CN112926739B - Network countermeasure effectiveness evaluation method based on neural network model - Google Patents
Network countermeasure effectiveness evaluation method based on neural network model Download PDFInfo
- Publication number
- CN112926739B CN112926739B CN202110265382.0A CN202110265382A CN112926739B CN 112926739 B CN112926739 B CN 112926739B CN 202110265382 A CN202110265382 A CN 202110265382A CN 112926739 B CN112926739 B CN 112926739B
- Authority
- CN
- China
- Prior art keywords
- layer
- neural network
- network
- countermeasure
- effectiveness
- 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
- 238000011156 evaluation Methods 0.000 title claims abstract description 32
- 238000003062 neural network model Methods 0.000 title claims abstract description 16
- 238000013528 artificial neural network Methods 0.000 claims abstract description 68
- 238000013210 evaluation model Methods 0.000 claims abstract description 44
- 238000012549 training Methods 0.000 claims abstract description 39
- 241000544061 Cuculus canorus Species 0.000 claims abstract description 21
- 230000008485 antagonism Effects 0.000 claims abstract description 16
- 238000010276 construction Methods 0.000 claims abstract description 13
- 238000000034 method Methods 0.000 claims description 22
- 210000002569 neuron Anatomy 0.000 claims description 17
- 235000005770 birds nest Nutrition 0.000 claims description 9
- 235000005765 wild carrot Nutrition 0.000 claims description 9
- 230000007123 defense Effects 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 5
- 238000000354 decomposition reaction Methods 0.000 claims description 5
- 238000007781 pre-processing Methods 0.000 claims description 5
- 244000000626 Daucus carota Species 0.000 claims description 4
- 230000000694 effects Effects 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 claims description 2
- 238000009825 accumulation Methods 0.000 abstract description 2
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 30
- 230000008569 process Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000003012 network analysis Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000002776 aggregation Effects 0.000 description 1
- 238000004220 aggregation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000005295 random walk Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a network countermeasure effectiveness evaluation method based on a neural network model, and relates to the technical field of network security. The invention builds a two-stage neural network countermeasure effectiveness evaluation model, avoids the complicated relationship in the combing index system, has strong self-learning, self-organizing and adapting ability, and can continuously and dynamically learn and train the model through training samples. By accumulation of historical samples, the challenge performance assessment model will have greater accuracy. In the learning of the neural network, an artificial intelligence algorithm-a cuckoo algorithm is adopted to find the optimal weight, the global searching capability is strong, the selection parameters are few, the convergence speed is extremely high, and the construction of the antagonism effectiveness evaluation model has higher efficiency.
Description
Technical Field
The invention relates to the technical field of network security, in particular to a network countermeasure effectiveness evaluation method based on a neural network model.
Background
With the evolution of informatization warfare, network antagonism plays an increasing role as a new combat effort in modern battlefield. The network countermeasure is that the fighter and the fighter use each element in the information system as the main fight object and the advanced information technology as the basic means to break down and destroy the enemy information system and protect the own information system. The network space countermeasure efficacy is comprehensively, reasonably and effectively evaluated, improvement of the weak links is facilitated, and the overall combat capability of the network space is improved.
At present, methods for evaluating the network space countermeasure efficacy mainly comprise social network analysis, a complex network, a hierarchical analysis method, an artificial neural network method and the like. Constructing a network space countermeasure effectiveness evaluation index system by using a social network analysis method, and obtaining the importance degree order of each index, wherein the influence weight value of each index on an evaluation target cannot be calculated; a networked index system framework is proposed by using a complex network theory, so that the aggregation relation between basic indexes and capability effects can be analyzed, but a final evaluation result cannot be obtained quantitatively; the analytic hierarchy process is used for evaluating the network space countermeasure efficacy, when the indexes are excessive, the data statistics are large, and the weight is difficult to calculate; when an artificial neural network method is used for constructing an evaluation model, a gradient descent method is generally adopted for correcting the weight coefficient, and the method has a low convergence rate in the learning process and is easy to fall into local optimum in the training process.
Disclosure of Invention
First, the technical problem to be solved
The invention aims to solve the technical problems that: how to design a method for constructing a network space countermeasure effectiveness evaluation model, so that the model construction is faster and more accurate.
(II) technical scheme
In order to solve the technical problems, the invention provides a network countermeasure efficacy evaluation method based on a neural network model, which comprises the following steps:
step 1, constructing a secondary neural network countermeasure effectiveness evaluation model by carrying out multi-level decomposition on network countermeasure comprehensive effectiveness indexes;
and 2, training a secondary neural network countermeasure effectiveness evaluation model based on a cuckoo algorithm.
