Evaluation of Attack Effect in Ad Hoc Networks Based on Variable

Weight TOPSIS Method

Linxi Guo

, Bin Wu

School of Cyberspace security, Beijing University of Posts and Telecommunications, No.10 Xitucheng Road, Beijing,

China

Keywords: Evaluation of attack effect, Variable weight theory, TOPSIS, Ad Hoc network security.

Abstract: Evaluation of Attack Effect in Ad Hoc networks is one of the key technologies for Ad Hoc network security

applications. In order to solve the traditional attack effect evaluation with the constant weighted summation

can’t adjust the relevant weights in time to the change of the attack situation, which causes the limitation and

one-sidedness of the judgment. This paper proposes an attack effect evaluation model based on variable

weight theory. First, comprehensively considering the relevance of the attack's own complexity and the attack

effect, establish an attack effect evaluation indicator system. Then, construct a state variable weight vector

expression, so that the weights are adjusted accordingly with the change of the situation. Finally, combined

with TOPSIS method, the attack effect evaluation model based on variable weight TOPSIS is constructed.

The experimental simulations show that the evaluation results obtained by the model are scientific and

reasonable, which makes up for the deviation caused by the decision of the constant weight, and provides a

theory and method for the evaluation of the attack effect in Ad Hoc network.

1 INTRODUCTION

Compared with traditional wireless networks, Ad Hoc

networks do not need infrastructure construction, and

have a high coverage and high dynamic self-

organizing network mechanism, which supports

various devices to access and exit the network at any

time, thus more in line with the actual mobile device

networking. In addition, because of their robustness

and low cost, Ad Hoc networks have broad

application prospects in many fields such as

intelligent transportation, disaster relief and military

communications. However, the Ad Hoc network is a

typical dynamic network with a flexible topology,

and there is no unified security control center.

Therefore, Ad Hoc networks are more vulnerable to

various types of attacks such as eavesdropping,

impersonation, tampering, etc., which may lead to

greater security threats (Aarti, D.S. 2013).

A lot of researches have been done on Ad Hoc

network attacks, but the number of studies on the

evaluation of Ad Hoc network attacks is very limited.

The evaluation of the effect of network attacks is an

important part of network security. The evaluation of

the network attack effect is an important part of

network security. The evaluation results can not only

test the effect and assessability of the specified attack

(operation plan) scheme, but also measure the

security of the network through simulated attacks,

thereby improving the security protection capability

of the network.

The research on the evaluation of network attack

effect firstly determines indicator weights according

to performance indicators of the target network. Then,

according to the membership function of the attack

effect, the comprehensive evaluation value is

obtained by the linear weighted comprehensive

method based on constant weights. Zeng, C. X.et

al.(2016) applied fuzzy mathematics theory to

analytic hierarchy process and established an

evaluation model based on FAHP, thus avoiding the

calculation of complex problems; Yuan, Z. and

Jianguo, H. (2014) proposed an attack effect

evaluation method based on network entropy, which

relieved the subjectivity of the evaluation to some

extent; Jajodia, S. et al.(2005) used gray theory into

evaluate calculations so that the evaluation results

implied the correlation between the evaluation

indicators to some extent. In the study of the DoS

attack effect evaluation of mobile Ad Hoc networks,

the idea of variable power was introduced for this

problem, but it was targeted for each specific attack

Guo, L. and Wu, B.

Evaluation of Attack Effect in Ad Hoc Networks Based on Variable Weight TOPSIS Method.

DOI: 10.5220/0008098602130221

In Proceedings of the International Conference on Advances in Computer Technology, Information Science and Communications (CTISC 2019), pages 213-221

ISBN: 978-989-758-357-5

213

(Chen, J. and Ma, T. 2012). The weight determination

method, which undoubtedly increases the complexity

of the evaluation, and the qualitative evaluation

results obtained by the gray fuzzy evaluation model

cannot measure the advantages and disadvantages of

different attacks in the same category. In order to

make the evaluation results more accurate and

reasonable, the attack effect evaluation models are

improved, and new evaluation models are

continuously proposed, but there are still some

problems, and the number of evaluation models that

can be applied to Ad Hoc networks is extremely

limited.

