CN101714198B - System based on Bayesian estimation for evaluating credibility of countermeasure information of computer network - Google Patents

Abstract

The invention discloses a system based on Bayesian estimation for evaluating the credibility of countermeasure information of a computer network, which comprises a service response time experimental data configuring module (11), a prior distribution establishing module (1), a prior distribution initial value configuring module (12), an evaluation data inputting module (2), a Bayes formula constructing and computing module (21), a parameter maximum posterior estimation computing module (22), a posterior distribution constructing module (23) and an information credibility computing module (3). By introducing the Bayesian statistical theory, the invention provides a specific evaluation method aiming at the content credibility of the service response time in the running state of network nodes, and the information credibility is obtained by the calculation according to the probability that the information occurs in the historical knowledge. The Bayesian statistical theory is widely applied to the posterior distribution for observation objects established on the basis of prior distribution and posterior information.

Description

Computer network antagonism information reliability assessment system based on Bayesian Estimation

Technical field

The present invention relates to a kind of computer network antagonism (CNO) information processing more particularly be said, be meant that a kind of computer network based on Bayesian Estimation resists (CNO) information reliability assessment system.

Background technology

Computer network antagonism (computer network operations is called for short CNO) is meant on computer network to obtaining Information Superiority and takes to strengthen and safeguard one's own side's information capability, and stops and weaken the activity and the behavior of this ability of the other side and effort.Information process is meant the process that information translation becomes information also to obtain for the user.Process comprises six intelligence activities that are mutually related: plan and indication, collect, handle and processing, analyze and produce, distribution and integrate, assessment and feedback.The research contents of this patent concentrates on the information credibility evaluation part of analyzing with the production phase.The research object of this patent is the service response time in the target operation state.Information is meant that the situation (state) of an antagonism target comprises: running status of Bian Huaing (status) and metastable operation platform feature (platform features) at any time.Running status refers to all properties of system state in certain value in a flash.Such as load, bandwidth and the service response time etc. of system, be the form of classification.Running status is the set of two class values.Be designated as:

STATUS::＝{info，level|info，level∈N，run(info)∧run(level)}(1)

Predicate run (x) expression x is the runtime value that can change at any time, and info represents the discrete values of information.Level represents the discrete classification of state, and level=0 represents that state closes, level=1 ..., n; The n level state that n 〉=1 expression is opened.Can determine the quantification classification of state according to expected degree.The evaluation object of this algorithm is a kind of of level, is a kind of classification of service response time is represented, service response time (service response time) is designated hereinafter simply as RT, concrete RT value representation be sometime the section.

Information input the status information that mainly (computer network utilization) collects from CNE of CNO intelligence channel about target, in reality, that not do not find or not have definite threat, imperfect information, enemy's deceptive practices etc. all be ever-present, so the information that each information source provides all has uncertainty to a certain degree.The generation reason of concrete analysis information uncertainty has two aspects.The one, the gap of the information that target truth and its external manifestation are come out mainly is because target itself may have duplicity; The 2nd, the gap between the collected and information that reports of information that the target external manifestation is come out and gatherer is because the reliability of gatherer is subjected to the influence of its position and current network environment.Directly the behavior of target being analyzed is the comparison difficulty, can directly touch but the behavior of gatherer and information content are people.By analysis with the accuracy of the reliability of gatherer and the information content and these two evaluation indexes comprehensively as three big ingredients of CNO information reliability assessment process.

The authenticity and the reliability of information that reliability assessment is clear and definite, it will have influence on the quality and the efficient of commanding and decision-making.Confidence level is to a kind of tolerance of trusting, and is meant that people are a judgement of genuine degree according to experience in the past to certain things or phenomenon, or perhaps people to certain things or phenomenon for really believing degree.(list of references: " artificial intelligence " Shi Zhongzhi, Wang Wenjie writes, National Defense Industry Press 2007.2) the information credibility assessment is the process that information is provided its confidence level with regard to its information source (in CNO, information source is meant gatherer) reliability and information content accuracy.Utilize association study between the information at the assessment of variety classes information.