Preferably, step 1 specifically includes:
(1) Constructing network countermeasure effectiveness evaluation index system
Layering the network countermeasure effectiveness into a comprehensive effectiveness layer, a capability element layer and an index element layer to form a network countermeasure effectiveness evaluation index system framework, wherein the uppermost layer is the comprehensive effectiveness layer, namely the network countermeasure comprehensive effectiveness; the middle layer is a capability element layer, namely the main capability decomposition of the network against comprehensive efficiency; the lowest layer is an index element layer, namely, judging each index on which each capability of the network against comprehensive efficiency depends;
(2) Construction of a model framework for evaluation of the effectiveness of a second-level neural network
Converting the network countermeasure effectiveness evaluation index system framework into a secondary neural network countermeasure effectiveness evaluation model, wherein each secondary neural network comprises an input layer, an hidden layer and an output layer; each neuron has an input connection and an output connection, each connection having a weight; the input layer of the first-stage neural network corresponds to an index element layer, and the output layer corresponds to a capability element layer; the input layer of the second-stage neural network corresponds to the capability element layer, and the output layer corresponds to the comprehensive performance layer.
Preferably, the step of constructing a framework of the second-level neural network performance evaluation model specifically includes:
(21) Construction of first-level neural network countermeasure effectiveness evaluation model framework
Index element of performance evaluation corresponding to input layer of first-stage neural networkLayer, input layer vector is defined as x 1 ,x 2 ,...,x M The hidden layer vector is defined as h 1 ,h 2 ,...,h L ,Wherein a is ij I.e. connecting the input layer neurons x i And hidden layer neuron h j Weight coefficient between, i=1,..m, j=1,.. L, M, L is a positive integer, the output function is defined as +.>Where j=1,.. then the layer vector is outputWhere i=1,..m, j=1, a., L, k=1..n, N is a positive integer, b jk Weights from hidden layer to output layer in the first-level neural network model;
(22) Construction of model framework for evaluating antagonism effectiveness of second-level neural network
The input layer of the second level neural network corresponds to the capability factor layer of the performance evaluation, and the input layer vector is defined as y 1 ,y 2 ,...,y N The hidden layer vector is defined as g 1 ,g 2 ,...,g P ,Wherein c kr I.e. connecting the input layer neurons y k And hidden layer neurons g r The weight coefficients in between, k=1..the term, N, r=1..the term, P are positive integers and the output function is defined as +.>Where r=1,..p, then the layer vector is outputWhere k=1,..n, r=1,.. r Is the implicit layer-to-output layer weight in the second level neural network model.
Preferably, step 2 specifically includes:
(1) Preprocessing the original sample
The original sample can be used as a training sample after being preprocessed, and the original sample is normalized by using a linear change method;
(2) Cuckoo algorithm optimization training
Initializing an objective function, a nest position and a minimum error; inputting a training sample obtained after preprocessing in the neural network antagonism effectiveness evaluation model, searching for an optimal nest position by using a cuckoo algorithm, generating a new weight by using the Lewy flight of the cuckoo algorithm and optimizing iteration, ending training when the absolute error of an actual output value and an expected value is smaller than a set minimum error, reserving a current optimal weight, and obtaining an optimal antagonism effectiveness evaluation model, wherein the optimal position is the optimal weight of the antagonism effectiveness evaluation model.
Preferably, the normalization processing of the original sample by using the linear variation method specifically includes: setting the original sample of the index element layer as x', and normalizing the training sample when the greater the index value is, the better the countermeasure efficacy isx′ min ,x′ max Respectively, the minimum and the maximum in x', when the index value is larger and the countermeasure efficacy is worse, the normalized training sampleLet the original sample of the ability factor layer be y', when the index value is bigger and the countermeasure effect is better, the normalized training sample is +.>y′ min 、y′ max Respectively, the minimum value and the maximum value in y', when the greater the index value is, the worse the countermeasure efficacy is, the normalized training sample is ∈ ->Comprehensive designThe original sample of the efficacy layer is E', and the normalized training sample is +.>E′ min 、E′ max The minimum and maximum values in E', respectively.