The commonly used constant weights vectors

reflect the overall goodness of the attack effect

evaluation to a certain extent, and the weight

coefficient corresponding to each evaluation

indicator reflects the importance of this indicator.

Therefore, the constant weights vector will play a

good role in most cases. However, regardless of the

value of the evaluation indicator attribute, the weights

vector remains unchanged, so the constant weights

vector cannot objectively reflect the change of the

state value of each attribute and the influence of the

attribute relevance on the weights. There are many

unreasonable phenomena in using the same weights

vector in different attack scenarios, mainly in the

following two types:

1) If the value of the indicator reaches a critical

value, it will have a greater impact on the evaluation

of the attack effect. For example, when the node

corruption reaches a critical value, it will have a great

impact on the reliability indicator of the node. The

network reliability will be poor, and the

corresponding security performance will be worse,

especially when the destroyed node is a critical node.

At the same time, when obtaining the attribute values

of the attack effect evaluation, there may be cases

where the individual indicator values are too low or

zero. Assume that there are two evaluation indicators

in the evaluation of Ad Hoc network attacks effect,

namely network performance and security

performance, and these two indicators are equally

important, that is, the weight 















. Then the comprehensive evaluation result

is 



 



. From the evaluation results,

the results of the attack effect obtained by the state

vector  and the state vector 

 are the same. However, the actual situation

is that the network performance of the target network

with the state vector  is already in a

state of paralysis, and the network availability is

significantly reduced. And the network and security

performance of the target network with state vector

 are still within acceptable limits. The

reason why the evaluation result is inconsistent with

the actual situation is that the constant weight vector

is independent of the value of each indicator, and it

does not affect the influence of the indicator values

on the comprehensive evaluation result.

2) When evaluating specific types of attacks, each

type of attacks focuses on different network security

performance metrics. For example, DoS attacks more

affect the network performance of the target network,

thereby destroying its reliability and availability.

While obtaining information-based attacks more

affect the security performance of the target network,

thereby undermining its confidentiality. Therefore,

different types of attack effects are not comparable.

In addition to the irrational problems caused by

constant weighted summation, the current attack

effect evaluation models are more subjective and

focus on the attack results more than the process,

ignoring the correlation between the complexity of

the attack behavior and the effect of the attacks.

In order to solve the above problems, this paper

innovatively proposes an attack effect evaluation

model based on variable weight TOPSIS. The

innovations of this paper mainly include the

following points:

 This paper comprehensively considers the

impact of attack complexity and proposes an

attack effect evaluation indicator system

suitable for Ad Hoc networks；

 This paper combines the variable weight theory

based on punishment and incentive mechanism

with the TOPSIS evaluation method, and

proposes a state variable weight vector

expression suitable for Ad Hoc network attack

evaluation. The calculation formula

appropriately adjusts the weight according to

the attribute value of the attack effect indicator.

Specifically, a penalty is imposed on the

indicator weight of the attribute value that is

low. While incentives are given to indicator

weights with high attribute values. Therefore,

this model solves the unreasonable problems

brought about by the evaluation of constant

weights.

Finally, we use the specific attack test in the

simulation experiment platform and obtain the real

and objective indicator data to verify the rationality

and effect of the proposed model.

The rest of the paper is organized as follows.

Section 2 proposes a standardized quantization

method for indicators and establishes an evaluation

system for attack effect. Section 3 describes in detail

the method of determining the variable weight vector.

