CN102456158B - Based on the air traffic control atm information system security assessment method of ANN BP model - Google Patents
Based on the air traffic control atm information system security assessment method of ANN BP model Download PDFInfo
- Publication number
- CN102456158B CN102456158B CN201010519836.4A CN201010519836A CN102456158B CN 102456158 B CN102456158 B CN 102456158B CN 201010519836 A CN201010519836 A CN 201010519836A CN 102456158 B CN102456158 B CN 102456158B
- Authority
- CN
- China
- Prior art keywords
- network
- blank pipe
- infosystem
- error
- neural network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000013528 artificial neural network Methods 0.000 claims abstract description 31
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 20
- 238000004891 communication Methods 0.000 claims abstract description 13
- 238000003062 neural network model Methods 0.000 claims abstract description 7
- 238000011156 evaluation Methods 0.000 claims description 53
- 238000012549 training Methods 0.000 claims description 42
- 230000006870 function Effects 0.000 claims description 30
- 238000012360 testing method Methods 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 16
- 238000004458 analytical method Methods 0.000 claims description 14
- 238000004088 simulation Methods 0.000 claims description 6
- 210000002569 neuron Anatomy 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 4
- 230000009466 transformation Effects 0.000 claims description 3
- 230000001419 dependent effect Effects 0.000 claims description 2
- 230000001131 transforming effect Effects 0.000 claims description 2
- 238000006243 chemical reaction Methods 0.000 claims 3
- 230000001537 neural effect Effects 0.000 claims 3
- 238000012546 transfer Methods 0.000 claims 2
- 241001269238 Data Species 0.000 claims 1
- 230000003292 diminished effect Effects 0.000 claims 1
- 238000013341 scale-up Methods 0.000 claims 1
- 238000011478 gradient descent method Methods 0.000 abstract 1
- 238000011160 research Methods 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 238000000513 principal component analysis Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 238000013278 delphi method Methods 0.000 description 3
- 238000011161 development Methods 0.000 description 3
- 238000004451 qualitative analysis Methods 0.000 description 3
- 238000011002 quantification Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 230000001186 cumulative effect Effects 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000013210 evaluation model Methods 0.000 description 2
- 238000007429 general method Methods 0.000 description 2
- 230000036541 health Effects 0.000 description 2
- 230000009545 invasion Effects 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000013178 mathematical model Methods 0.000 description 2
- 230000003449 preventive effect Effects 0.000 description 2
- 238000012847 principal component analysis method Methods 0.000 description 2
- 238000004445 quantitative analysis Methods 0.000 description 2
- VUFNLQXQSDUXKB-DOFZRALJSA-N 2-[4-[4-[bis(2-chloroethyl)amino]phenyl]butanoyloxy]ethyl (5z,8z,11z,14z)-icosa-5,8,11,14-tetraenoate Chemical compound CCCCC\C=C/C\C=C/C\C=C/C\C=C/CCCC(=O)OCCOC(=O)CCCC1=CC=C(N(CCCl)CCCl)C=C1 VUFNLQXQSDUXKB-DOFZRALJSA-N 0.000 description 1
- 241000700605 Viruses Species 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000009395 breeding Methods 0.000 description 1
- 230000001488 breeding effect Effects 0.000 description 1
- 238000005291 chaos (dynamical) Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 230000001066 destructive effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000012854 evaluation process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 238000010230 functional analysis Methods 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000006855 networking Effects 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000001902 propagating effect Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 238000005303 weighing Methods 0.000 description 1
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Do you the present invention is based on ANN? the air traffic control atm information system security assessment method of BP model, utilizes nerual network technique principle, carries out safety assessment to blank pipe infosystem, is the new method to air traffic control system assessment in information security field.Blank pipe infosystem comprises four major parts, communication system, navigational system, surveillance and automated system four Iarge-scale system.Neural network, from the index affecting blank pipe infosystem, will affect the input of specific targets as neural network of four Iarge-scale system securities in blank pipe infosystem.Whole network adopts error backpropagation algorithm, and network exports and target output contrasts, if the difference of the output valve of this network and desired value is outside error range, then adopts gradient descent method to carry out weighed value adjusting, until meet the scope of error permission.Network finally exports the safe class into whole blank pipe infosystem.The security of this blank pipe infosystem entirety can be assessed accurately by this neural network model.
Description
Technical field
The present invention is that one utilizes nerual network technique to carry out the method for safety assessment (SecurityEvaluation) to the empty traffic administration of aviation (being called for short: blank pipe) ATM (AirTrafficManagement) infosystem.It relates to blank pipe infotech and safety technique, belongs to safety assessment field in information security field.
Background technology
Blank pipe information resources guarantee to realize the basic resource of air traffic safety management.Blank pipe informatization has entered the important period pushed forward comprehensively and accelerate development.Through building for many years, built much information system in each important service department of air traffic control system, Computer information network has become the groundwork means of blank pipe service operation.In the face of in large scale, apply various, that user is huge and business degree of dependence is high blank pipe infosystem, extensive information-based and networking has become the outstanding feature that air traffic control system develops.