Preferably, the specific steps of the cuckoo algorithm optimization training are as follows:
1) Initializing an objective function
Wherein E is the actual output of the second-stage neural network, E d Y is the expected value of the integrated performance layer k For the actual output of the first level neural network,for the expected value of the capability element layer, the rejection probability P, P E [0,1 ] is initialized];
Initializing the positions of n bird's nest:
ω s (0) =[a (0) 11 ,a (0) 12 ,..,a (0) ML ,b (0) 11 ,b (0) 12 ,..,b (0) LN ,..,c (0) 11 ,c (0) 12 ,..,c (0) NP ,d (0) 1 ,d (0) 2 ,..,d (0) P ] T ,s=1,...,n;
2) Calculating an objective function value of each nest position, and selecting a nest with the optimal current objective function;
3) Keeping the optimal nest position of the previous generation objective function, and updating the nest position by utilizing the Lewy flight;
the updated formula of the position of the bird nest of the cuckoo is omega s (t+1) =ω s (t) +α.L (β), wherein ω s (t) Representing the position of the s-th nest at the t-th iteration; alpha generationA table step; l (β) obeys the lewye distribution:
0<β≤2,
wherein u, v obeys the normal distribution,
ω i' (t) 、ω j' (t) the position of any two bird nest is the t iteration;
4) Comparing the current position function value with the function value of the optimal nest position of the previous generation, if the current position function value is better, updating the current position function value into the current optimal function value, and if the current position function value is not better, reserving the optimal function value of the previous generation;
5) After the position is updated, a number r epsilon [0,1 ] is randomly generated]If r>P, P, omega s (t+1) Continuously updating, comparing the updated nest position function values, and calculating the global optimal position at the moment;
6) Judging whether the maximum iteration number or the minimum error requirement is met, if so, outputting a global optimal position, namely, each connection weight of the countermeasure effectiveness evaluation model, and if not, returning to the step 2) to continue iteration;
7) And (5) bringing the optimal weight into the neural network model to obtain an optimal secondary neural network antagonism effectiveness evaluation model.
Preferably, when the network countermeasure effectiveness evaluation index system is constructed, based on various common network information equipment combat demands in the network space attack and defense countermeasure field, the network countermeasure effectiveness is decomposed into various capability elements such as network reconnaissance, network attack, network defense and command decision, and each capability element is refined and divided into a plurality of index elements.
Preferably, the value range of the comprehensive efficiency E is 0-1.
Preferably, α=1 is taken.
The invention also provides application of the method in the technical field of network security.
(III) beneficial effects
The invention builds a two-stage neural network countermeasure effectiveness evaluation model, avoids the complicated relationship in the combing index system, has strong self-learning, self-organizing and adapting ability, and can continuously and dynamically learn and train the model through training samples. By accumulation of historical samples, the challenge performance assessment model will have greater accuracy.
In the learning of the neural network, an artificial intelligence algorithm-a cuckoo algorithm is adopted to find the optimal weight, the global searching capability is strong, the selection parameters are few, the convergence speed is extremely high, and the construction of the antagonism effectiveness evaluation model has higher efficiency.
Drawings
FIG. 1 is a topological structure diagram of a secondary neural network performance evaluation model of the present invention;
FIG. 2 is a general flow chart of the construction of a secondary neural network evaluation model in the present invention.
Detailed Description
For the purposes of clarity, content, and advantages of the present invention, a detailed description of the embodiments of the present invention will be described in detail below with reference to the drawings and examples.
The invention provides a network countermeasure effectiveness evaluation method based on a neural network model. According to the method, a performance evaluation model is built through two stages of neural networks, and each stage of neural network comprises an input layer, a hidden layer and an output layer. The second-stage neural network takes the effective value output by the first-stage neural network as network input, and the final output value is the network countermeasure comprehensive efficiency. And training samples are continuously trained by using a cuckoo algorithm, so that the weight of the neural network is continuously optimized, and the final evaluation model is more accurate. The invention is oriented to the evaluation of the comprehensive performance exerted by common red and blue parties in network attack and defense countermeasures.
The technical scheme for solving the technical problems generally comprises two steps: firstly, constructing a two-level neural network countermeasure effectiveness evaluation model, and decomposing network countermeasure comprehensive effectiveness indexes into a comprehensive effectiveness layer, a capability element layer and an index element layer in a multi-level manner, wherein the index element layer corresponds to an input value of a first-level neural network; the capability element layer corresponds to the output value of the first-stage neural network and is also the input value of the second-stage neural network; the comprehensive efficiency layer corresponds to the output value of the second-stage neural network. The topology of the secondary neural network performance evaluation model is shown in fig. 1. Training a secondary neural network countermeasure effectiveness evaluation model, learning a training sample by using a cuckoo algorithm, and continuously optimizing and iterating the weight until the optimized weight can meet that the error between the actual output and the target value of the neural network is smaller than the expected error.