CTISC 2019 - International Conference on Advances in Computer Technology, Information Science and Communications

214

On the basis of using the analytic hierarchy process

to determine the weight of the indicator constant, the

construction and application of the state variable

weight vector applicable to the evaluation are mainly

studied. Then, in Section 4, the variable weight theory

is combined with the TOPSIS method to describe the

specific evaluation process of the variable weight

TOPSIS model. Afterwards ， in Section 5, the

rationality and effect of the proposed model are

proved by experimental simulation. Finally, Section

6 contains our conclusions.

2 ESTABLISH AN EVALUATION

INDICATOR SYSTEM

The establishment of the evaluation model for Ad

Hoc network attack effects can be divided into the

following three steps: establishing an evaluation

indicator system for attack effects, determining the

weight value of the evaluation indicators, and using

the comprehensive evaluation algorithm to calculate

the evaluation results.

The evaluation indicator system is the

infrastructure of the entire assessment process.

Therefore, it is a basis for Mobile Ad Hoc Network

Attack Effect of effective evaluation to establish a

reasonable evaluation indicator system, which is an

important basis to reflect the effect of the attack.

This paper proposes an evaluation indicator

system of attack effect for Ad Hoc network, based on

the correlation between attack complexity and attack

effect, and gives a standardized quantification method

of the indicators.

2.1 Ad Hoc Network Attack Effect

Evaluation Indicator System

The basic idea of establishing the evaluation indicator

system of attack effect for Ad Hoc networks is as

follows: Firstly, according to the security

vulnerabilities of Ad Hoc networks and the impact of

common attacks, the basic evaluation indicators are

selected, and the three-level indicator system of

“target-criteria-indicators” is established (Lai C et al.

2015).

Considering the correlation between the complexity

of the attack and the effect of the attack, this paper

establishes an attack effect evaluation indicator

system for Ad Hoc networks based on attributes of

attack process and performance factors of attack

results, as shown in Figure 1.

Figure 1: Ad Hoc network attack effect evaluation

indicators system.

2.2 Standardized Quantification of

Evaluation Indicators

2.2.1 Standardized Quantification of

Qualitative Indicators

In order to quantitatively describe the effects of

different types of attacks, qualitative indicators such

as cost of attack, attack target level and get

permission level need to be assigned from high to low,

and the data sources can be obtained by experts. The

specific scoring criteria are as follows:

 Cost of attack: It mainly refers to the technical

requirements and the amount of equipment

resources consumed to implement the attack.

For this qualitative indicator, the possible states

of the indicator can be listed and the reference

segment value is assigned according to the

degree of importance, as shown in Table 1.

Table 1: Quantitative reference value of Cost of

attack.

Indicator state

Reference score

Number of malicious nodes

0~3

Resource and equipment

consumption

0~3

Human resources

0~2

Financial consumption

0~4

Other

0~4

Then normalize, that is, the ratio of the initial

attribute value to the reference total score value.

 Attack target level: It mainly refers to the

importance of the network. It is a qualitative

Evaluation of Attack Effect in Ad Hoc Networks Based on Variable Weight TOPSIS Method

215

indicator with order, which can be quantified

according to the information, as shown in Table

Table 2: Quantitative value of Attack target level.

Indicator state

Quantitative value

Single network

0.3

Partial network

0.5

Entire network

0.8

 Get permission level: It refers to the level of

permission obtained through an exploit method

during the attack process and quantifying it

according to the degree of importance, as

shown in Table 3.

Table 3: Quantitative value of Get permission level.

Indicator state

Quantitative value

Single network

0.3

Partial network

0.5

Entire network

0.8

2.2.2 Standardized Quantification of

Quantitative Indicators

Different indicators have different dimensions,

ranges of variation and confrontational problems.

