The D-S evidence theory belongs to the method of uncertainty reasoning and has the ability to process uncertain information. Using the D-S evidence theory, multiple pieces of evidence can be fused to obtain a judgment result for a certain situation event, and based on this, an evaluation can be made on the intelligence related to the event. This is why the D-S evidence theory can be used in a trust analysis system.

[0051] Beneficial effects: The present invention uses a credibility analysis technology based on Bayesian networks and evidence theory to calculate the credibility of intelligence, which can effectively screen and sort open-source intelligence quickly, screen out highly credible open-source intelligence, greatly reduce the time consumed by manual screening, and greatly reduce the risk of intelligence obsolescence caused by the long time-consuming intelligence screening. Brief Description of the Drawings

[0052] The following further specifically describes the present invention in conjunction with the drawings and specific embodiments, and the above and / or other advantages of the present invention will become clearer.

[0053] Figure 1 It is a schematic diagram of accurate extraction of web page content based on a template.

[0054] Figure 2 It is a schematic diagram of internal fusion of open-source intelligence.

[0055] Figure 3 It is a structure diagram of a Bayesian network for attack events. Specific Embodiments

[0056] The present invention provides an open-source intelligence trust analysis method based on Bayesian networks and evidence theory, including the following steps:

[0057] (1) Acquisition, identification, and extraction of open-source intelligence, continuously collecting the latest open-source intelligence from multiple intelligence sources

[0058] (2) Fusion of open-source intelligence, integrating different forms of intelligence to form a data set with a unified format

[0059] (3) Conduct credible analysis on open source intelligence based on Bayesian network and evidence theory. First, comprehensively consider various credibility influencing factors, including the Bayesian network prediction probability of the situation events related to open source intelligence, calculate the credibility of the intelligence, and use the membership function to convert the credibility value of the intelligence into a trust evaluation triple. Finally, use the D-S evidence theory to fuse multiple evidences to obtain the judgment result of a certain situation event, and accordingly evaluate the intelligence related to the event.

[0060] Regarding step (1) of the solution, use a web crawler program to collect open source intelligence from important domestic and foreign military news websites, national official websites, large well-known commercial websites, etc. Adopt the idea of accurate extraction of web page data based on templates, extract templates from specific web page configurations, then locate the information to be extracted in the templates, and finally store it in the database. The accurate extraction method of web pages based on templates converts web pages into structured data through templates, and the change of web page format does not require modifying the code of the extraction system. Use the XSL module to extract data, and the data flow is HTML → XML → database, gradually changing from unstructured to structured, and finally converting into available and well-defined structured data. The process is as Figure 1 shown.

[0061] The specific description of the web page content extraction algorithm is as follows:

[0062] The first step: Normalize the obtained web page format and convert it into a DOM tree using JDOM;

[0063] The second step: Obtain the data reference points and write an XSL template to find the data reference points;

[0064] The third step: Locate the elements and attributes in the XML document and map the HTML file to an XML file;

[0065] The fourth step: Scan the XML file forward to obtain the attribute tags and attribute contents;

[0066] The fifth step: Construct a HashMap hash table, where the key corresponds to the XML tag and the value corresponds to the tag content;

[0067] The sixth step: Store the data extracted from the XML file in the database.

[0068] Regarding step (2) of the solution, the fusion of open source intelligence mainly refers to the internal fusion of open source intelligence. The internal fusion of multi-source data means integrating information with different data structures obtained through different collection channels to form a data set with a unified format and wide application. The process is as Figure 2 shown.

[0069] Regarding the open source intelligence credible analysis technology based on Bayesian network and evidence theory in step (3) of the solution. The implementation steps are as follows:

[0070] At the beginning, factors affecting the credibility of open-source intelligence are determined. The following influencing factors are proposed in this paper:

[0071] 1) Credibility of the intelligence source

[0072] Generally, the higher the trust value of the intelligence source is considered, the more credible the intelligence sent by this source is.

[0073] 2) Timeliness

[0074] Timeliness mainly considers the time elapsed from the generation to the use of the intelligence. Because the actual situation changes rapidly, the newer the intelligence is, the greater its reference value.

[0075] 3) Recognition rate

[0076] The recognition rate mainly considers how many times the intelligence has been forwarded from generation to reception. According to the forwarding strategy of the intelligence, it is generally believed that the more times a piece of intelligence is forwarded, the more credible it is.

[0077] 4) Battlefield situation

[0078] The battlefield situation mainly considers the prediction of the probability of an event occurring based on historical experience. According to the characteristics of the event, factors affecting the occurrence of the event are extracted to determine the variable set of the Bayesian network. Then, based on expert knowledge, etc., structure learning is carried out to determine the causal relationship between variables, forming the structure of the Bayesian network. Each node in the Bayesian network represents a variable, and the connection lines between nodes represent the causal relationship between variables. No connection line indicates conditional independence between variables.

[0079] First, the variable set of the Bayesian network needs to be determined. Taking the event of whether the Red Army launches an attack in a certain battle as an example, three influencing factors, namely weather, the Red Army's troops, and the terrain where the enemy is located, are selected to obtain the variable set X = {Weather, Force, Terrain, attackAccident}. The possible values of each variable are shown in Table 1.