FIG. 2 is a general flow chart of a two-level neural network evaluation model construction, and the method of the invention specifically comprises the following steps:
step 1, constructing a secondary neural network countermeasure efficacy evaluation model
(1) Construction of network countermeasure effectiveness evaluation index system framework
Classifying and layering the network countermeasure effectiveness into a comprehensive effectiveness layer, a capability element layer and an index element layer to form a network countermeasure effectiveness evaluation index system framework. The uppermost layer is a comprehensive performance layer, namely the network antagonism comprehensive performance; the middle layer is a capability element layer, namely the main capability decomposition of the network against comprehensive efficiency; the lowest layer is an index element layer, namely, each index on which each capability of the network against comprehensive efficiency depends is judged.
In this embodiment, the network countermeasure effectiveness is decomposed into various capability elements such as network reconnaissance, network attack, network defense, command decision and the like by researching various common network information equipment combat demands in the network space attack and defense countermeasure field, and each capability element is refined and divided into a plurality of index elements. The method does not need classification, only needs layering, and avoids the problems of complicated and unclear classification between the capability element layer and the index element.
(2) Construction of a model framework for evaluation of the effectiveness of a second-level neural network
And converting the network countermeasure effectiveness evaluation index system framework into a secondary neural network countermeasure effectiveness evaluation model. Each level of neural network comprises an input layer, an implicit layer and an output layer; each neuron has an input connection and an output connection, each connection having a weight; the input layer of the first-stage neural network corresponds to an index element layer, and the output layer corresponds to a capability element layer; the input layer of the second-stage neural network corresponds to the capability element layer, and the output layer corresponds to the comprehensive performance layer.
The method specifically comprises the following steps:
(21) And constructing a first-stage neural network antagonism effectiveness evaluation model framework.
The input layer of the first level neural network corresponds to the index element layer of the performance evaluation, and the input layer vector is defined as x 1 ,x 2 ,...,x M The hidden layer vector is defined as h 1 ,h 2 ,...,h L ,Wherein a is ij I.e. connecting the input layer neurons x i And hidden layer neuron h j Weight coefficient between, i=1,..m, j=1,.. L, M, L is a positive integer, the output function is defined as +.>Where j=1,.. then the layer vector is outputWhere i=1,..m, j=1, a., L, k=1..n, N is a positive integer, b jk For the implicit layer to output layer weights in the first level neural network model, the value of each neuron output needs to be multiplied by this weight and summed into the next neuron.
(22) And constructing a second-level neural network antagonism effectiveness evaluation model framework.
The input layer of the second-level neural network corresponds to the capability element layer of the performance evaluation, and the input layer is orientedThe quantity is defined as y 1 ,y 2 ,...,y N The hidden layer vector is defined as g 1 ,g 2 ,...,g P ,Wherein c kr I.e. connecting the input layer neurons y k And hidden layer neurons g r The weight coefficients in between, k=1..the term, N, r=1..the term, P are positive integers and the output function is defined as +.>Where r=1,..p, then the layer vector is outputWhere k=1,..n, r=1,.. r Is the implicit layer-to-output layer weight in the second level neural network model.
Step 2, training a secondary neural network countermeasure effectiveness evaluation model
(1) Preprocessing the original sample
The original sample needs to be preprocessed to be used as a training sample. And (3) normalizing the original sample row by using a linear change method, wherein the measurement units of all parameters in the network countermeasure effectiveness evaluation index system are different.
The normalization processing of the original sample by adopting the linear change method comprises the following steps: setting the original sample of the index element layer as x', and normalizing the training sample when the greater the index value is, the better the countermeasure efficacy isx′ min ,x′ max Respectively, the minimum and the maximum in x', when the index value is larger and the countermeasure efficacy is worse, the normalized training sample is ∈ ->Let the original sample of the ability element layer be y', when the index value is bigger and the countermeasure efficiency is better, the normalized training sampley′ min 、y′ max Respectively, the minimum value and the maximum value in y', when the greater the index value is, the worse the countermeasure efficacy is, the normalized training sample is ∈ ->Let the original sample of the comprehensive efficiency layer be E' and the normalized training sample beE′ min 、E′ max The minimum and maximum values in E' are respectively, the range of the comprehensive efficiency E is 0-1, and the closer the E value is 1, the better the comprehensive efficiency is represented.