Therefore, they cannot be directly used for attack

effect evaluation. It is necessary to dimensionless and

normalize the original data of the indicator. In this

paper, extreme value processing method will be used

to standardize the results of dimensionless processing

as [0,1]. Considering the problem of different

confrontation among indicators, the indicators are

divided into two types: benefit-oriented indicators

and cost-oriented indicators, which are standardized

and quantified separately:

 Benefit-oriented indicators: The greater the

attribute value, the better the attack effect. For

this type of indicator attribute value 



, the

pre-treatment formula is:











 

















(1)

 Cost-oriented indicators: The smaller the

attribute value, the better the attack effect. For

this type of indicator attribute value 



, the

pre-treatment formula is:











 







   







(2)

Benefit-oriented indicators include attack target

level, getting permission level, getting data volume,

average end-to-end delay, network packet drop rate,

normalized routing overhead, data loss rate, the

amount of data tampering, degree of node destruction

and unit time power consumption. Cost-oriented

indicators include cost of attack, attack time,

bandwidth utilization and throughput scaling.

3 VARIABLE WEIGHT THEORY

TO DETERMINE INDICATORS

WEIGHT

This paper focuses on the application of variable

weight theory in the evaluation of attack effects,

proposes a state variable weight vector expression

suitable for the model and determines the value of the

parameter, which is on the basis of determining the

constant weight of indicators by AHP method. The

state variable weight vector is the key of variable

weight theory in practical application. And It is one

of the most important innovation of this paper.

3.1 Determination of Constant Weight

of Indicators Based on AHP

Analytic Hierarchy Process (AHP) is a method for

determining the weight of indicators combined with

qualitative analysis and quantitative analysis. It can

quantify multiple uncertainties and fuzziness in the

decision process. AHP requires that the problem to be

solved be decomposed into several parts, each Parts

are divided into different hierarchical structures.

Compare each indicator at the same level and

determine the weight of the indicator based on the

importance of the indicator (Sun Z et al. 2012).

3.1.1 Constructing Judgment Matrix

According to the expert opinion, the pairwise

comparison factors are quantified using the 1–9 ratio

scale comparison table with reference to expert

opinions, as shown in Table 4.

Table 4: 1–9 ratio scale comparison table.

Nine-scale

Meaning

ai is as important as a j

ai is a little bit important than a j

ai is obvious important than a j

ai is consuming important than a j

ai is extreme important than a j

Remarks: Take 2, 4, 6, 8 between adjacent judgment

values

According to Table 4 and the comparison of the

advantages and disadvantages of each evaluation

indicator, the following judgment matrix can be

constructed:

CTISC 2019 - International Conference on Advances in Computer Technology, Information Science and Communications

216













 











 



   









 







(3)

Where: 





 



is the ratio

of the i-th factor to the importance of the j-

th factor.



 and 



 





3.1.2 Calculate Indicator Weights and

Consistency Check

For all pairs of comparison matrices, consistency

check is required. The purpose of consistency check

is to avoid the self-contradictory phenomenon of

subjective judgment.

Define 





. If CR<0.1, the judgment matrix

is considered to satisfy the consistency, where 









; 



is the maximum eigenvalue of the

judgment matrix A;  is the order; RI is the average

random consistency indicator. Table 5 gives the

corresponding RI values of matrix 1-14.

Table 5: The value of the random consistency indicator

RI(n).

0.52

0.89

1.12

1.26

1.36

1.41

1.46

1.49

1.52

1.54

1.56

1.58

When the judgment matrix satisfies the complete

consistency, the eigenvector corresponding to the

eigenvalue 



is the constant coefficient of the

indicator 







 







Since the judgment matrix is constructed by the

subjective judgment of the expert, if it does not meet

the consistency, the data need to be adjusted.

3.2 Determination Indicators Weight

by Variable Weight Theory

The introduction of variable weight theory can solve

the problem that the weights of indicators in constant

weight assessment cannot be changed according to

the change of the attack situation, leading to the

decision bias. Therefore, how to apply variable

weight theory to the field of attack effect evaluation,

which can make the change of weight better reflect

the attack situation, is the key of this paper.

3.2.1 Variable Weight Theory

Variable weights are relative to constant weights. The

concept of variable weight vector and state variable

weight vector were first proposed by Wang, P.X.