At present, domestic and international network security situation allows of no optimist, and blank pipe infosystem faces huge threat (invasion, attack and virus etc.).Local and overseas hostile force is becoming increasingly rampant for the attack destructive activity of satellite, wireless and ground network and the invasion breaking-up activity that utilizes information network to carry out.Once blank pipe infosystem is attacked, radio communications system is interfered, the system failure etc., flight safety can be made to be subject to serious threat, the orchestration of blank pipe will be had a strong impact on, and the lighter can cause flight to operate interruption normally, and severe one can critical flight safety; Serious harm public interest and national security.
The important support point of whole aviation development is not only by blank pipe, and is the important duty of any one sovereign state.So want the relation of conscientiously correctly process blank pipe and air defence, strengthen blank pipe air defence My Perspective On The Co-building, make to organically combine between system, interconnect, information sharing and operation are efficiently.Any Information Security Risk all likely will directly affect the safe and stable operation of whole Civil Aviation System, affect the normal work running of civil aviaton, even may injure country, people life property safety, so blank pipe information security is the major issue being directly connected to national economy.
Being an important technology of Prevention-Security to the security evaluation of air traffic control system, is also the important component part of information security engineering.By safety assessment, scientifically analyze the security status of blank pipe infosystem, evaluation is made to the general safety situation of system, takes preventive measures.Setting up sound Civil Aviation ATM security of system assessment indicator system is implement the important guarantee of civil aviaton's information security strategy, by security assurance information assessment indicator system, united analysis and aspect ratio are carried out comparatively to China's aeronautical information system and core business system, conclusion China's aviation information safety defense situation being made to quantification will be contributed to, for CAAC provides decision support, to the planning of China's aviation Information Security Construction, the input of Information Security Construction, and even information security management policy making, the research and development of information security technology is all significant.
Need certain foundation and guidance method to the assessment of security risk, the current evaluation method obtaining comparing widespread use has: Delphi method, fault tree analysis, analytical hierarchy process, principal component analysis (PCA), DEA Method, fuzzy analysis and gray theory etc.
Delphi method (Delrhimethod) is not also referred to as Experts consultation method, result from sciemtifec and technical sphere at first, be applied to the prediction in any field afterwards gradually, as military forecasting, population forecast, health care prediction, managed and demand forecast, educational forecasting etc.In addition, be also used for carrying out evaluating, decision-making, managing communication and planning.Delphi method is a kind of Qualitative Forecast Methods, by the method that back-to-back group decision-making is seeked advice from, group member works alone separately, then with system, independently mode their judgement comprehensive, overcome the shortcoming for some authority is driven, reduce the psychological pressure of respondent, forecasting reliability is increased.
Fault tree analysis (FaultTreeAnalysis) model is proposed in 1961 by the Waston.H.A of Bell phone testing laboratory, as the mathematical model of analytic system reliability, has now become fairly perfect systems reliability analysis technology.Fault Tree Analysis can be divided into quantitative and qualitative analysis two kinds of modes.The qualitative analysis of fault tree is exactly the minimal cut set by asking fault tree, obtains whole fault modes of top event, and to find the weakest environment in system architecture or most critical position, the key position of concentrating strength on minimal cut set finds is strengthened.
Analytical hierarchy process (AnalyticHierarchyProcess) is put forward in 1970's by the operational research expert Satie of famous American, is a kind of Multi-objective Decision Analysis method of combination of qualitative and quantitative analysis.At present, the AHP educational circles that planned strategies for is considered as simple and effective Multiobjective Decision Making Method.The range of application of AHP is expanding gradually, up to now, AHP has been applied to the analysing and decision in a lot of field, as: economic analysis and plan, human behavior science, health care, accounting, sociology, education, the talent, military commanding, geography, operational research Methods evaluation, architecture, scientific technological advance, environment, law etc.AHP, as a kind of policy-making thought mode, plays a part more and more important to the various decision process of people.
Principal component analysis (PrincipalComponentAnalysis) is a branch of multivariate statistical analysis, and be applied to nonrandom vector by Karl, Pearson before this, then it has been generalized to random vector by Hotelling.Principal Component Analysis Method is application mathematical statistics and linear algebra knowledge, the former random vector relevant by its component, by means of an orthogonal transformation, change into the incoherent new random vector of its component, and using variance estimating as quantity of information, dimension-reduction treatment application decision analysis and functional analysis knowledge again being carried out to new random vector, by constructing suitable cost function, further low-dimensional system being changed into unidimensional system.Principal Component Analysis Method, with features such as the objectivity of the terseness of its theory, tax power, is widely used in the evaluation of numerous object in the fields such as economy, society, science and education, environmental protection.
DEA (DataEnvelopmentAnalysis) is based on relative efficiency concept, drops into a kind of new method of carrying out relative effectiveness or benefit evaluation with the unit (department or enterprise) of multi objective output to identical type according to multi objective.Since 1978, by the famous scholar of planning strategies for Cha Ensi, Ku Bai and Lodz, first C2R model is proposed and for since evaluating interdepartmental relative effectiveness, DEA method constantly improves, often be applied to the evaluation of enterprise operation overall efficiency, industry production situation, community service department facility etc. particularly at the community service department to non-simple profit, as school, hospital, the evaluation aspect of some cultural facility etc. is considered to a kind of effective method.