(2) Cuckoo algorithm optimization training
Initializing an objective function, a nest position and a minimum error; inputting the training sample obtained after the processing in the neural network countermeasure effectiveness evaluation model, searching for the optimal nest position by using a cuckoo algorithm, utilizing the Levy flight of the cuckoo algorithm, generating new weight by optimizing iteration through a random walk mode of short-distance exploration and occasional long-distance walk, ending training when the absolute error of an actual output value and an expected value is smaller than a set minimum error, reserving the current optimal weight, and obtaining the optimal countermeasure effectiveness evaluation model, wherein the optimal position is the optimal weight of the countermeasure effectiveness evaluation model.
The optimization training of the cuckoo algorithm comprises the following specific steps:
1) Initializing an objective function
Wherein E is the actual output of the second-stage neural network, E d Y is the expected value of the integrated performance layer k For the actual output of the first level neural network,for the expected value of the capability element layer, the rejection probability P (the probability that the host finds a new coming egg and discards the egg) is initialized, P ε [0,1]。
Initializing the positions of n bird's nest:
ω s (0) =[a (0) 11 ,a (0) 12 ,..,a (0) ML ,b (0) 11 ,b (0) 12 ,..,b (0) LN ,..,c (0) 11 ,c (0) 12 ,..,c (0) NP ,d (0) 1 ,d (0) 2 ,..,d (0) P ] T ,s=1,...,n。
2) An objective function value for each nest location is calculated (which varies with the weight in the objective function) and the nest for which the current objective function is optimal is selected.
3) And (3) reserving the optimal nest position of the previous generation objective function, and updating the nest position by utilizing the Lewy flight type.
The updated formula of the position of the bird nest of the cuckoo is omega s (t+1) =ω s (t) +α.L (β), wherein ω s (t) Representing the position of the s-th nest at the t-th iteration; α represents the step size, typically taking α=1; l (β) obeys the lewye distribution:
wherein u, v obeys the normal distribution,
ω i' (t) 、ω j' (t) is the position of any two bird's nest at the t-th iteration。
4) And comparing the current position function value with the function value of the optimal nest position of the previous generation, updating to the current optimal function value if the current position function value is better, and if the current position function value is not better, reserving the optimal function value of the previous generation.
5) After the position is updated, a number r epsilon [0,1 ] is randomly generated]If r>P, P, omega s (t+1) And continuing to update, comparing the updated nest position function values, and calculating the global optimal position at the moment.
6) And judging whether the maximum iteration number or the minimum error requirement is met, and if so, outputting a global optimal position, namely, each connection weight of the countermeasure effectiveness evaluation model. If not, returning to the step 2) to continue iteration.
7) And (5) bringing the optimal weight into the neural network model to obtain an optimal secondary neural network antagonism effectiveness evaluation model.
The foregoing is merely a preferred embodiment of the present invention, and it should be noted that modifications and variations could be made by those skilled in the art without departing from the technical principles of the present invention, and such modifications and variations should also be regarded as being within the scope of the invention.