(1985). It was emphasized that the weight of indicator

should change with the change of attribute value of

indicator in order to overcome the deviation caused

by constant weight decision-making. Li, H. X. (1995)

further gave the axiomatic definition of variable

weight and state variable weight vector.

Let 







 



 be the state variable and





















  be the variable

weight of the relative constant weight 



. According

to the variable weight theory, the variable weight

vector satisfies the following axiom (Deqing, L.

2002):

1) Normalized condition:

 

























；

2) Continuity: The variable weight vector





















is continuous with respect to each

independent variable 



;

3) Monotonicity: The variable weight vector





















is monotonically decreasing

(punitive variable weight) or increasing (incentive

variable weight) with respect to the independent

variable 



.

The variable weight vector x mainly relies on the

construction of state variable weight vector s.

According to the configuration level of attributes of

attack effect evaluation indicator, the weight values

of each indicator are adjusted. In addition to

satisfying continuity and monotonicity, the state-

varying weight vector also satisfies the Hadamard

product:











 







 



















(4)

According to the above definition, it can be seen

that the purpose of the punitive state variable weight

vector is to punish the low-level indicator attribute to

ensure the balance of the evaluation indicator by

increasing the weight of the indicator with the

decrease of the state value; The purpose of the

incentive state variable weight vector is to stimulate

the high-level indicator attribute by increasing the

indicator weight with the increase of the state value.

3.2.2 Constructing State Variable Weight

Vector

The key to the application of variable weight theory

to the actual variable weight problem lies in the

construction and selection of the state variable weight

vector 







. Therefore, the characteristics of the

existing various types of state variable weight vectors

and the requirements of equilibrium for the decision-

making problem should be fully considered in

practical applications. Next, after analysing the

Evaluation of Attack Effect in Ad Hoc Networks Based on Variable Weight TOPSIS Method

217

evaluation process of attack effect for Ad Hoc

networks, it is proposed that in order to meet the

variable weight requirements proposed in this paper,

the constructed state variable vector should satisfy the

following characteristics:

1) The indicator value is too large or too small, and

the weight is increased. If a certain indicator is too

high in the evaluation process, which means that a

certain performance indicator of the target network is

too low or the attack cost is very small, it will affect

the rise of the overall attack effect level regardless of

the size of the constant weight. Therefore, the weight

of the indicator needs to be increased. Similarly, if a

certain attribute value is very low, which means that

a certain performance of the target network is not

affected by the attack or the attack cost is too high, it

will also affect the overall evaluation level of attack

effect to a certain extent even if the constant weight

of this indicator is very small. So, the weight of the

indicator also needs to be increased.

2) Incentive range is greater than punishment range.

Due to the complexity of Ad Hoc networks, the

relationship between the proposed attack effect

evaluation indicators is relatively large and inevitably

there are redundant indicators. By analyzing the value

of single indicator separately, it is found that when the

value of single indicator is high, such as the average

end-to-end delay is too high, the overall attack effect

is significantly improved. However, when the value

of single indicator is low, the overall attack effect is

not significantly reduced. Therefore, considering the

balance of the evaluation indicator system, the state

variable weight vector in the evaluation model of the

attack effect of Ad Hoc network should satisfy the

requirement that the incentive range is greater than

the penalty range.

3) The punishment and incentive of the indicator

with relatively large constant weight are also

relatively large. The constant weight reflects the

relative importance of each indicator attribute to a

certain extent. The evaluation result of the attack

effect is more dependent on the indicator with

relatively large weight. Therefore, the state variable

weight vector should be able to punish and motivate

the indicators with relatively large constant weight.