[0080] Table 1

[0081]

[0082] Then, based on expert knowledge, etc., structure learning is carried out to determine the causal relationship between variables, forming the structure of the Bayesian network. Each node in the Bayesian network represents a variable, and the connection lines between nodes represent the causal relationship between variables. No connection line indicates conditional independence between variables. The network structure is as Figure 3 shown.

[0083] Then, parameter learning is performed based on the determined network structure to establish a conditional probability table between nodes. The results after parameter learning are shown in Table 2 (Bayesian network conditional probability table).

[0084] Table 2

[0085]

[0086] The probability is calculated using the following formula:

[0087]

[0088] Where D is the training data, which comes from the historical messages in the Tangle network. P(D) represents the prior probability of the training data D. h represents the candidate hypothesis in the hypothesis space H. P(h) is called the prior probability of h, which represents the initial probability before training. P(D|h) represents the probability of observing the training data D when the hypothesis h is true. P(D|h) represents the probability of observing the training data D when the candidate hypothesis h is true, reflecting the probability that h is true after observing the training data D. P(h|D) represents the posterior probability of the candidate hypothesis h.

[0089] After obtaining the conditional probability of an event occurring under different conditions, the trained Bayesian network model can be used to predict the event e under any conditions. j The probability of occurrence p j .

[0090] According to the above four credibility influencing factors, the intelligence credibility calculation method is shown in formula (2).

[0091]

[0092] in It is the information released by intelligence source k about event e j Credibility of intelligence, Cr k Indicates the trust value of the intelligence sender, max(Cr m ) represents event e j All corresponding information M j The sender with the largest trust value, Indicates how long it takes for the information to be generated and calculated, and b is an undetermined coefficient. Indicates the number of hops from the generation to the reception of the information, p j represents the probability of the event occurring after Bayesian network reasoning, α is the weight of each item, and α1+α2+α3+α4=1. The intelligence receiver calculates the probability of the event e according to formula (2). j The credibility value of all intelligence is obtained from the set

[0093] At this time, the credibility values of each piece of information about event e j are obtained. To increase the consideration of information uncertainty and maintain the flexibility of judgment results, the membership function in fuzzy set theory is introduced. The credibility value of the information is input into the membership function to obtain the trust, uncertainty, and degree of trust of the information. The membership function adopts the Gaussian membership function shown in Equation (3).

[0094]

[0095] where T(c) represents the membership degree, x represents the credibility value of the message, and δ represents the variance of the Gaussian distribution. Then, using the evidence theory, the results of the membership functions corresponding to the credibility values of all information about event e j are synthesized to obtain the comprehensive trust value T j , as shown in Equation (4):

[0096]

[0097] As shown in Equation (3), the output of T is a triple (x, y, z) for the evaluation of information trust. x represents the disbelief degree of the information, y represents the uncertainty degree of the information, and z represents the credibility degree of the information. For example, after inputting the credibility value of a certain piece of information into the membership function of Equation (3), the output (0, 0, 1) is obtained. The disbelief and uncertainty degrees of this information are 0, and the credibility degree is 1, indicating that the credibility of this information is extremely high. Finally, the D-S evidence theory is used to fuse the information trust values represented by the triple, and the specific method is as follows.

[0098] First, define the trust space T represents trusting this information. Conversely, represents not trusting this information. Then define the hypothesis S = {T}, indicating that the information is credible, represents that the information is not credible; the hypothesis means that it is uncertain whether the information is credible.

[0099] Define the mass function m() as follows:

[0100] m(S) = α(5) m(I) = 1 - α - β

[0101] (7) where α and β are intermediate parameters, 0 ≤ α, β ≤ 1 and α + β ≤ 1. The synthesis method of the mass function is:

[0102]

[0103]

[0104] where \(m_1(I)\) and \(m_2(I)\) represent the mass functions of information source 1 and information source 2 respectively, and \(K\) is the normalization constant, which is calculated as follows:

[0105] Finally, the event \(e\) j The trust evaluation triple after fusion is obtained, and whether to believe this event is selected according to the evaluation result. Based on this, an evaluation of the information related to the event is generated to judge whether the information is credible.

[0106] For example, according to Equation (3), the trust evaluation triples of information source 1 and information source 2 for a certain piece of information are obtained as \((0.01, 0.05, 0.94)\) and \((0.1, 0.2, 0.7)\) respectively. Now, the evidence theory is used to fuse the information trust values of these two information sources. Let \(m_1\) represent the mass function of information source 1 and \(m_2\) represent the mass function of information source 2. According to Equations (5), (6), and (7), the content shown in Table 3 can be obtained.

[0107] Table 3

[0108]

[0109] Finally, according to Equation (11), the value of \(k\) can be calculated, and then substituting it into Equations (8), (9), and (10), the trust evaluation triple after fusion can be obtained, that is, the unbelievability, uncertainty, and believability of a certain piece of information.