Claims (4)
1. The network countermeasure effectiveness evaluation method based on the neural network model is characterized by comprising the following steps of:
step 1, constructing a secondary neural network countermeasure effectiveness evaluation model by carrying out multi-level decomposition on network countermeasure comprehensive effectiveness indexes;
step 2, training a secondary neural network countermeasure effectiveness evaluation model based on a cuckoo algorithm;
the step 1 specifically comprises the following steps:
(1) Constructing network countermeasure effectiveness evaluation index system
Layering the network countermeasure effectiveness into a comprehensive effectiveness layer, a capability element layer and an index element layer to form a network countermeasure effectiveness evaluation index system framework, wherein the uppermost layer is the comprehensive effectiveness layer, namely the network countermeasure comprehensive effectiveness; the middle layer is a capability element layer, namely the main capability decomposition of the network against comprehensive efficiency; the lowest layer is an index element layer, namely, judging each index on which each capability of the network against comprehensive efficiency depends;
(2) Construction of a model framework for evaluation of the effectiveness of a second-level neural network
Converting the network countermeasure effectiveness evaluation index system framework into a secondary neural network countermeasure effectiveness evaluation model, wherein each secondary neural network comprises an input layer, an hidden layer and an output layer; each neuron has an input connection and an output connection, each connection having a weight; the input layer of the first-stage neural network corresponds to an index element layer, and the output layer corresponds to a capability element layer; the input layer of the second-stage neural network corresponds to the capability element layer, and the output layer corresponds to the comprehensive performance layer;
the step of constructing the framework of the secondary neural network efficiency evaluation model specifically comprises the following steps:
(21) Construction of first-level neural network countermeasure effectiveness evaluation model framework
The input layer of the first level neural network corresponds to the index element layer of the performance evaluation, and the input layer vector is defined as x 1 ,x 2 ,...,x M The hidden layer vector is defined as h 1 ,h 2 ,...,h L ,Wherein a is ij I.e. connecting the input layer neurons x i And hidden layer neuron h j Weight coefficient between, i=1,..m, j=1,.. L, M, L is a positive integer, the output function is defined as +.>Where j=1,.. output layer vector +.>Where i=1,..m, j=1, a., L, k=1..n, N is a positive integer, b jk Weights from hidden layer to output layer in the first-level neural network model;
(22) Construction of model framework for evaluating antagonism effectiveness of second-level neural network
The input layer of the second level neural network corresponds to the capability factor layer of the performance evaluation, and the input layer vector is defined as y 1 ,y 2 ,...,y N The hidden layer vector is defined as g 1 ,g 2 ,...,g P ,Wherein c kr I.e. connecting the input layer neurons y k And hidden layer neurons g r The weight coefficients in between, k=1..the term, N, r=1..the term, P are positive integers and the output function is defined as +.>Where r=1,..p, then output layer vector +.>Where k=1,..n, r=1,.. r Is the weight from the hidden layer to the output layer in the second-level neural network model;
the step 2 specifically comprises the following steps:
(1) Preprocessing the original sample
The original sample can be used as a training sample after being preprocessed, and the original sample is normalized by using a linear change method;
(2) Cuckoo algorithm optimization training
Initializing an objective function, a nest position and a minimum error; inputting a training sample obtained after preprocessing in a neural network antagonism effectiveness evaluation model, searching for an optimal nest position by using a cuckoo algorithm, generating a new weight by using the Lewy flight of the cuckoo algorithm and optimizing iteration, ending training when the absolute error of an actual output value and an expected value is smaller than a set minimum error, reserving a current optimal weight, and obtaining an optimal antagonism effectiveness evaluation model, wherein the optimal position is the optimal weight of the antagonism effectiveness evaluation model;
the original sample is normalized by adopting a linear change methodThe one-step treatment is specifically as follows: setting the original sample of the index element layer as x', and normalizing the training sample when the greater the index value is, the better the countermeasure efficacy isx′ min ,x′ max Respectively, the minimum and the maximum in x', when the index value is larger and the countermeasure efficacy is worse, the normalized training sample is ∈ ->Let the original sample of the ability factor layer be y', when the index value is bigger and the countermeasure effect is better, the normalized training sample is +.>y′ min 、y′ max Respectively, the minimum value and the maximum value in y', when the greater the index value is, the worse the countermeasure efficacy is, the normalized training sample is ∈ ->
Let the original sample of the comprehensive efficiency layer be E' and the normalized training sample be
E′ min 、E′ max Respectively the minimum and maximum values in E';
the optimization training of the cuckoo algorithm comprises the following specific steps:
1) Initializing an objective function
Wherein E is the actual output of the second-stage neural network, E d Y is the expected value of the integrated performance layer k For the actual output of the first level neural network,d yk for the expected value of the capability element layer, the rejection probability P, P E [0,1 ] is initialized];
Initializing the positions of n bird's nest:
ω s (0) =[a (0) 11 ,a (0) 12 ,..,a (0) ML ,b (0) 11 ,b (0) 12 ,..,b (0) LN ,..,c (0) 11 ,c (0) 12 ,..,c (0) NP ,d (0) 1 ,d (0) 2 ,..,d (0) P ] T ,s=1,...,n;
2) Calculating an objective function value of each nest position, and selecting a nest with the optimal current objective function;
3) Keeping the optimal nest position of the previous generation objective function, and updating the nest position by utilizing the Lewy flight;
the updated formula of the position of the bird nest of the cuckoo is omega s (t+1) =ω s (t) +α.L (β), wherein ω s (t) Representing the position of the s-th nest at the t-th iteration; alpha represents the step size; l (β) obeys the lewye distribution:
wherein u, v obeys the normal distribution,
ω i' (t) 、ω j' (t) the position of any two bird nest is the t iteration;
4) Comparing the current position function value with the function value of the optimal nest position of the previous generation, if the current position function value is better, updating the current position function value into the current optimal function value, and if the current position function value is not better, reserving the optimal function value of the previous generation;
5) After the position is updated, a number r epsilon [0,1 ] is randomly generated]If r > P, for ω s (t+1) Continuously updating, comparing the updated nest position function values, and calculating the global optimal position at the moment;
6) Judging whether the maximum iteration number or the minimum error requirement is met, if so, outputting a global optimal position, namely, each connection weight of the countermeasure effectiveness evaluation model, and if not, returning to the step 2) to continue iteration;
7) Bringing the optimal weight into a neural network model to obtain an optimal secondary neural network countermeasure efficacy evaluation model;
when the network countermeasure effectiveness evaluation index system is constructed, based on various common network information equipment combat demands in the network space attack and defense countermeasure field, the network countermeasure effectiveness is decomposed into various capability elements such as network reconnaissance, network attack, network defense and command decision, and each capability element is refined and divided into a plurality of index elements.