In view of the advantages of exponential state

variable weight vectors, such as obvious decision-

making requirements, flexible parameter setting and

strong model expansion ability, this paper constructs

the expression of state variable weight vectors





























 









as Equation 5 by

drawing on the relevant research results of variable

weight theory and satisfying the above three points of

analysis.























































(5)

Where: n is the number of indicators;;















is the average attack effect indicator value;

 are the penalty amplitude coefficient and the

excitation amplitude coefficient respectively and 

;  is the penalty threshold coefficient;

when the value of the j-th indicator status value is not

higher than the penalty threshold or not lower than the

incentive threshold ， the weight is increased by

changing the weight to achieve the purpose of

punishment or incentive. In practical applications, the

evaluator should set and k according to the

specific requirements of the attack effect evaluation.

4 CONSTRUCTION OF

VARIABLE WEIGHT TOPSIS

EVALUATION MODEL

The basic principle of Technique for Order

Preference by Similarity to Ideal Solution (TOPSIS)

is to rank the evaluation objects by means of the

positive ideal solution and the negative ideal solution

in the multi-objective decision problem.

Theoretically, the positive ideal solution and the

negative ideal solution are the optimal solution and

the worst solution respectively, which are often not

achieved in reality. When evaluating the attack effect

of an attack scheme, the scheme should be judged

from the distance between the two ideal solutions to

determine the situation between the different schemes.

On this basis, the paper proposes a variable

weight TOPSIS model.

4.1 Construction of Normalized Multi-

Attribute Evaluation Matrix

There are m attack schemes to form a scheme set 









 



, and n evaluation indicators constitute

the indicator set 











, then the

evaluation sample value 



of 



to 



constitutes

the multi-attribute evaluation matrix X.













 











 



   









 







(6)

CTISC 2019 - International Conference on Advances in Computer Technology, Information Science and Communications

218

4.2 Normalization of Evaluation

Matrix

The indicators are divided into qualitative indicators

and quantitative indicators. Qualitative indicators can

be normalized according to the quantitative principles

in Table 1, Table 2 and Table 3. The quantitative

indicators can be further divided into benefit

indicators and cost indicators, which are normalized

according to formula (1) and formula (2) respectively

to obtain the normalized matrix U as follows.













 











 



   









 







(7)

4.3 Determining Indicator Weights

Based on Variable Weight Theory

Firstly, the AHP method is used to calculate the

indicator constant weights, and the indicator constant

weight coefficient vector 











 





obtained. Then according to the state variable weight

vector, that is, formula (5), the indicator weights















































 are

calculated.

4.4 Establishing the Weighted

Normalization Evaluation Matrix

Based on Variable Weight Vectors

The weighted normalization evaluation matrix Y is

obtained by multiplying the corresponding items of

the matrix U and the matrix W(U), and is expressed

as follows:





















 































 















  

















 



















(8)

4.5 Calculating the Closeness of Attack

Schemes

1) Positive and negative ideal solutions are as follows:











 















































(9)

2) The distance values between each attack

scheme and the ideal solutions are as follows:

























 



























 















(10)

3) The closeness of each attack plan and positive

ideal solution is calculated according to the following

formula:





















 





 







(11)

Where： 





represents the closeness degree of

each evaluation scheme and the positive ideal

solution, and also indicates the degree of distance

from the negative ideal solution. Therefore, each

scheme can be evaluated and ranked by sorting 





descending order.

5 CASE ANALYSIS

5.1 Simulation Environment Setting

The simulation environment is established under NS2.

The settings of network parameters and environment

parameters in the scenario are shown in Table 6.

Table 6: Simulation parameters setting.

Network

parameter

Set value

Scene

parameter

Set value

Simulation

area size

1000m*10

00m

Channel

attenuatio

n model

TwoRayG

round

Number of

network nodes

Antenna

type

Omni

Antenna

Channel type

Channel/

Wireless

PHY

protocol

Phy/Wirel

essPhy

Channel

bandwidth

2Mbps

MAC

Protocol

MAC/802

_11

Maximum

movement

rate

30m/s

Routing

Protocol

AODV

Transmission

distance

250m

Interface

Queue/dro

ptail

Data packet

size

512 Bytes

Wireless

network

interface

At the beginning of the simulation, the initial

energy of all nodes is consistent, the packet loss rate

is maintained between 0% and 15%, and the attack

duration is 300 seconds.