Nineteen sixty-five, U.S. kybernetics scholar L.A.Zadeh professor has delivered the famous paper of " fuzzy set " (FuzzySet) this section on " InformationandControl " magazine, this concept of proposition subordinate function describes the middle transition of phenomenon difference, thus breaches in classical set theory the absolute relation belonging to or do not belong to.Zadeh teaches the work of this initiative, indicates one of mathematics new branch---the birth of fuzzy mathematics.Fuzzy theory is also for Comprehensive Evaluation Problem provides a kind of new method.Application fuzzy theory is set up and is evaluated mathematical model, evaluation index quantification qualitatively can be made, quantitative fuzzy evaluation mark sense accuracy is approached, evaluation method is made to have more science, practicality, be widely used in the various fields such as society, economy, military affairs, engineering at present, obtain a large amount of achievements in research.
Gray system theory is that Chinese scholar Deng Julong teaches and first proposes in nineteen eighty-two, through the development of more than 20 years, substantially sets up the structural system of a new branch of science.So-called gray system to refer in system not only adularescent parameter (known parameters) but also has the system of black parameter (unknown parameter), and its research contents comprises the quantification, modeling, prediction, decision-making, control etc. of objective things.Gray system theory is the theory from the non-completeness research of information and process complication system, it is not leave for Study system from the rule that internal system is special, but by the observational data of a certain level of system in addition mathematics manipulation, reach and understand the mechanism such as internal system variation tendency, mutual relationship on higher level.Under the drive of gray system theory, also in succession create grey geology, grey breeding, gray control theory, grey-chaos theory, regional economy Grey System Analysis, grey axiology, grey combine a collection of emerging cross disciplines such as anti-.
In evaluation problem, the relation overwhelming majority between objective attribute target attribute is nonlinear relationship, and general method is difficult to this relation of reflection.The information source of many problems is imperfect, and evaluation rule is usually conflicting, and irrationality can follow sometimes.People are difficult to each mutuality of objectives of description scheme exactly usually, more cannot express weight allocation between them with quantitative relation formula, if existing scheme and evaluation result thereof can be utilized, feature according to given new departure just directly can make evaluation to scheme, then not only can reduce artificial uncertain factor, improve the accuracy of evaluation result, greatly can also alleviate the burden of estimator.
Nerual network technique can solve the problem effectively.The Nonlinear Processing ability of neural network breaches the limitation of the existing evaluation method based on linear process; General evaluation method is ambiguous in information, imperfect, exist in the complex environments such as contradiction and be often difficult to application, and nerual network technique then can cross over this obstacle.
Summary of the invention
First the present invention analyzes the demand for security of blank pipe, proposes the blank pipe infosystem 3 layers of neural network assessment models based on BP neural network.Then according to the BP neural network model set up, using blank pipe infosystem primary safety index as training sample, the data provided are by learning and training the inner link found out between constrained input, with the BP network trained, blank pipe infosystem is assessed, and compare with traditional appraisal procedure.Experimental result shows, this network has very strong adaptivity and fault-tolerant ability, this model is used for the safety assessment of blank pipe infosystem, is consistent with actual result, and has very large advantage and potential.
The object of the invention is, overcome the impact of the deficiencies in the prior art and artificial subjective factor.Because in evaluation problem, the relation overwhelming majority between objective attribute target attribute is nonlinear relationship, and general method is difficult to this relation of reflection; The information source of many problems is imperfect, and evaluation rule is usually conflicting, and irrationality can follow sometimes; People are difficult to each mutuality of objectives of description scheme exactly usually, more cannot express weight allocation between them with quantitative relation formula, if existing scheme and evaluation result thereof can be utilized, feature according to given new departure just directly can make evaluation to scheme, then not only can reduce artificial uncertain factor, improve the accuracy of evaluation result, greatly can also alleviate the burden of estimator.Nerual network technique can solve the problem effectively.The Nonlinear Processing ability of neural network breaches the limitation of the existing evaluation method based on linear process; General evaluation method is ambiguous in information, imperfect, exist in the complex environments such as contradiction and be often difficult to application, and nerual network technique then can cross over this obstacle.Utilize the method that neural network is assessed, can potential safety hazard and risk in Timeliness coverage blank pipe infosystem, input for safety of making rational planning for, takes corresponding safe precaution measure, increases economic efficiency.
Realizing technical solution of the present invention is: according to the feature of blank pipe infosystem, set up neural network assessment models.The security of feature to system according to network is analyzed, and set up the network model being applicable to blank pipe infosystem, its concrete performing step is:
1) as shown in Figure 1, this model is a multi input, single system exported to the safety assessment neural network model of the blank pipe infosystem of building.The input of this model is 12 evaluation index values of blank pipe infosystem, and evaluation index as shown in Figure 2.Replace traditional appraisal procedure with the BP network trained, provide assessment result by neural network.
2) the choosing of evaluation index
Civil Aviation ATM system is a complexity and huge system, relate to the data information running the aspects such as relevant communication, navigation, supervision, meteorology, information, air traffic control to blank pipe, and related hardware, as communication facilities, communication media, radar navigation set, meteorological equipment, navigational intelligence equipment etc.By the layering of system orderliness, it is the prerequisite of carrying out safety assessment.