[0110] In specific implementation, the present application provides a computer storage medium and a corresponding data processing unit. Among them, the computer storage medium can store a computer program, and when the computer program is executed by the data processing unit, it can run the content of the invention of an open-source intelligence credibility analysis method based on Bayesian network and evidence theory and some or all of the steps in each embodiment. The storage medium can be a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM), etc.

[0111] Those skilled in the art can clearly understand that the technical solutions in the embodiments of the present invention can be implemented by means of a computer program and its corresponding general hardware platform. Based on such an understanding, the technical solutions in the embodiments of the present invention, in essence or the part that contributes to the prior art, can be embodied in the form of a computer program, that is, a software product. This computer program software product can be stored in a storage medium, including several instructions to enable a device (which can be a personal computer, a server, a single-chip microcomputer, a MUU or a network device, etc.) containing a data processing unit to execute the methods described in various embodiments or some parts of the embodiments of the present invention.

[0112] The present invention provides an open-source intelligence credible analysis method based on Bayesian networks and evidence theory. There are many methods and ways to specifically implement this technical solution. The above is only the preferred embodiment of the present invention. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present invention, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present invention. Each component not clearly defined in this embodiment can be implemented by the prior art.

Claims

1. A trustworthy analysis method for open-source intelligence based on Bayesian networks and evidence theory, characterized by: The following steps are involved: Step 1: Obtain open source intelligence and perform identification and extraction; Step 2: Conduct open-source intelligence fusion to integrate different forms of intelligence into a unified data set; Step 3: Conduct credibility analysis of open-source intelligence based on Bayesian networks and evidence theory; In step 1, a web crawler is used to collect open source intelligence from the website, extract templates from specific web page configurations, extract the templates used by the website, locate the information to be extracted in the templates, and finally store the information in the database; In step 2, the open source intelligence fusion refers to integrating information with different data structures obtained through different collection channels to form a data set with a unified format; Step 3 includes: Step 3-1: Identify the following factors that influence the credibility of open-source intelligence: source credibility, timeliness, recognition, and battlefield situation. Step 3-2, perform parameter learning based on the determined network structure and establish a conditional probability table between nodes; Step 3-3, calculate the credibility of intelligence based on the factors of open source intelligence credibility in step 3-1; Step 3-4: Introduce the membership function T(c) from fuzzy set theory, input the credibility value of the intelligence into the membership function, and obtain the trust, uncertainty, and degree of trust of the intelligence; Step 3-5: Use DS evidence theory to fuse the intelligence trust values represented by the triples; Step 3-2 includes calculating the probability using the following formula: Where D is the training data, P(D) represents the prior probability of the training data D, h represents the candidate hypothesis in the hypothesis space H, P(h) is called the prior probability of h, P(D|h) represents the probability of observing the training data D when the candidate hypothesis h is true, and P(h|D) represents the posterior probability of the candidate hypothesis h; After obtaining the conditional probability of events occurring under different conditions, we can predict the event e under any conditions based on the trained Bayesian network model. j The probability of occurrence p j ; Step 3-3 includes: calculating the credibility of the intelligence using the following formula: in is the information released by the kth intelligence source about event e j Credibility of intelligence, Cr k Indicates the trust value of the intelligence sender, Indicates event e j All corresponding information M j The sender with the largest trust value, It represents the time from the generation of information to its calculation, b is the coefficient to be determined, Indicates the number of hops from the generation to the reception of intelligence. represents the probability of an event obtained through Bayesian network reasoning, e is a natural constant, α1, α2, α3, α4 are weights, and α1+α2+α3+α4=1; the information receiver calculates the probability of event e according to formula (2) j The credibility value of all intelligence is obtained as a set The value of k ranges from 1 to M.

2. The method according to claim 1, characterized in that In step 3-4, the membership function T(c) is shown in formula (3): Where T(c) represents the membership degree, and c represents the credibility value of the intelligence; Represents the variance of the Gaussian distribution, using evidence theory, for the event e j The membership function results corresponding to all intelligence credibility values are synthesized to obtain the comprehensive trust value T j : in Indicates the information released by the Mth intelligence source about event e j Membership function of the credibility of intelligence.

3. The method according to claim 2, characterized in that Steps 3-5 include: As shown in formula (3), the output of T(c) is a triple (x, y, z) about the trust evaluation of intelligence, where x represents the unreliability of intelligence, y represents the uncertainty of intelligence, and z represents the credibility of intelligence; define the trust space T stands for trust in intelligence, Indicates distrust of intelligence; define the assumption S = {T}, S = {T} indicates that the intelligence is credible, Indicates that the intelligence is unreliable; assume expressing uncertainty about the credibility of the intelligence; Define the mass function m() as follows: m(S)=α(5) m(I)=1-α-β (7) Where α and β are intermediate parameters, 0≤α, β≤1 and α+β≤1, and the synthesis method of mass function is: Where m1(I) and m2(I) represent the mass function of intelligence source 1 and the mass function of intelligence source 2 respectively, and K is a normalization constant, which is calculated as follows: Finally, we get event e j After fusion, the trust evaluation triplet is formed, and based on the evaluation results, it is decided whether to believe the event. Based on this, an evaluation is made on the intelligence related to the event to determine whether the intelligence is credible.

Images