Evaluation of Attack Effect in Ad Hoc Networks Based on Variable Weight TOPSIS Method

219

5.2 Quantitative Measurement of

Indicators

1) In the simulation experiment analysis of the Hello

flood attack scenario, the attack nodes are randomly

selected. as shown in Figure 2. The attack payload is

20 kb/s, and the number of attack nodes is 4, 8, and

12. The normalized values of the indicators are as

shown in Case 1, Case 2 and Case 3 in Table 7.

Figure 2: Simulation of the Hello Flood attack scenario.

2) In the simulation experiment analysis of the

wormhole attack scenario, the attack nodes are

randomly selected. as shown in Figure 3. The number

of malicious nodes is 4 and 8. The normalized values

of the indicators are as shown in Case 4 and Case 5 in

Table 7.

Figure 3: Simulation of the wormhole attack scenario.

Table 7: Indicator normalization values.

Indicator

Case1

Case2

Case3

Case4

Case5

Number

malicious

nodes

0.68

0.43

0.31

0.58

0.39

0.30

0.40

0.42

0.67

0.33

0.43

0.62

0.38

0.44

0.30

0.42

0.51

0.33

0.42

0.17

0.26

0.33

0.46

0.28

0.43

0.62

0.52

0.68

C10

0.15

0.20

0.42

0.16

0.22

C11

0.32

0.46

0.45

0.46

0.58

C12

C13

0.083

0.10

0.13

0.067

C14

0.11

0.17

0.24

0.087

0.10

5.3 Obtaining Indicator Weights Based

on Variable Weight Theory

1) The basis weight of each indicator is determined

based on the AHP method. The judgment matrix of

the criterion level indicator 





















based on the expert opinion is established as follows:

























1/5

1/6

1/5

1/4





1/2









1/2





1/2

1/3

1/2

CR=0.0145<0.1, which satisfies the consistency

requirement, and the weights of the criterion layer can

be calculated as W= (0.0901,0.224,0.301,0.224,

0.160). Similarly, the weights of each indicator layer

can be calculated as shown in Table 8.

Table 8: Indicator weight values of each layer.

Aggressive

indicator

(0.485,0.340,0.175)

Confidentiality

indicator

(0.667,0.333)

Usability

indicator

(0.222,0.097,0.169,0.384,0.128)

Integrity

indicator

(0.667,0.333)

Reliability

indicator

(0.5,0.5)

Finally, the resulting constant weight vector is W =

(0.0441,0.0309,0.0159,0.149,0.0746,0.0668,

0.0292,0.0509,0.116,0.039,0.149,0.0746,0.08,0.08).

2) Determine the variable weight vector matrix.

According to the variable weight state vector, set

=0.5,, k=0.7. Combining the formulas (5),

(4) and the constant weight vector W, we can obtain

the variable weight vector matrix as:

W(U)=







    

    

    

    

    

    

    

    

    

    

    

    

    

    









CTISC 2019 - International Conference on Advances in Computer Technology, Information Science and Communications

220

5.4 Comprehensive Evaluation Results

and Analysis Based on Variable

Weight TOPSIS Method

Based on the weighted vector matrix W(U), the

weighted normalized evaluation matrix is further

calculated to obtain the closeness of the attack

scheme, as shown in Table 9. The rationality and

effect of the variable weight TOPSIS evaluation

model are verified by comparison with the calculation

results of the constant-weight TOPSIS evaluation

model.

Table 9: Closeness of each attack scheme.

Case1

Case2

Case3

Case4

Case5

variable-

weight

TOPSIS

0.1892

0.3341

0.4938

0.5079

0.7562

constant

weight

TOPSIS

0.1673

0.3304

0.4919

0.4831

0.7835

Using variable weight TOPSIS evaluation

method, the evaluation results rank of the attack effect

is Case 1 < Case 2 < Case 3 < Case 4 < Case 5. If

constant-weight TOPSIS method is used, the attack

effect evaluation rank is Case 1 < Case 2 < Case 4 <

Case 3 < Case 5. Comparative analyses of the results

of different methods are as follows:

1) Case 1 < Case 2 < Case3 and Case 4 < Case

5 are satisfied simultaneously. It means that for the

same attack scheme, the more malicious nodes, the

better the attack effect. According to this, the

rationality of the variable weight TOPSIS model has

been confirmed.

2) Case 4 is a wormhole attack initiated by four

malicious nodes. By establishing a fake malicious

channel to steal data packets, the network

performance and security performance of the target

network are simultaneously reduced. In comparison,

although Case 3 has more malicious nodes, the flood

attack only affects the network performance of the

target system. Therefore, the attack effect of Case 4 is

stronger than Case 3. The evaluation results of

variable weight TOPSIS model have shown that the

proposed model solves the limitation of the problems

in constant-weight TOPSIS model to some extent,

and is a more effective evaluation method.

6 CONCLUSIONS

Aiming at the problem of attack effect evaluation of

Ad Hoc networks, based on the comprehensive

consideration of the correlation between attack

complexity and attack effect, this paper constructs a

comprehensive evaluation indicator system of attack

effect. And by introducing the variable weight theory,

an attack effect evaluation model based on variable

weight TOPSIS is proposed. The model can

reasonably adjust the weights based on the change of

the attribute values of each indicator, and can obtain

a more reasonable evaluation result. The proposed

evaluation method overcomes the limitations of

traditional attack effect evaluation methods and

provides an effective reference processing method for

Ad Hoc network attack effect evaluation.

REFERENCES

Aarti, D.S. 2013. Tyagi, “Study Of Manet: Characteristics,

challenges, application and security attacks”.

International Journal of Advanced Research in

Computer Science and Software Engineering, 3(5),

pp.252-257.

Zeng, C. X., Quan-Yuan, W. U., Ai-Ping, L. I. and Jiang, R.

2016. Research on fahp based trojan attack effect

evaluation. Chinese Journal of Network & Information

Security.

Yuan, Z. and Jianguo, H. 2014. Estimation method for

information support efficiency of network attacks based

on system entropy. Journal of Pla University of Science

& Technology.

Jajodia, S., Noel, S. and O’Berry, B. 2005. Topological

analysis of network attack vulnerability. In Managing

Cyber Threats. Springer, Boston, MA, pp. 247-266.

Chen, J. and Ma, T. 2012. DoS attack effect assessment

method in mobile Ad Hoc network. Dianguang yu

Kongzhi(Electronics, Optics & Control), 19(3), pp.86-

89.

Lai, C., Chen, X., Chen, X., Wang, Z., Wu, X. and Zhao, S.

2015. A fuzzy comprehensive evaluation model for

flood risk based on the combination weight of game

theory. Natural Hazards, 77(2), pp.1243-1259.

Sun, Z. and Liu, M. 2012. Application of Fuzzy AHP

Method in the Effect Evaluation of Network Attack. In

2nd International Conference on Electronic &

Mechanical Engineering and Information Technology.

Atlantis Press.

Wang, P.Z. 1985. Fuzzy sets and random colony shadow.

Beijing: Beijing Normal University Press, pp. 93−102.

Li, H.X. 1995. Factor space theory and knowledge

representation of Mathematical framework(VIII):

variable weight synthesis. Fuzzy Systems and

Mathematics, 9(3), pp.1−9.

Deqing, L. 2002. The properties and construction of state

variable weight vectors. Journal of Beijing Normal

University, 38(4), pp.455-461.

Evaluation of Attack Effect in Ad Hoc Networks Based on Variable Weight TOPSIS Method

221