Theory and model major part about information evaluation all is to seek a kind of appraisal procedure that is applicable to all situations from the macroscopic view attempt at present, and do not occur as yet at the appraisal procedure of CNO information confidence level.

Summary of the invention

The research that the present invention is directed to concrete CNO field information assessment algorithm plays an important role for the implementation result of evaluation work.The objective of the invention is to propose a kind of concrete evaluating system at the content reliability of service response time in the running status, this system includes service response time experimental data configuration module (11), prior distribution is set up module (1), prior distribution initial value configuration module (12), assessment data load module (2), Bayes formula construction and computing module (21), the estimation of parameter maximum a posteriori computing module (22), sample posteriority distributed structure module (23) and information credibility computing module (3); Under the non-existent situation of assessment data prior distribution, the user generates user's experimental data configuration file by service response time experimental data configuration module (11) input experimental data; Perhaps pass through prior distribution initial value configuration module (12) under assessment data prior distribution and the non-existent situation of user's experimental data, the prior distribution value of input sample and distribution parameter, comprise sample average, sample variance, mean parameter and parameter variance, and generate prior distribution initial value configuration file, export to prior distribution and set up module (1).And prior distribution is set up module (1) under the situation that the assessment data prior distribution exists, and the prior distribution historical information in the reading database is also exported to the Bayes formula construction and computing module (21); And under the non-existent situation of assessment data prior distribution, the user's experimental data configuration file or the prior distribution initial value configuration file that receive are handled, generate the prior distribution historical information and export to the Bayes formula construction and computing module (21).After the prior distribution historical information successfully obtains, from database, read service response time information to be assessed and evaluate parameter set by assessment data load module (2), and generate message file to be assessed and export to Bayes formula construction and computing module (21) and make up Bayes formula at the posteriority Distribution calculation of distribution parameter, and the posteriority distributed intelligence that calculates distribution parameter exports to the parameter maximum a posteriori and estimates that computing module (22) is used to calculate the maximum a posteriori estimated information of distribution, and output maximum a posteriori estimated information is given sample posteriority distributed structure module (23).Sample posteriority distributed structure module (23) is used to calculate the posteriority distribution of the affiliated sample of information to be assessed, and information credibility computing module (3) is given in the distributed intelligence of output sample posteriority.Last information credibility computing module (3) draws final every service response time Reliability of Information information based on posteriority distributed intelligence structure confidence level computing formula.

Intelligence activity is the key activities in the network antagonism commanding and decision-making.Wherein, the information reliability assessment then is one of basis that constitutes intelligence activity and process thereof.Information is that decision making forms the requisite foundation of action scheme, the authenticity and the reliability of information that reliability assessment is clear and definite, and it will have influence on the quality and the efficient of commanding and decision-making.Information credibility in the evaluating system that the present invention proposes is the probability calculation that occurs in historical knowledge (priori) according to information.Bayesian statistical theory is widely used in setting up on based on the basis of prior distribution and posterior information the posteriority of the object of observation being distributed.Therefore the present invention introduces Bayesian statistical theory and solves information content confidence level computational problem in the running status.Advantage of the present invention is: the calculating based on the ideological guarantee information credibility of Bayesian statistics all is based on up-to-date knowledge accumulation; At service response time method for designing with statistical law, make full use of evaluation object and have these characteristics of statistical law, stronger with respect to general reliability assessment method availability, specific aim.

Description of drawings

Fig. 1 is the structured flowchart that the present invention is based on the CNO information reliability assessment method of Bayesian Estimation.

Among the figure: 1. prior distribution is set up module 2. assessment data load modules, 3. information credibility computing module 11. service response time experimental data configuration modules, 12. prior distribution initial value configuration module 21.Bayes formula constructions and computing module 22. parameter maximum a posteriori estimation computing module 23. sample posteriority distributed structure modules

Embodiment

The present invention is described in further detail below in conjunction with drawings and Examples.