2. The method of claim 1, wherein the integrated efficiency E is in the range of 0 to 1.
3. The method of claim 1, wherein α = 1.
4. Use of a method according to any of claims 1 to 3 in the field of network security technology.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110265382.0A CN112926739B (en) | 2021-03-11 | 2021-03-11 | Network countermeasure effectiveness evaluation method based on neural network model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202110265382.0A CN112926739B (en) | 2021-03-11 | 2021-03-11 | Network countermeasure effectiveness evaluation method based on neural network model |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112926739A CN112926739A (en) | 2021-06-08 |
CN112926739B true CN112926739B (en) | 2024-03-19 |
Family
ID=76172648
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202110265382.0A Active CN112926739B (en) | 2021-03-11 | 2021-03-11 | Network countermeasure effectiveness evaluation method based on neural network model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112926739B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113420293A (en) * | 2021-06-22 | 2021-09-21 | 北京计算机技术及应用研究所 | Android malicious application detection method and system based on deep learning |
CN117235477B (en) * | 2023-11-14 | 2024-02-23 | 中国电子科技集团公司第十五研究所 | User group evaluation method and system based on deep neural network |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107153869A (en) * | 2017-03-29 | 2017-09-12 | 南昌大学 | A kind of Diagnosis Method of Transformer Faults based on cuckoo chess game optimization neutral net |
CN107222333A (en) * | 2017-05-11 | 2017-09-29 | 中国民航大学 | A kind of network node safety situation evaluation method based on BP neural network |
CN107919983A (en) * | 2017-11-01 | 2018-04-17 | 中国科学院软件研究所 | A kind of space information network Effectiveness Evaluation System and method based on data mining |
CN108337223A (en) * | 2017-11-30 | 2018-07-27 | 中国电子科技集团公司电子科学研究院 | A kind of appraisal procedure of network attack |
WO2019002603A1 (en) * | 2017-06-30 | 2019-01-03 | Royal Holloway And Bedford New College | Method of monitoring the performance of a machine learning algorithm |
CN109241591A (en) * | 2018-08-21 | 2019-01-18 | 哈尔滨工业大学 | Anti-ship Missile Operational Effectiveness assessment and aid decision-making method |
CN109547431A (en) * | 2018-11-19 | 2019-03-29 | 国网河南省电力公司信息通信公司 | A kind of network security situation evaluating method based on CS and improved BP |
CN110930054A (en) * | 2019-12-03 | 2020-03-27 | 北京理工大学 | Data-driven battle system key parameter rapid optimization method |
CN111163487A (en) * | 2019-12-31 | 2020-05-15 | 上海微波技术研究所(中国电子科技集团公司第五十研究所) | Method and system for evaluating comprehensive transmission performance of communication waveform |
-
2021
- 2021-03-11 CN CN202110265382.0A patent/CN112926739B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107153869A (en) * | 2017-03-29 | 2017-09-12 | 南昌大学 | A kind of Diagnosis Method of Transformer Faults based on cuckoo chess game optimization neutral net |
CN107222333A (en) * | 2017-05-11 | 2017-09-29 | 中国民航大学 | A kind of network node safety situation evaluation method based on BP neural network |
WO2019002603A1 (en) * | 2017-06-30 | 2019-01-03 | Royal Holloway And Bedford New College | Method of monitoring the performance of a machine learning algorithm |
CN107919983A (en) * | 2017-11-01 | 2018-04-17 | 中国科学院软件研究所 | A kind of space information network Effectiveness Evaluation System and method based on data mining |
CN108337223A (en) * | 2017-11-30 | 2018-07-27 | 中国电子科技集团公司电子科学研究院 | A kind of appraisal procedure of network attack |
CN109241591A (en) * | 2018-08-21 | 2019-01-18 | 哈尔滨工业大学 | Anti-ship Missile Operational Effectiveness assessment and aid decision-making method |