Facts have proved, a good air traffic control system safe evaluation method should meet following requirement: evaluation index can reflect situation and the technical quality feature of air traffic control system all-sidedly and accurately; Evaluation model is simple and clear, workable, is easy to grasp; Evaluation conclusion can reflect the rationality of air traffic control system, integrity and safe reliability; The data adopted in evaluation are easy to obtain, and data processing work amount is little; Each evaluation index has clear and definite evaluation criterion.Based on above condition, from communication, navigation, monitor, robotization 4 aspects have chosen 12 main influence factors as evaluation index, constitute a scientific and reasonable air traffic control system assessment indicator system.
The index affecting ATM safety mainly comprises communication, navigation, monitors, robotization four aspects.Satellite system that specific targets are respectively (1), (2) VHF (VeryHighFrequency, very high frequency(VHF)) system, (3) interior telephone system, (4) DVOR (DopplerVHFOmnidirectionalRange, Doppler VHF omnirange) system, (5) ILS (InstrumentLandingSystem, instrument landing system) system, (6) DME (DistanceMeasuringEquipment, Distnace determination device) system, (7) SSR (SecondarySurveillanceRadar, secondary surveillance radar) system, (8) ADS (AutomaticDependentSurveillance, automatic dependent surveillance) system, (9) GPS (GlobalPositionSystem, GPS) system, (10) flight planning, (11) flight information, (12) supervisory system.The comprehensive assessment index system of blank pipe infosystem as shown in Figure 2.
3) division of safe class
Usually assessment result is divided into four grades, represents very safe, safer respectively, dangerous, danger close.Shown in table 1:
Table 1 safety status classification table
Very safe | Safer | Dangerous | Danger close |
0.85-1 | 0.75-0.85 | 0.6-0.75 | 0-0.6 |
A | B | C | D |
4) neuron number of BP network is determined
The input layer number of BP network depends on the number of the safety indexes affecting blank pipe infosystem.Can be obtained by the Security Evaluation Model of blank pipe infosystem, its number is 12.And its output layer is exactly the safety indexes number weighing blank pipe infosystem, its number is 1, and net result is safe class.The function newrb in Matlab is utilized to create a BP neural network.
5) BP network hidden neuron is chosen
For the design of network, wherein in hidden layer, neuronic number affects the test performance of network to a great extent.The defeated people's layer of this network and hidden layer, and the transport function between hidden layer and output layer adopts logarithmic function; Consider specification and the learning time of network, select Trainlm function to train function, maximum train epochs epochs is 1000; The least error goal of setting is 0.01; Show is 20.Other parameters are default value.
6) training sample is chosen
Training sample is chosen 12 groups of blank pipe stations and is adopted data in fact, and training sample is as shown in table 2, and by repetition test, the performance finally choosing network when neuronic number is 25 in hidden layer is best.The service check of BP neural network.Matlab is utilized to emulate network.Network, through initialization, utilizes function Trainlm to train network, and after training to 3 steps, network error reaches the error requirements of setting, and training terminates, as shown in Figure 5.
Table 2 training sample
7) network performance inspection
In order to check the performance of the rear network of training further, further simulation analysis is done to training result below.Utilize postreg function can export the Output rusults of network simulation and target and do nonlinear regression analysis, and obtain both related coefficients, thus can as the distinguishing rule of network training result quality.We utilize the linear regression analysis between the simulation data vector target vector of network, and the related coefficient that the target vector obtained is exported network is as the important evaluation mark of network performance.If network performance is good, the network analog value so obtained should be equal with network real output value, is namely on the diagonal line of coordinate axis first quartile, intercept equals 0, slope equals 1, and degree of fitting equals 1, usually gets degree of fitting R and be greater than 0.80 just passable in practical application.Finally obtain BP neuron network simulation value and actual exports between Nonlinear regression equation be: A=0.998T+ (-0.00418) (R=0.99), result is satisfied with, and network performance is fine.As shown in Figure 6.
8) network test
Utilize the above network trained below, the performance of network tested, after getting four groups as test sample book, test sample book is as shown in table 3.
Table 3 test sample book
As can be seen from the test result of network, not only demonstrate the network feasibility of design, and the stable performance of network, accuracy is high, and error is very little, and test result and actual result match.
The self-learning capability that this network has greatly facilitates again memory and the extraction of knowledge; Network, by study, can extract comprised rule, association's process particular problem, and can carry out completion to Incomplete information from typical example.From the angle evaluated, neural network is by the study to existing program and evaluation result thereof, the experience of implicit people wherein, knowledge and to intuitive thoughts such as the views of each target importance can be obtained, once be used for evaluating, network just can reproduce these experiences, knowledge and intuitive thought, rational judgement is made to challenge, had both embodied the subjective judgement of people thus, greatly reduce again the impact of disadvantageous human factor in evaluation procedure.Visible, neural network is the effective way of Multi-attribute synthetic evaluation.
9) assessment result explanation
BP neural network has very strong Approximation.The every content comprised with " safe and reliable, the economical rationality " of air traffic control system and its definition is for foundation, from the main composition communication of blank pipe body system, navigation, supervision and automated system four levels establish main 12 typical evaluation indexes of air traffic control system, adopt the assessment indicator system of set of system, set up the air traffic control system Model for Safety Evaluation based on BP neuroid, and solve.The test result of this model shows, the air traffic control system assessment indicator system adopted reflects the situation of air traffic control system well, and the BP network algorithm adopted meets the nonlinear characteristic of air traffic control system, may be used for the safety evaluatio of the complicated air traffic control systems such as civil aviaton.
Accompanying drawing explanation
Fig. 1 ATM safety assessment neural network model figure
Fig. 2 blank pipe infosystem evaluation index system figure
Fig. 3 BP network model figure
Fig. 4 BP flow through a network figure
The convergence of Fig. 5 network training, error curve diagram
The non-linear regression figure of Fig. 6 network
Embodiment
Shown in Fig. 2, Civil Aviation ATM system is a complexity and huge system, relate to the data information running the aspects such as relevant communication, navigation, supervision, meteorology, information, air traffic control to blank pipe, and related hardware, as communication facilities, communication media, radar navigation set, meteorological equipment, navigational intelligence equipment etc.As being evaluated object, air traffic control system is made up of multiple core system, belongs to typical multi input, multiple-target system.By the layering of system orderliness, it is the prerequisite of carrying out safety evaluation.
After having had the evaluation index system of blank pipe infosystem, assess to the security of whole blank pipe infosystem, in conjunction with Artificial Neural Network, Fig. 1 is the illustraton of model to whole system evaluation process.
Formulate evaluation grade standard: according to the feature of blank pipe infosystem, be A, B, C, D fourth class by safety status classification.Difference correspondence is very safe, safer, dangerous, danger close.The corresponding corresponding point number interval of each grade of model, the comprehensive assessment score value of gained is higher, illustrates that the security of whole blank pipe infosystem is higher.
Network evaluation mainly adopts BP network model, i.e. error backpropagation algorithm.BP network model is that a kind of forward solves, reverse propagated error reach the network model connecting flexible strategy between amendment network layer, and it is divided into input layer, hidden layer and output layer usually, and wherein hidden layer also may more than one.In forward-propagating, input signal, from input layer by transforming function transformation function to hidden layer successively forward-propagating, then, network is according to the size of training error, by obtaining the result of output layer compared with desired output, if deviation exceeds allowed band, then automatically regulate weights and threshold, namely the inverse of error transmits, output error signal is reduced, thus make network export force into desired output, usually reach error mean square difference minimum till.The study of network broadcasts repeatedly alternately the realizing against propagation with error by the saequential transmission of pattern.Generally, BP network model adopts the strategy of momentum method and learning rate adjustment, thus improves the performance of network, decreases it and be absorbed in local minimum and improve speed of convergence.As shown in Figure 3, detailed derivation slightly for BP network model.
For whole network algorithm realization flow process as shown in Figure 4.The performing step of BP back-propagation algorithm is as follows:
Step 1. provides input information vector P and object vector T;
Step 2. carries out standardization to input P;
Step 3. calculates the actual output of hidden layer and output layer
Step 4. asks object vector and the actual deviation exported; If total P is to training sample, the corresponding different sample of network has different errors
can using whole sample output error square to carry out cumulative again square as total output error, also can by total output error of representative network maximum in all errors, more employing root-mean-square errors in reality
as the total error of network.
If step 5. error in requiring, then forwards step 10 to;
Step 6. calculates hidden layer elemental error;
Step 7. asks error gradient;
Step 8. pair weights and bias is modified, and error signal is returned along original connecting path, by revising the neuronic weights of each layer, successively propagating to input layer and going to calculate.Again through forward-propagating process, these two processes carry out making error signal minimum repeatedly;
Step 9. gets back to step 3;
Whether step 10. calculates whole error and meets the demands, and then learns, otherwise get back to step 6 as met;
There are two kinds of weighed value adjusting methods in actual applications at present.As can be seen from above step, in standard BP algorithm, often input a sample. all will return error and adjust weights, this weighed value adjusting method to each sample training in rotation is also called single sample training.Due to single sample training follow be live in the present " selfish departmentalism " and principle. the error only produced for each sample adjusts, and turns round and look at this and loses quilt unavoidably, the number of times of whole training is increased, causes speed of convergence excessively short.Another kind method is after all sample inputs. the total error E of computational grid
always:
Then calculate the error signal of each layer according to total error and adjust weights, the batch processing mode of this cumulative errors is called batch (Batch) training or cycle (Epoch) training.Because batch training have followed to reduce " collectivism " principle that global error is target, thus can ensure that total error is to reduction power to change.When sample number is more, batch training is than fast convergence rate during single sample training.Training process flow diagram is as figure.The program of this problem adopts batch BP algorithm flow of training.
After establishing the neural network assessment models of blank pipe infosystem, BP error backpropagation algorithm is utilized to carry out training and testing to network, when the scope that network training allows to error, network training terminates, by test data, the network trained is tested, when test result meets the demands, show that network performance is good, can be applicable in the assessment of blank pipe infosystem.Training and testing for network adopts MATLAB instrument to carry out emulating and analyzing.Emulation and analysis result are as Fig. 5 and Fig. 6.Concrete implementation procedure is as follows, wherein introduces the primary function related to.
Utilize matlab instrument, in conjunction with above algorithm, introduce the primary function that this program is used.Utilize newff () function creation network, function net.iw{1,1} and net.b{1} are initial weight and threshold function table.Function net.trainParam.show function is figure display interval; Function net.trainParam.lr is learning rate; Function net.trainParam.ePochs is maximum permission step number; Function net.trainParam.goal is permissible error scope; Train () is network training function; Function sim () is network simulation function; Function postreg () realizes the nonlinear fitting to network.
The blank pipe infosystem neural network model finally utilizing training and testing good carries out comprehensive assessment to blank pipe information system security, finally exports evaluation grade and provides corresponding assessment report.Make corresponding strick precaution and innovative approach according to assessment report, risk can be reduced to minimum level and even avoid dangerous generation, take preventive measures.And improve the overall economic benefit of air traffic control system, national product is played to the effect of promotion.
Claims (1)
1. the air traffic control atm information system security assessment method based on ANNBP model, it is characterized in that air traffic control (ATM, AirTrafficManagement, being called for short: blank pipe) infosystem comprises four major parts, is communication system, navigational system, surveillance and automated system four Iarge-scale system respectively, neural network ANN (ArtificialNeuralNetwork) is based on from the index affecting blank pipe infosystem, the input of 12 specific targets as neural network of four Iarge-scale system securities will be affected in blank pipe infosystem, be respectively (1) satellite system, (2) VHF (VeryHighFrequency, very high frequency(VHF)) system, (3) interior telephone system, (4) DVOR (DopplerVHFOmnidirectionalRange, Doppler VHF omnirange) system, (5) ILS (InstrumentLandingSystem, instrument landing system) system, (6) DME (DistanceMeasuringEquipment, Distnace determination device) system, (7) SSR (SecondarySurveillanceRadar, secondary surveillance radar) system, (8) ADS (AutomaticDependentSurveillance, automatic dependent surveillance) system, (9) GPS (GlobalPositionSystem, GPS) system, (10) flight planning, (11) flight information, (12) supervisory system, neural net method is adopted to carry out safety assessment to described blank pipe infosystem index, wherein neural network algorithm adopts error back propagation (BackPropagation, BP) algorithm, set up the neural network assessment models of blank pipe infosystem, whole system is assessed, finally provides assessment result and assessment report,
Described neural network model assessment comprises the following steps: the safety comprehensive evaluation index system 1) determining blank pipe infosystem, blank pipe infosystem is made up of four Iarge-scale system, wherein affects the evaluation index system of described 12 indexs as blank pipe infosystem of this four Iarge-scale system security; 2) the evaluation grade standard of blank pipe infosystem is formulated; 3) neural network model being applicable to blank pipe infosystem is set up; 4) choose described 12 achievement datas and input data as network, adopt BP algorithm to train network; 5) by repetition training, and by after simulation analysis, the optimum network structure in selection training process is as network training model; 6) choose other several groups of data, non-training data data as the test data of network, and is tested network performance; 7) comprehensive assessment is carried out to whole air traffic control system, provide assessment result and assessment report; 8) Problems existing and suggestion are proposed;
Described BP algorithm of neural network, wherein weighed value adjusting is the use in conjunction with gradient descent algorithm:
1. network error
When network input and desired output do not wait, there is output error E, be defined as follows:
Wherein, d represents desired output, O represent sample be input to neural network after calculating output valve, have l node at output terminal, d
krepresent the desired output of k node, o
krepresent the calculating output valve of k node;
Be expanded to input layer further to have:
Wherein, net
jthe weighted sum of neuronic output before representing being input as of Current neural unit, v
ijrepresent the weights of input layer to hidden layer, ω
jkrepresent the weights of hidden layer to output layer, f (∑ v
ijx
i) represent hidden layer transport function, x
irepresent input value;
2. weighed value adjusting
Network error originated from input is each layer weights ω
jk, v
ijfunction, therefore adjust weights can change error E, adjustment weights principle be that error is constantly diminished, the adjustment amount of weights therefore should be made to be directly proportional to the Gradient Descent of error, that is:
In formula, negative sign represents Gradient Descent, and constant η ∈ (0,1) represents scale-up factor, learning rate is reflected in training, BP algorithm belongs to δ learning law class, is the gradient descent algorithm of error, and the weighed value adjusting formula of the BP learning algorithm of three-layer network is:
Wherein, Δ v
ijrepresent the weights variable quantity of input layer to hidden layer, Δ ω
jkrepresent the weights variable quantity of hidden layer to output layer,
represent that BP neural network is connected to the weights error on output layer neurode k,
represent that BP neural network is connected to the weights error on hidden layer neurode j, y
jrepresent the output valve of hidden layer node, x
irepresent the input value of neural network;
12 indexs of the information system integrated evaluation index system of described blank pipe, as the input of network, and screen corresponding blank pipe service data, and these data are normalized, the neuron of BP algorithm all adopts sigmoid transfer function, the absolute value because inputting can be prevented after conversion excessive and make neuron output saturation, then making weighed value adjusting enter the flat region of error surface; The output of sigmoid transfer function is [0,1] or [-1,1] between, as the output data of teacher signal if do not carried out conversion process, the output component absolute error that numerical value certainly will be made large is large, and the output component absolute error that numerical value is little is little, only for the total error adjustment weights exported during network training, in total error, consequently account for the little output component relative error of share comparatively large, therefore will carry out change of scale to output quantity; In addition, when inputting or each component dimension of input vector is different, tackling different components and converting respectively in its span, when each component physical significance identical and be same dimension time, maximal value x should be determined in whole data area
maxwith minimum value x
mincarry out unified conversion process;
Value inputoutput data being transformed to [0,1] interval commonly uses following transforming function transformation function:
Wherein x
irepresent inputoutput data, x
minrepresent the minimum value of variation range, x
maxrepresent the maximal value of variation range; Using the real input of the data after process as network, adopt BP algorithm of neural network to train network, training is until error stops after meeting allowed band, and now network training terminates;
To described blank pipe infosystem, adopt described blank pipe infosystem neural metwork training network, adopt proper data to test network; If test result is satisfied, network performance is good, if test result is unsatisfied with, returns step 4) re-training network; Through repeatedly repetition training, find the neural network structure of the most applicable blank pipe infosystem, and set up neural network assessment models;
Adopt described BP algorithm of neural network to carry out Information Security Evaluation to described blank pipe infosystem, provide comprehensive assessment result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201010519836.4A CN102456158B (en) | 2010-10-26 | 2010-10-26 | Based on the air traffic control atm information system security assessment method of ANN BP model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201010519836.4A CN102456158B (en) | 2010-10-26 | 2010-10-26 | Based on the air traffic control atm information system security assessment method of ANN BP model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN102456158A CN102456158A (en) | 2012-05-16 |
CN102456158B true CN102456158B (en) | 2015-11-11 |
Family
ID=46039329
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201010519836.4A Active CN102456158B (en) | 2010-10-26 | 2010-10-26 | Based on the air traffic control atm information system security assessment method of ANN BP model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN102456158B (en) |
Families Citing this family (24)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103226900B (en) * | 2013-03-21 | 2015-10-28 | 北京工业大学 | A kind of space domain sector division methods based on weighted graph model |
US9540118B2 (en) | 2014-11-10 | 2017-01-10 | Federal Express Corporation | Risk assessment framework |
CN105095408A (en) * | 2015-07-09 | 2015-11-25 | 百度在线网络技术(北京)有限公司 | Method and apparatus for judging reliability of network expert |
CN105118333B (en) * | 2015-09-30 | 2017-12-15 | 中国民用航空总局第二研究所 | A kind of air traffic control analog simulation method for detecting abnormality and device based on multiple regression model |
CN105261241B (en) * | 2015-09-30 | 2017-11-07 | 中国民用航空总局第二研究所 | Air traffic control analog simulation method for detecting abnormality and device based on Hopfield neural network |
CN105575186B (en) * | 2015-12-08 | 2018-10-19 | 成都民航空管科技发展有限公司 | Air traffic control automation system secondary radar answering machine encodes fixed direction allocation method |
CN106897474B (en) * | 2015-12-21 | 2021-03-26 | 中国航空工业集团公司西安飞机设计研究所 | Method for evaluating control performance of aircraft cabin pressure control system |
CN106027550B (en) * | 2016-06-29 | 2019-04-12 | 北京邮电大学 | A kind of defence policies systematic analytic method and device |
CN107729949A (en) * | 2017-11-03 | 2018-02-23 | 中国民航大学 | A kind of cluster and analysis method of control operation sub-health state |
CN108133623B (en) * | 2018-01-31 | 2020-09-01 | 中国民航大学 | Method for establishing air cross point grading index |
CN108459582B (en) * | 2018-03-01 | 2021-03-02 | 中国航空无线电电子研究所 | IMA system-oriented comprehensive health assessment method |
CN110689032A (en) * | 2018-07-04 | 2020-01-14 | 北京京东尚科信息技术有限公司 | Data processing method and system, computer system and computer readable storage medium |
CN109409587A (en) * | 2018-10-09 | 2019-03-01 | 南京航空航天大学 | A kind of airport excavated based on weather data is into traffic flow forecasting method of leaving the theatre |
CN109495296B (en) * | 2018-11-02 | 2022-05-13 | 国网四川省电力公司电力科学研究院 | Intelligent substation communication network state evaluation method based on clustering and neural network |
CN109544999B (en) * | 2019-01-14 | 2021-03-26 | 中国民航大学 | Cloud model-based air traffic network reliability evaluation method |
CN111091266A (en) * | 2019-11-12 | 2020-05-01 | 青岛民航空管实业发展有限公司 | Grade determination method and device of empty management station and storage medium |
CN110928752A (en) * | 2019-11-14 | 2020-03-27 | 青岛民航空管实业发展有限公司 | Method, device and equipment for evaluating health degree of air traffic control station |
CN111610517B (en) * | 2020-06-09 | 2022-06-07 | 电子科技大学 | Secondary radar signal processing method based on deep four-channel network |
CN111610518B (en) * | 2020-06-09 | 2022-08-05 | 电子科技大学 | Secondary radar signal denoising method based on depth residual separation convolutional network |
CN111785095B (en) * | 2020-07-31 | 2021-06-01 | 北京航空航天大学 | Method for forming monitoring performance evaluation index |
CN112069740B (en) * | 2020-09-21 | 2023-11-24 | 吴学松 | Rock burst risk assessment method and platform |
CN113256143A (en) * | 2021-06-07 | 2021-08-13 | 中国传媒大学 | Assessment method for media-melting organization, electronic equipment and readable storage medium |
CN113850350B (en) * | 2021-11-30 | 2022-04-22 | 中哲国际工程设计有限公司 | Urban building land intelligent planning system and method |
CN114818396B (en) * | 2022-06-29 | 2022-09-20 | 湖南大佳数据科技有限公司 | Network security shooting range system and drilling method for satellite navigation system |
-
2010
- 2010-10-26 CN CN201010519836.4A patent/CN102456158B/en active Active
Non-Patent Citations (5)
Title |
---|
BP神经网络算法改进及应用研究;黄丽;《中国优秀硕士学位论文全文数据库信息科技辑》;20100131;I140-40 * |
I. S. SAEH, A. KHAIRUDDIN.STATIC SECURITY ASSESSMENT USING ARTIFICIAL.《POWER AND ENERGY CONFERENCE, 2008, PECON 2008. IEEE 2ND INTERNATIONAL》.2008, * |
中国民用航空总局.民用航空空中交通通信导航监视设备使用.《民用航空空中交通通信导航监视设备使用》.2002, * |
基于BP神经网络的民用航空航段安全风险评估;王浩锋;《信息与电子工程》;20101025;第8卷(第5期);612-615 * |
航空企业基于SHEL模型的神经网络安全评价研究;王起全;《中国安全科学学报》;20100228;第20卷(第2期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN102456158A (en) | 2012-05-16 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN102456158B (en) | Based on the air traffic control atm information system security assessment method of ANN BP model | |
Turskis et al. | A hybrid model based on fuzzy AHP and fuzzy WASPAS for construction site selection | |
De Ambroggi et al. | Modelling and assessment of dependent performance shaping factors through Analytic Network Process | |
CN102496069B (en) | Cable multimode safe operation evaluation method based on fuzzy analytic hierarchy process (FAHP) | |
Nearing et al. | Benchmarking and process diagnostics of land models | |
Kim | An assessment of deforestation models for reducing emissions from deforestation and forest degradation (REDD) | |
CN106411854A (en) | Network security risk assessment method based on fuzzy Bayes | |
Li et al. | An assessment method of operator’s situation awareness reliability based on fuzzy logic-AHP | |
Liu et al. | A risk assessment method based on RBF artificial neural network-cloud model for urban water hazard | |
CN107491694A (en) | Method for quantitative evaluation SCADA system information security fragility | |
Zichun | The BP artificial neural network model on expressway construction phase risk | |
Pessoa et al. | Cost forecasting of public construction projects using multilayer perceptron artificial neural networks: a case study | |
Guo et al. | Delphi method for estimating membership function of uncertain set | |
Kvyatkovskaya et al. | Methodology of a support of making management decisions for poorly structured problems | |
Dong et al. | Analysis of performance evaluation model for higher education informatization by means of PLS-BP model | |
Sun et al. | A BP neural network model based on genetic algorithm for comprehensive evaluation | |
Zhang | Benefit evaluation of human resource management in agricultural enterprises based on convolutional neural network. | |
Ling et al. | Analyzing the relationship between urban macroeconomic development and transport infrastructure system based on neural network | |
Zhu | Big Data’s Analysis and Prediction Method of Art Education Based on the BP Neural Network | |
Huang | Modeling and Planning on Urban Logistics Park Location Selection Based on the Artificial Neural Network. | |
Adnan et al. | Improvement of the method of estimation and forecasting of the state of the monitoring object in intelligent decision support systems | |
Yang | Evaluation and Analysis of Multimedia Collaborative Building Design Relying on Particle Swarm Optimization Algorithm | |
Sohag et al. | Demand forecasting for a domestic airport-A case study | |
Hrinchenko et al. | Qualimetric approaches to assessing sustainable development indicators | |
Pu et al. | Research on the application of a high-dimensional Bayesian network in the evaluation of Ethnic-Chinese bilingual teaching |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
TR01 | Transfer of patent right | ||
TR01 | Transfer of patent right |
Effective date of registration: 20230505 Address after: Unit 516, Zone A, Comprehensive Building, No. 3 Huatian Road, Huayuan Industrial Zone, Tianjin Binhai High tech Zone, 300300 (settled in Innovation Service Company) Patentee after: TIANJIN YUN'AN TECHNOLOGY DEVELOPMENT Co.,Ltd. Address before: 300300 Tianjin city Dongli District North Road No. 2898 Patentee before: CIVIL AVIATION University OF CHINA |