The present invention is a kind of CNO information reliability assessment system based on Bayesian Estimation, this system solved the evaluator with the target communication process in, the probability that same information occurs under similar situation is big more just thinks that this information credibility is high more.

Referring to shown in Figure 1, based on the CNO information reliability assessment system of Bayesian Estimation include service response time experimental data configuration module (11), prior distribution set up module (1), prior distribution initial value configuration module (12), assessment data load module (2), Bayes formula construction and computing module (21), parameter day posteriority estimate computing module (22), sample posteriority distributed structure module (23) and information credibility computing module (3).

The basic thought of this patent is to utilize the accumulation of historical knowledge to set up the prior distribution of service response time to be assessed, and the confidence level of RT is got the probability that this time period occurs in prior distribution.Therefore present patent application is elaborated by following three aspects:

(A) set up the prior distribution of evaluation object

In order to set up the prior distribution of certain service RT, need obtain the data set that is used to estimate distribution pattern in advance, may come from two approach: the operation same services obtains by experiment under similar constraint condition, and resulting RT is continuous time value RT (t); From the historical information of the target of similar constraint condition or target itself, obtain, resulting RT be branch the discrete values RT (l) of grade.The constraint condition here may comprise the platform features of server, the time period of service operation (being suitable under the totally different situation of this service performance in one day different time sections) etc., makes and sets up data when distributing try one's best the operation background of destination service.When there was not any priori in certain destination service response time, the data that can use first kind of approach to obtain were analyzed the prior distribution type by the SPSS statistical tool.Finish this step by following three modules:

Service response time experimental data configuration module (11) by reading user's experimental data, generates user's experimental data configuration file FILE under the non-existent situation of assessment data prior distribution ₁₁, export to prior distribution and set up module (1).

Prior distribution initial value configuration module (12) reads the sample of user's input and the prior distribution value of distribution parameter under assessment data prior distribution and the non-existent situation of user's experimental data, and generates prior distribution initial value configuration file FILE ₁₂, export to prior distribution and set up module (1).

Prior distribution is set up module (1) under the situation that the assessment data prior distribution exists, the prior distribution historical information S in the reading database ₁And export to Bayes formula construction and computing module (21).Under the non-existent situation of assessment data prior distribution, to the user's experimental data configuration file FILE that receives ₁₁Perhaps prior distribution initial value configuration file FILE ₁₂Handle, generate prior distribution historical information S ₁And export to Bayes formula construction and computing module (21).

In the present invention, user's experimental data configuration file FILE of service response time experimental data configuration module (11) output ₁₁Form be EvaluateTarget＜TargetID, ResponseTime＜rtset 〉, wherein, EvaluateTarget＜TargetID target under the expression experimental data,＜TargetID〉accord with for target designation; ResponseTime＜rtset〉the service response time data (time value is a successive value) of expression reporting of user,＜rtset〉gather for service response time.

The prior distribution initial value configuration file FILE of prior distribution initial value configuration module (12) output ₁₂Form be EvaluateTarget＜TargetID, PriorDistrlist＜prior_distrlist 〉, EvaluateTarget＜TargetID wherein〉expression is the same, PriorDistrlist＜prior_distrlist〉parameter list of expression assessment data prior distribution,＜prior_distrlist〉be the following initial parameter value of user according to the input of target designation symbol, comprise sample average＜distrlist_sampleMean 〉, sample standard deviation＜distrlist_sampleDeviation 〉, distribution parameter average＜distrlist_parameterMean 〉, distribution parameter standard deviation＜distrlist_parameterDeviation 〉.

(B) adjust distribution parameter based on Bayesian statistics

The theoretical basic thought of Bayes statistical study: going to toward parametric statistics Model parameter θ is had some priori obtaining sample observations x, is exactly prior distribution about the mathematical description of the priori of θ.The principal feature of Bayes statistics is to use prior distribution, and is obtaining sample observations X=(x ₁, x ₂..., x _n) ^TAfter, by the information that X and prior distribution provide, form more complete posterior information.The posteriority that utilizes the Bayes formula to calculate parameter θ afterwards distributes.

Use the Bayes formula in the present invention and need to be grasped before following information:

1. the statistical model of sample is the parametric statistics model, and statistical parameter is θ, and this is the priori about the RT sample distribution, and we have drawn this information when the setting up of a last joint prior distribution;

2. possess priori, can get, also can obtain with past some experimental knowledge to θ by some data analysis in past about θ;

3. new sample observations X=(x ₁, x ₂..., x _n) ^T, i.e. X=(RT ₁, RT ₂..., RT _n) ^T

Above several information have been arranged, just can construct the Bayes formula of the posteriority conditional probability density of calculating distribution parameter θ:

$h\left(\mathrm{\θ}\right|\mathrm{RT})=\frac{P\left\{\mathrm{RT}\right|\mathrm{\θ}\}\mathrm{\π}\left(\mathrm{\θ}\right)}{\underset{\mathrm{\Θ}}{\∫}P\left\{\mathrm{RT}\right|\mathrm{\θ}\}\mathrm{\π}\left(\mathrm{\θ}\right)\mathrm{d\θ}}---\left(2\right)$

In the formula, h (θ | RT) be illustrated in the posteriority conditional probability density that sample RT divides the distribution parameter θ that plants, P{RT| θ } represent that the condition of sample RT distributes, the distribution probability density of π (θ) expression parameter θ,

The joint distribution of expression sample RT and parameter θ.

Because parameter θ is a stochastic variable, and denominator Irrelevant with θ, so have:

h(θ|RT)∝π(θ)L(θ|RT)(3)

In the Bayes statistics, generally take this form convenience of calculation.After the posteriority that obtains parameter θ distributes, just can utilize Bayes parameter point estimation approach to provide the estimated value of θ, can use the maximum a posteriori estimation, get make posteriority distribution h (θ | RT) reach peaked point

Maximum a posteriori as θ estimates that the posteriority that has also just obtained RT distributes.

$h\left({\widehat{\mathrm{\θ}}}_{\mathrm{MD}}\right)=\underset{\mathrm{\θ}\∈\mathrm{\Θ}}{\sup }h\left(\mathrm{\θ}\right|\mathrm{RT})---\left(4\right)$

Finish this step by following four modules:

Assessment data load module (2) reads service response time information to be assessed and evaluate parameter set from database, and generates message file FILE to be assessed ₂Export to Bayes formula construction and computing module (21);

Bayes formula construction and computing module (21) are used to make up the Bayes formula at the posteriority Distribution calculation of distribution parameter, and calculate the posteriority distributed intelligence S of distribution parameter ₂₁Export to the parameter maximum a posteriori and estimate computing module (22);

The parameter maximum a posteriori estimates that computing module (22) is used to calculate the maximum a posteriori estimated information of distribution, and output maximum a posteriori estimated information S ₂₂Give sample posteriority distributed structure module (23);

Sample posteriority distributed structure module (23) is used to calculate the posteriority distribution of the affiliated sample of information to be assessed, and output sample posteriority distributed intelligence S ₂₃Give information credibility computing module (3);

The message file FILE to be assessed of assessment data load module (2) output ₂Form be ResponseTime＜rtset, EvaluateParalist＜evaluate_paralist 〉, EvaluateTarget＜TargetID wherein 〉, ResponseTime＜rtset〉expression implication the same, EvaluateParalist＜evaluate_paralist〉expression evaluate parameter tabulation, comprise target EvaluateTarget to be assessed＜TargetID 〉, data summation SampleLength＜sample_length to be assessed, standard packet counts GroupStandard＜group_standard.The input data here are discrete time grade, next step the time participate in again adjusting distributing after need being converted to three continuous time values earlier.

(C) acquired information confidence level

The content reliability of RT (l) is the probability that occurs in its prior distribution under the constraint condition of being set up, because its expression is the time period of a response time, so its probability of in prior distribution, occurring probability of occurring of this time period just.Promptly

$C\left(\mathrm{RT}\left(l\right)\right|\mathrm{conditions})={\∫}_{l\left(\mathrm{lowerbound}\right)}^{l\left(\mathrm{upperbound}\right)}f\left(\mathrm{RT}\right|\mathrm{conditions})\mathrm{dRT}---\left(5\right)$

The conditions here is meant constraint condition, and l (upper bound) is meant the upper bound of the time period that RT represents, and l (lower bound) represents lower bound.

Information credibility computing module (3) is used for based on posteriority distributed intelligence structure confidence level computing formula and draws final every service response time Reliability of Information information.

Embodiment:

CNO target in this experiment is the Beijing Institute of Aeronautics Website server, and therefore reliability assessment need set up the prior distribution of Beijing Institute of Aeronautics homepage response time to liking the Beijing Institute of Aeronautics homepage response time.Wireshark[10 is adopted in experiment] network packet when networkprotocol analyzer intercepts and captures each visit Beijing Institute of Aeronautics homepage, therefrom calculate the response time.In this experiment, the response time is defined as follows.

The homepage response time: server was to the response TCP message capture time of first HTTP message of user after the TCP three-way handshake connection was set up.Wireshark is to the 5th message of each visit data bag intercepting and capturing generally speaking.

Consider the influence of the network traffic condition of different time sections in a day to test result, the experimental data that the response time prior distribution is set up was separately gathered at three different periods in one day.Test 50 data respectively in the morning, afternoon and evening, in SPSS, these data are drawn histogram and matched curve, can observe the websites response time is to meet normal distribution substantially, whether the distribution very originally of the QQ illustrated handbook of drawing data is similar to normal distribution, from QQ figure, can observe scatter diagram approximate point-blank near, can think that data distribute from normal population.RT～N (μ, σ ²), its density function is

$f\left(\mathrm{RT}\right)=\frac{1}{\sqrt{2\mathrm{\π}}\mathrm{\σ}}\·\exp [-\frac{1}{2}\frac{{(\mathrm{RT}-\mathrm{\μ})}^2}{{\mathrm{\σ}}^2}]---\left(6\right)$

In the formula, μ and σ are average and the standard deviation of response time RT.

RT～RT (RT ₁, RT ₂..., RT ₅₀) ^TBe from normal population N (μ, σ ²) simple random sampling, the sample data above utilizing obtains according to the statistical method of routine: $\widehat{\mathrm{\μ}}=1.520\mathrm{ms},$ ${\widehat{\mathrm{\σ}}}^2=0.226\mathrm{ms}.$ Needs are set up the prior distribution of μ before according to new data average being adjusted, normal distribution average N (μ ₀, τ ²), μ wherein ₀And τ ²Belong to super parameter, by 50 data are divided into groups, the average in each group of asking obtains drawing according to conventional statistical method after the average sample ${\widehat{\mathrm{\μ}}}_0=1.511\mathrm{ms},$ ${\widehat{\mathrm{\τ}}}^2=0.146\mathrm{ms}.$ Obtained the prior distribution of distribution parameter thus.

Gatherer after the CNO target is collected up-to-date information, form posterior information set RT '=(RT ' ₁, RT ₂' .., RT ₁₀') ^T, by formula (2),

$h\left(\mathrm{\μ}\right|\mathrm{RT})\∝\exp \{-\frac{1}{2{\widehat{\mathrm{\σ}}}^2}\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{({\mathrm{RT}}_i^{\′}-\mathrm{\μ})}^2-\frac{{(\mathrm{\μ}-{\widehat{\mathrm{\μ}}}_0)}^2}{2{\mathrm{\τ}}^2}\}$

$=\exp \{-\frac{1}{2}[\frac{10{\mathrm{\μ}}^2-2\·10\mathrm{\μ}\stackrel{\‾}{{\mathrm{RT}}^{\′}}+\underset{i=1}{\overset{10}{\mathrm{\Σ}}}{\mathrm{RT}}_i^{\′2}}{{\widehat{\mathrm{\σ}}}^2}+\frac{{\mathrm{\μ}}^2-2{\widehat{\mathrm{\μ}}}_0\mathrm{\μ}+{{\widehat{\mathrm{\μ}}}_0}^2}{{\mathrm{\τ}}^2}\left]\right\}---\left(7\right)$

Intermediate computations is omitted, and the posteriority that draws μ distributes

Wherein

${\widehat{\mathrm{\μ}}}^{\′}={(\frac{10}{{\mathrm{\σ}}^2}+\frac{1}{{\mathrm{\τ}}^2})}^{-1}(\frac{10\stackrel{\‾}{{\mathrm{RT}}^{\′}}}{{\mathrm{\σ}}^2}+\frac{{\mathrm{\μ}}_0}{{\mathrm{\τ}}^2})=1.601\mathrm{ms},---\left(8\right)$

${\mathrm{\γ}}^2={(\frac{10}{{\mathrm{\σ}}^2}+\frac{1}{{\mathrm{\τ}}^2})}^{-1}=0.212\mathrm{ms}.$

Get

Maximum a posteriori as average is estimated And

${\widehat{\mathrm{\σ}}}^2=\frac{1}{m}\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{({\mathrm{RT}}_i^{\′}-{\widehat{\mathrm{\μ}}}_{\mathrm{MD}})}^2=0.150\mathrm{ms}---\left(9\right)$

Calculate the information RT=8 confidence level of (express time section 1.4ms to 1.7ms, RT represent that the division of grade is determined by gatherer), shown in formula (5),

$\left(\mathrm{RT}\right)=\underset{1.4}{\overset{1.7}{\∫}}\frac{1}{\sqrt{2\mathrm{\π}}{\widehat{\mathrm{\σ}}}_{\mathrm{MD}}}\·\exp [-\frac{1}{2}\frac{{(\mathrm{RT}-{\widehat{\mathrm{\μ}}}_{\mathrm{MD}})}^2}{{{\widehat{\mathrm{\σ}}}_{\mathrm{MD}}}^2}]\mathrm{dRT}=65.52\%---\left(10\right)$

Use the present invention, at first, under the non-existent situation of assessment data prior distribution, the user generates user's experimental data configuration file by service response time experimental data configuration module (11) input experimental data; Perhaps pass through prior distribution initial value configuration module (12) under assessment data prior distribution and the non-existent situation of user's experimental data, the prior distribution value of input sample and distribution parameter, comprise sample average, sample variance, mean parameter and parameter variance, and generate prior distribution initial value configuration file, export to prior distribution and set up module (1).And prior distribution is set up module (1) under the situation that the assessment data prior distribution exists, and the prior distribution historical information in the reading database is also exported to the Bayes formula construction and computing module (21); And under the non-existent situation of assessment data prior distribution, the user's experimental data configuration file or the prior distribution initial value configuration file that receive are handled, generate the prior distribution historical information and export to the Bayes formula construction and computing module (21).After the prior distribution historical information successfully obtains, from database, read service response time information to be assessed and evaluate parameter set by assessment data load module (2), and generate message file to be assessed and export to Bayes formula construction and computing module (21) and make up Bayes formula at the posteriority Distribution calculation of distribution parameter, and the posteriority distributed intelligence that calculates distribution parameter exports to the parameter maximum a posteriori and estimates that computing module (22) is used to calculate the maximum a posteriori estimated information of distribution, and output maximum a posteriori estimated information is given sample posteriority distributed structure module (23).Sample posteriority distributed structure module (23) is used to calculate the posteriority distribution of the affiliated sample of information to be assessed, and information credibility computing module (3) is given in the distributed intelligence of output sample posteriority.Last information credibility computing module (3) draws final every service response time Reliability of Information information based on posteriority distributed intelligence structure confidence level computing formula.

The information reliability assessment system that the present invention proposes based on Bayesian Estimation, emphasize the vital role of prior distribution when the evaluator determines, utilize service response time under the concrete condition to present the characteristics of specific distribution, find out and set up the rule of information distribution, the probability that information to be assessed is occurred in prior distribution is as its confidence level.Experiment has described the foundation of a prior distribution and the computation process of confidence level in detail, proves that Bayesian statistics can adjust prior distribution in time according to fresh information, guarantees that reliability assessment is based on that up-to-date knowledge provides.This algorithm except can among the IICEM vertically association provide the concrete computing method, provide possible reference frame for analyzing information goal behavior model and information uncertainty simultaneously.

Claims (5)

1. one kind based on the computer network of Bayesian Estimation antagonism information reliability assessment system, it is characterized in that: include that service response time experimental data configuration module (11), prior distribution are set up module (1), prior distribution initial value configuration module (12), assessment data load module (2), Bayes formula construction and computing module (21), the parameter maximum a posteriori is estimated computing module (22), sample posteriority distributed structure module (23) and information credibility computing module (3);

Service response time experimental data configuration module (11) by reading user's experimental data, generates user's experimental data configuration file FILE under the non-existent situation of assessment data prior distribution ₁₁, export to prior distribution and set up module (1);

Described user's experimental data configuration file FILE ₁₁Form be EvaluateTarget＜TargetID, ResponseTime＜rtset 〉, wherein, EvaluateTarget＜TargetID target under the expression experimental data,＜TargetID〉accord with for target designation; ResponseTime＜rtset〉the service response time data of expression reporting of user, wherein time value is a successive value,＜rtset〉gather for service response time;

Prior distribution initial value configuration module (12) reads the sample of user's input and the prior distribution value of distribution parameter under assessment data prior distribution and the non-existent situation of user's experimental data, and generates prior distribution initial value configuration file FILE ₁₂, export to prior distribution and set up module (1);

Prior distribution is set up module (1) under the situation that the assessment data prior distribution exists, the prior distribution historical information S in the reading database ₁And export to Bayes formula construction and computing module (21); Under the non-existent situation of assessment data prior distribution, to the user's experimental data configuration file FILE that receives ₁₁Perhaps prior distribution initial value configuration file FILE ₁₂Handle, generate prior distribution historical information S ₁And export to Bayes formula construction and computing module (21);

Assessment data load module (2) reads service response time information to be assessed and evaluate parameter set from database, and generates message file FILE to be assessed ₂Export to Bayes formula construction and computing module (21);

Bayes formula construction and computing module (21) are used to make up the Bayes formula at the posteriority Distribution calculation of distribution parameter, and calculate the posteriority distributed intelligence S of distribution parameter ₂₁Export to the parameter maximum a posteriori and estimate computing module (22);

Described posteriority conditional probability density Bayes formula is $h\left(\mathrm{\θ}\right|\mathrm{RT})=\frac{P\left\{\mathrm{RT}\right|\mathrm{\θ}\}\mathrm{\π}\left(\mathrm{\θ}\right)}{\underset{\mathrm{\Θ}}{\∫}P\left\{\mathrm{RT}\right|\mathrm{\θ}\}\mathrm{\π}\left(\mathrm{\θ}\right)\mathrm{d\θ}},$ Because parameter θ is a stochastic variable, and denominator Irrelevant with θ, thus have h (θ | RT) ∝ π (θ) L (θ | RT); After the posteriority that obtains parameter θ distributes, utilize Bayes parameter point estimation approach to provide the estimated value of θ, use maximum a posteriori to estimate, get make posteriority distribution h (θ | RT) reach peaked point

Maximum a posteriori as θ estimates that the posteriority that has also just obtained RT distributes

The content reliability of RT (l) is the probability that occurs in its prior distribution under the constraint condition of being set up, because its expression is the time period of a response time, so its probability of in prior distribution, occurring probability of occurring of this time period just

$C\left(\mathrm{RT}\left(l\right)\right|\mathrm{conditions})={\∫}_{l\left(\mathrm{lower\; bound}\right)}^{l\left(\mathrm{upper\; bound}\right)}f\left(\mathrm{RT}\right|\mathrm{conditions})\mathrm{dRT};$

H (θ | RT) be illustrated in the posteriority conditional probability density that sample RT divides the distribution parameter θ that plants;

P{RT| θ } represent that the condition of sample RT distributes, the distribution probability density of π (θ) expression parameter θ;

The joint distribution of expression sample RT and parameter θ;

Conditions is meant constraint condition, and l (upper bound) is meant the upper bound of the time period that RT represents, and l (lower bound) represents lower bound;

The parameter maximum a posteriori estimates that computing module (22) is used to calculate the maximum a posteriori estimated information of distribution, and output maximum a posteriori estimated information S ₂₂Give sample posteriority distributed structure module (23);

Sample posteriority distributed structure module (23) is used to calculate the posteriority distribution of the affiliated sample of information to be assessed, and output sample posteriority distributed intelligence S ₂₃Give information credibility computing module (3);

Information credibility computing module (3) is used for based on posteriority distributed intelligence structure confidence level computing formula and draws final every service response time Reliability of Information information.

2. the computer network antagonism information reliability assessment system based on Bayesian Estimation according to claim 1 is characterized in that: the prior distribution initial value configuration file FILE of prior distribution initial value configuration module (12) output ₁₂Form be EvaluateTarget＜TargetID, PriorDistrlist＜prior_distrlist 〉, EvaluateTarget＜TargetID wherein target under the expression experimental data,＜TargetID〉accord with for target designation; PriorDistrlist＜prior_distrlist〉parameter list of expression assessment data prior distribution,＜prior_distrlist〉be the user according to the following initial parameter value of target designation symbol input, comprise sample average＜distrlist_sampleMean 〉, sample standard deviation＜distrlist_sampleDeviation, distribution parameter average＜distrlist_parameterMean, distribution parameter standard deviation＜distrlist_parameterDeviation.

3. the computer network antagonism information reliability assessment system based on Bayesian Estimation according to claim 1 is characterized in that: the message file FILE to be assessed of assessment data load module (2) output ₂Form be ResponseTime＜rtset, EvaluateParalist＜evaluate_paralist 〉, EvaluateTarget＜TargetID wherein 〉, ResponseTime＜rtset〉target under the expression experimental data,＜TargetID〉accord with for target designation; EvaluateParalist＜evaluate_paralist〉expression evaluate parameter tabulation, comprise target EvaluateTarget to be assessed＜TargetID 〉, data summation SampleLength＜sample_length to be assessed, standard packet counts GroupStandard＜group_standard; The input data here are discrete time grade, next step the time participate in again adjusting distributing after need being converted to three continuous time values earlier.

4. the computer network antagonism information reliability assessment system based on Bayesian Estimation according to claim 1, it is characterized in that: service response time information posteriority distributed structure is made up of three parts, comprise Bayes formula construction and computing module (21), parameter maximum a posteriori estimation computing module (22), sample posteriority distributed structure module (23), to the prior distribution historical information S that receives ₁With message file FILE to be assessed ₂Handle, adjusted the also posteriority distribution parameter average＜distrlist_parameterMean of tectonic information distribution parameter before this〉and distribution parameter standard deviation＜distrlist_parameterDeviation, get the parameter of the maximum a posteriori estimated value of parameter posteriority distribution as the distribution of information posteriority, the posteriority distribution characteristics value that constructs information is sample average＜distrlist_sampleMean〉and sample standard deviation＜distrlist_sampleDeviation.

5. the computer network antagonism information reliability assessment system based on Bayesian Estimation according to claim 1, it is characterized in that: it is the posteriority distributed intelligence S of information that information credibility computing module (3) receives sample ₂₃Structure confidence level computing formula is tried to achieve probability that time period of information representation to be assessed occurs as its confidence level in posteriority distributes.

Images