CN109547431A (en) * | 2018-11-19 | 2019-03-29 | 国网河南省电力公司信息通信公司 | A kind of network security situation evaluating method based on CS and improved BP |
CN110930054A (en) * | 2019-12-03 | 2020-03-27 | 北京理工大学 | Data-driven battle system key parameter rapid optimization method |
CN111163487A (en) * | 2019-12-31 | 2020-05-15 | 上海微波技术研究所(中国电子科技集团公司第五十研究所) | Method and system for evaluating comprehensive transmission performance of communication waveform |
Non-Patent Citations (6)
Title |
---|
Evaluation of NCM Effectiveness Base on SOM-BP Cloud Neural Networks;Zhang GuangHui 等;《Applied Mechanics and Materials》;第241-244卷(第03期);1779-1784 * |
基于BN-and-BP神经网络融合的陆空联合作战效能评估;周兴旺 等;《火力与指挥控制》;第43卷(第04期);3-8 * |
基于布谷鸟搜索优化BP神经网络的网络安全态势评估方法;谢丽霞 等;《计算机应用》;第37卷(第07期);1926-1930 * |
空间信息网络对抗效能评估指标体系研究;徐志明 等;《计测技术》(第02期);11-13 * |
网络对抗效能评估的指标体系研究;李雄伟 等;《无线电工程》(第03期);14-16+52 * |
网络空间信息防御作战指挥效能评估研究;王劲松 等;《现代防御技术》;第46卷(第05期);143-151+158 * |
Also Published As
Publication number | Publication date |
---|---|
CN112926739A (en) | 2021-06-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110197282B (en) | Threat estimation and situation assessment method based on genetic fuzzy logic tree | |
CN112926739B (en) | Network countermeasure effectiveness evaluation method based on neural network model | |
Asadi et al. | ACORI: A novel ACO algorithm for rule induction | |
CN111275074B (en) | Power CPS information attack identification method based on stacked self-coding network model | |
CN110766138A (en) | Method and system for constructing self-adaptive neural network model based on brain development mechanism | |
Hussain et al. | Analysis of techniques for anfis rule-base minimization and accuracy maximization | |
CN115409099A (en) | Internet of things flow anomaly detection model establishing method and detection method | |
Gylberth et al. | Differentially private optimization algorithms for deep neural networks | |
CN114494771B (en) | Federal learning image classification method capable of defending back door attack | |
Mythili et al. | Deep learning with particle swarm based hyper parameter tuning based crop recommendation for better crop yield for precision agriculture | |
Li et al. | Using sparrow search hunting mechanism to improve water wave algorithm | |
Zhou et al. | Improving robustness of random forest under label noise | |
CN116632834A (en) | Short-term power load prediction method based on SSA-BiGRU-Attention | |
CN115412332B (en) | Internet of things intrusion detection system and method based on hybrid neural network model optimization | |
Qi et al. | Battle damage assessment based on an improved Kullback-Leibler divergence sparse autoencoder | |
CN113887807B (en) | Robot game tactics prediction method based on machine learning and evidence theory | |
Lu et al. | Dynamic evolution analysis of desertification images based on BP neural network | |
Alifiah et al. | Prediction of COVID-19 Using the Artificial Neural Network (ANN) with K-Fold Cross-Validation | |
Sova et al. | Development of a methodology for training artificial neural networks for intelligent decision support systems | |
Liu et al. | Outlier detection algorithm based on SOM neural network for spatial series dataset | |
Zhipeng et al. | Cultural Events Classification using Hyper-parameter Optimization of Deep Learning Technique | |
CN113807005B (en) | Bearing residual life prediction method based on improved FPA-DBN | |
Khamees et al. | Train the Multi-Layer Perceptrons Based on Crow Search Algorithm | |
Xu et al. | Defense against adversarial attacks with an induced class | |
CN113283537B (en) | Method and device for protecting privacy of depth model based on parameter sharing and oriented to membership inference attack |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |