Research on Security Evaluation Methods of Edge Computing

Information Systems

Keqi Chen

College of Economic and Management, Tiangong University, Tianjin, China

Keywords: Edge Information System, Security Evaluation Grading, Grey Clustering, Subjective and Objective Weighting.

Abstract: With the rapid development of economy and science and technology, the demand for edge computing

technology has expanded. Edge computing can provide caching and processing functions to realize the

localization of computing power and resource storage. Under the high requirements of edge information

system, the accuracy of its security evaluation and classification is also required to be improved. At the same

time, due to the limited resources, the data scale should not be too large and not strictly obey the normal

distribution, so the adaptability and accuracy of the traditional first-order grey clustering are insufficient, and

the similarity of the maximum membership degree adjacent membership degree may lead to fuzzy judgment

and reverse sorting results. Single empowerment method using subjective right confirmation method or

objective right confirmation method alone will lead to inaccurate results. This review will discuss the security

evaluation and grading model for edge information systems from the following aspects: overview of

mainstream methods, overview of data sets and evaluation criteria, the current level of this direction, and the

discussion of current problems and solutions. In this paper, some processing methods of the information

system security evaluation and grading model of edge computing are discussed. In the subsequent work,

further research can be carried out from the following two aspects: the more optimized solution of grey

clustering, and whether the security evaluation and grading model can be applied to various scenarios.

1 INTRODUCTION

The development of edge information system is

becoming more and more mature, and the security

issues in it have also been widely concerned, so the

security evaluation and classification is also very

important. In the determination of index weight,

various weighting methods have their own

advantages and disadvantages, and how to decide the

weighting method is the most important. At present,

there has been preliminary improvement in security

evaluation in various fields. For big data network

security problems, the network security evaluation

method based on rough set can simplify the core

values that can represent network security attributes

from a large number of high-dimensional dynamic

data, so as to carry out the security evaluation of

network security (Liu, 2019). In the study of the

security of medical big data information sharing, the

researchers selected the security sharing indicators of

medical big data according to the Delphi method, and

https://orcid.org/0009-0007-0867-5677

established the hierarchical evaluation model of the

indicators by using the analytic hierarchy process.

The evaluation method can effectively evaluate the

security of medical big data sharing, and provide a

new technical platform and solution for hospital

information management (Shan, 2018). In the aspect

of network security assessment, a network security

assessment model based on multi-data layer is

proposed, which realizes qualitative and quantitative

assessment functions and provides a new idea for

network risk assessment (Ma, 2011).

In the edge information system, the data size

needs to be controlled in a reasonable range, and the

data distribution often shows the characteristics of

non-strictly normal. The traditional first-order grey

clustering evaluation system has exposed obvious

limitations in this situation, and its adaptability and

accuracy are difficult to meet the actual needs.

Especially in the judgment of evaluation results,

because the values of the maximum membership

degree and the adjacent membership degrees are

Chen, K.

Research on Security Evaluation Methods of Edge Computing Information Systems.

DOI: 10.5220/0013703900004670

Paper published under CC license (CC BY-NC-ND 4.0)

In Proceedings of the 2nd International Conference on Data Science and Engineering (ICDSE 2025), pages 667-672

ISBN: 978-989-758-765-8

667

close to each other, the judgment process is often

ambiguous, and even the ranking results are

unreasonably reversed, which seriously affects the

reliability and effectiveness of the evaluation.

In order to overcome these problems, a high-

accuracy security evaluation grading model based on

subjective and objective weighting and second-order

grey clustering, namely C-SG model, came into

being (Guo, 2024). The model is divided into seven

modules with clear and interrelated functions. The

security analysis module is taken as the starting step,

that is, the selection of evaluation objects and the

determination of security objectives. Then, the

security policy decomposition module undertakes

the important task of in-depth analysis of the security

requirements, potential threats and vulnerabilities of

the edge information system. Secondly, the mapping

and delineation of security metrics involves aligning

chosen indicators with the edge information system

security evaluation framework. Next, the data

acquisition and preprocessing are used to eliminate

the influence of the order of magnitude and

dimension. Then the determination module of index

weight, and finally the security evaluation grading

module and effectiveness analysis module. Among

them, the determination module of index weight

adopts the method of combining subjective and

objective weighting, scientifically and reasonably

determines the relative importance weight of each

safety index in the evaluation system, and avoids the

one-sidedness and limitations of single weighting

method. According to the second-order grey

clustering algorithm (Guo, 2024), the security

evaluation grading module performs in-depth

analysis and calculation on the processed data, and

obtains accurate security evaluation grading results,

which effectively solves the shortcomings of the

traditional first-order grey clustering method in

processing the data of the edge information system.

This review will introduce some existing evaluation

methods in the safety evaluation and grading,

including the determination of index weight, as well

as their scientific nature and limitations, and put

forward the corresponding solutions.

2 OVERVIEW OF MAINSTREAM

METHODS

2.1 Subjective and Objective

Empowerment

1)Analytic hierarchy process combined with entropy

weight method (Wen, 2022): Analytic hierarchy

process (AHP) builds a hierarchical structure, and

experts score the importance of each index to

determine the weight, which is subjective, but can

comprehensively consider the logical relationship

between indicators. The entropy weight method

determines the weight according to the information

entropy of the index data, and the objectivity is

outstanding. The combination of the two can

complement each other, make the weight

determination more scientific and reasonable, and

make the subsequent security evaluation grading

more accurate.

2)The coefficient of variation method is paired

with the expert scoring method: The variation

coefficient method (Yang, Wang, Zheng, et al, 2024)

using the variation degree of each index data to

measure its impact on the overall weight, reflect the

characteristics of the data is discrete; The expert

scoring method integrates professional experience to

give weight judgment, and the combination of the two

can take into account the difference between

objective data and subjective professional cognition,

which is helpful to improve the quality of weight

setting in the evaluation model.

2.2 Second-order Grey Clustering

Aspect

1)Optimization of grey clustering coefficient

calculation (Guo, 2024): The traditional way of

calculating grey clustering coefficient is improved.

For example, the determination of different gray class

boundary values of each index is considered more

finely, and a more reasonable whitening weight

function is used, so that the clustering process can

more accurately divide the evaluation object into the

corresponding security level category, and improve

the accuracy of classification.

2)Combined with big data sample correction:

With the increasing amount of available data, the

parameters of the second-order grey clustering are

constantly corrected and improved by using a large

number of actual security evaluation related sample

data, so that the model can better adapt to the security

evaluation needs in different scenarios, reduce errors,

and achieve more accurate security evaluation

grading.

3)Integration with other intelligent algorithms:

For example, combining with intelligent algorithms

such as smart grid fault prediction and diagnosis (Xu,

2024) and mathematical optimization algorithm (Yan,

2024), the scope of safety level is initially divided by

second-order grey clustering, and then the learning

and optimization capabilities of intelligent algorithms

ICDSE 2025 - The International Conference on Data Science and Engineering

668

are used to further adjust accurately, give full play to

their respective advantages, and jointly improve the

accuracy and reliability of safety evaluation grading.

3 COMMONLY USED DATASETS

AND EVALUATION CRITERIA

The dataset in this direction is constructed based on

the original security evaluation data. In the process of

data collection, the secondary index is clearly set as

the minimum unit of index collection, and the

standardized processing method is used to convert the

original data. Thus, the computational obstacles

caused by different orders of magnitude and units are

eliminated. In terms of determining index weights,

the C-SG model (Guo, 2024) adopts a combination

weighting method based on the optimal function

theory and the combination of subjective and

objective weights. It is divided into PageRank

weighting method based on indicator correlation

(Song, Gong, 2023) and entropy weighting method

based on indicator information (Yang, Zhang, 2024).

Finally, the subjective and objective combination

weighting method based on optimal function theory

organically integrated the above two weighting

results. The optimal function is constructed to balance

the advantages and disadvantages of subjective and

objective weighting.

In this direction, the determination of index

weight is divided into subjective weight confirmation

methods, such as Delphi method (Xu,2023) and

analytic hierarchy process (Kang, Xu, Yang, et, al,

2024). Objective weight confirmation rules include

entropy weight method (Yang, Zhang, 2024) and

CRITIC method (Chen, Lv, Yang, 2022). At present,

a combination weight method combining subjective

and objective weights based on optimal function

theory is also proposed (Qin, Yuan, Zhou, et, al,2023).

It includes PageRank weighting based on indicator

relevance (Song, Gong, 2023) and entropy weighting

method based on indicator information (Yang, Zhang,

2024), and uses the least square method to realize the

combination of the two weights. In the security

evaluation grading, the whitening weight function is

constructed, the standardized gray clustering

coefficient is calculated, and then the second-order

correction factor is introduced to calculate the

comprehensive decision vector to further increase the

discrimination of the original unit decision vector

(Guo,2024).

4 EXISTING PROBLEMS AND

POSSIBLE SOLUTIONS ARE

DISCUSSED AND ANALYZED

4.1 Limitations of C-SG Model

Evaluation Methods

There are problems: In the module 7 validity analysis

module of C-SG model, the model is only evaluated

by the membership difference coefficient and the

head to tail consistency rate (Guo,2024), which has

certain limitations. Although these two indicators can

reflect the consistency and membership

discrimination of model evaluation results to a certain

extent, for a complex security evaluation grading

model, they cannot fully cover the overall

performance of the model under different application

scenarios, different data characteristics and various

actual requirements. The availability and

effectiveness of a model are multi-dimensional

concepts, involving the adaptability of the model to

various types of data, the ability to accurately

distinguish different security situations, and the

operating efficiency in resource-constrained

environments. It is difficult to build a complete and

reliable effectiveness evaluation system only by

relying on these two indicators, which may lead to

misjudgment or inadequate evaluation of the overall

performance of the model. Thus, it affects the

promotion and application of the model in the actual

safety evaluation work. Solution: To ensure that the

usability and effectiveness of the model are

comprehensively and accurately evaluated, it is

necessary to introduce more diverse evaluation

methods. First, evaluation metrics based on confusion

matrix (Cao, Yin, Li, et al, 2024) can be used, such as

precision, recall, F1 score, etc. By comparing the

prediction results of the model with the real security

situation, constructing a confusion matrix, and then

calculating these indicators, researchers can

intuitively understand the ability of the model in

correctly identifying the security level. For example,

precision reflects the proportion of correct

predictions, recall reflects the ability of the model to

capture positive samples (for a given security level),

and the F1-score takes both precision and recall into

account to provide a more balanced evaluation.

Second, cross-dataset validation is performed.

Several datasets from different edge information

systems with different data distribution

characteristics were used to verify the model. This

allows you to examine how well your model

generalizes in different scenarios, avoiding situations

Research on Security Evaluation Methods of Edge Computing Information Systems

669

where your model works well on a particular dataset

and fails in other real-world applications. By

evaluating various performance indicators of the

model on multiple data sets, and analyzing its stability

and adaptability, the confidence in the availability

and effectiveness of the model is further enhanced.

4.2 The One-sidedness of the

Combinatorial Weighting Method

There are problems: In the current combination

weighting method combining subjective and

objective weights, it only focuses on the PageRank

weighting method based on indicator correlation and

the entropy weighting method based on indicator

information (Guo,2024). This limitation makes the

research on combination weighting method not

comprehensive. Although the advantages of these two

methods are described, other possible weighting

methods and their shortcomings are not deeply

discussed, which makes it impossible to examine the

rationality of method selection from a broader

perspective. Different weighting methods have their

own characteristics and application scenarios.

Ignoring other methods may mean missing out on a

more optimal weight determination strategy, or

failing to fully recognize the possible limitations of

existing methods under specific conditions. This may

cause the adaptability and accuracy of the combined

weighting method to be questioned in the face of

complex and changing security evaluation data and

requirements, and then affect the reliability and

effectiveness of the whole security evaluation grading

model. Solution: In order to determine the rationality

of the existing method selection, it is necessary to

deeply explore other subjective and objective

weighting methods and analyze their shortcomings.

For example, for the direct scoring method in the

subjective weighting method, its advantage is that it

is simple and direct, and it can quickly obtain the

intuitive judgment of experts on the importance of

indicators. However, the disadvantages of the method

are that it completely relies on the subjective opinions

of experts, lacks consideration of the inherent laws of

data, and is easily affected by experts' personal

preferences, knowledge limitations, and subjective

arbitrariness, which leads to imprecise weight

allocation, especially in the face of large-scale and

complex data and index system, there may be weight

imbalance. This method determines the weight

according to the degree of dispersion of the index

data. The greater the degree of dispersion, the greater

the weight is given to the index, which highlights the

influence of data difference on the weight. But its

disadvantage is that it places too much emphasis on

the variability of the data and may ignore the actual

importance implications of the indicators themselves.

In some cases, even if the degree of dispersion of an

index is small, it is of key significance from the

perspective of professional domain knowledge, and

the deviation maximization method may

underestimate its weight, thus affecting the rationality

of the evaluation results. Through a comprehensive

analysis of these and other related weighting

methods, compared with PageRank weighting and

entropy weighting methods, the advantages and

disadvantages of different methods can be more

clearly understood. Thus, according to the specific

goals of security evaluation, data characteristics,

application scenarios and other factors, The

rationality of the combination weighting method

combining subjective and objective weights selected

at present is determined rationally, which lays a solid

methodological foundation for constructing a more

scientific and accurate safety evaluation grading

model.

4.3 Limitations of the Scoring Method

There are problems: In the entropy-TOPSIS method

comprehensive evaluation, the initial scoring of the

index system is crucial. The expert scoring method

initially adopted is based on the experts' own

experience and professional knowledge to score

various indicators (Chou, Ding, Gao, et,al, 2024).

However, this method has obvious limitations in the

scoring process because it lacks verification of its

own rationality. In practice, different scoring methods

have their own advantages and disadvantages. For

example, although the subjective scoring method can

reflect the subjective judgment of experts, it is easily

affected by personal bias and subjective factors.

Objective scoring methods, such as those based on

statistics, can reduce the interference of human

factors, but may ignore the internal relationship

between indicators. Solution: In order to ensure the

scientific and accurate scoring, we need to

comprehensively evaluate multiple scoring methods.

By analyzing the advantages and disadvantages of

various scoring methods, the most suitable scoring

method is selected, so as to provide a more reliable

basis for the comprehensive evaluation of entropy

weight-TOPSIS method. Only after such a

comprehensive consideration and analysis, can the

initial score be reasonably assigned, so that the

entropy weight-TOPSIS method is more scientific

and effective in the comprehensive evaluation. For

example, consider the analytic Hierarchy process,

ICDSE 2025 - The International Conference on Data Science and Engineering

670

which is a systematic analysis method that

decomposes a complex problem into multiple levels

and determines the weight of each factor through the

steps of establishing a hierarchical structure model,

constructing a judgment matrix, calculating the

weight vector, and performing a consistency check.

Firstly, the hierarchical structure model should be

established, and the problem is divided into goal

level, criterion level and scheme level. Then the

judgment matrix is constructed. For each element of

the same level, the importance of each element

relative to the element of the previous level is

compared by experts or decision makers, and the

value is assigned by 1-9 scale method to construct the

judgment matrix. Then, the weight vector was

calculated, the largest eigenvalue of the judgment

matrix and its corresponding eigenvector were

calculated, and the weight vector of each element was

obtained after normalizing the eigenvector. Finally,

the consistency check is carried out to ensure the

rationality of the judgment. Analytic Hierarchy

Process (AHP) can make complex problems

organized and hierarchical, so as to make decision-

making more scientific and reasonable, but it also has

some limitations such as strong subjectivity in

constructing judgment matrix and complex

calculation for problems with many indicators.

Therefore, the combination of this method and

entropy weight method can realize the

complementary advantages of subjective and

objective.

4.4 The Disadvantages of Grey

Relational Analysis

There are problems: The grey relational analysis

method (Xu, Huang, 2024) is adopted in the research

of disaster risk assessment based on AHP-entropy

combination weighting. This method is not

demanding on the amount of data in theory, but in

practical application, if the amount of data is too

small, it will lead to unstable calculation results of

correlation degree, lack of sufficient representation

and reliability, thus affecting the analysis conclusion.

At the same time, the default data of this method is

equal spacing and regularity, but the actual data may

not meet it, such as non-equal interval time series data

or data with abnormal fluctuations, which will make

the analysis results deviate. Solution: Develop a

comprehensive data collection plan to collect as much

relevant data as possible before the study begins. If

the original data is insufficient, supplementary data

can be obtained through various channels, such as

collecting data from other relevant databases,

industry reports, questionnaires, etc., to increase the

richness and representation of data. If the original

data is insufficient, supplementary data can be

obtained through various channels, such as collecting

data from other relevant databases, industry reports,

questionnaires, etc., to increase the richness and

representation of data. For non-equally spaced data,

interpolation or smoothing can be used to transform it

into approximately equally spaced data. For data with

abnormal fluctuations, robust statistical methods are

used, such as eliminating outliers, performing data

smoothing or using robust statistics such as median

instead of mean for analysis. According to the actual

distribution characteristics of the data, we can also

choose the appropriate grey relational analysis to

improve the model. For the data with obvious

seasonality or periodicity, the method of seasonal

adjustment or cycle decomposition can be introduced

to preprocess the data, and then the grey relational

analysis can be carried out.

5 CONCLUSIONS

In this paper, a high accurate safety assessment

grading model based on subjective and objective

weighting and second-order grey clustering, risk

assessment based on entropy weight-TOPSIS method

and disaster risk assessment based on AHP-entropy

combination weighting are summarized. The validity

and shortcomings of these models are discussed and

analyzed, and the solutions are proposed. At the same

time, combined with the current status of the edge

information system, the accuracy of the security

evaluation grading model was improved. In the future

work, we will devote to study the more optimal

solution of grey clustering and whether the security

evaluation grading model can be applied to various

scenarios.

REFERENCES

Cao, M., Yin, S. X., Li, S. Q., Wang, X. & Miao, Y.L.,

(2024). Study on risk assessment of water inrush from

mining floor with pressure based on confusion matrix.

Coal science and technology 1-13.

Chen, L., Lu, L. & Yang, X.D., (2022). Grey correlation

network security situation assessment method based on

improved CRITIC. Telecommunication technology

(04),517-525.

Chou, W. H., Ding, X. X., Gao, J. S., Zhu, J. & Song, W.

H., (2025). Risk assessment of water supply facilities in

public buildings based on entropy weight -TOPSIS

Research on Security Evaluation Methods of Edge Computing Information Systems

671

method. Water conservancy planning and design

(01),64-69+81.

Guo, Z. Y., (2024). Research on Information System

Security Evaluation Model for Edge Computing (Ph. D.

Dissertation, Beijing University of Posts and

Telecommunications).PhDhttps://link.cnki.net/doi/10.2

6969/d.cnki.gbydu.2024.000232doi:10.26969/d.cnki.g

bydu.2024.000232

Kang, D. C., Xu, L., Yang, L. X., Shan, B., Duan, J., Liang,

X. F. & Zhang, Q. P., (2024). Evaluation of geological

hazard susceptibility in Shapotou region based on

analytic hierarchy Process and GIS. Value engineering

(36),29-32.

Liu, B., (2019). Research on Network Security Evaluation

of Big Data Center Based on Rough

Set.(eds.)Proceedings of the 16th International

Conference on Innovation and Management(pp.368-

372).School of Electrical and Electronic Engineering,

Wuhan Polytechnic University;doi:10.26914/c.

cnkihy.2019.095123.

Ma, X. H., (2011). A New Network Security Evaluation

Model based on Multi-data and

Layer.(eds.)Proceedings of 2011 International

Conference on Advanced Materials and Computer

Science(ICAMCS 2011 Part2)(pp.648-654).Ningxia

Polytechnic University;

Qin, F. T., Yuan, X. J., Zhou, C. & Fan, Y. W., (2023). Grey

evaluation method of network security level based on

comprehensive empowerment. Computer

science(S2),873-878.

Shan, M. J., (2018). Analysis and Evaluation on the Sharing

Security of Medical Big Data based on Analytic

Hierarchy Process. (eds.) Proceedings of 2018

International Conference on Computational Science

and Engineering (ICCSE 2018) (pp.118-124).

Department of Information Science and Technology,

East China University of Political Science and Law;

Song, K. & Gong, Z. W., (2023). Social network group

decision making method based on cloud model and

PageRank algorithm. Operation research and

management (05),56-61.

Wen, Z. C., (2022). Research on the coupling evaluation

model of structural coal reservoir heterogeneity by

hierarchy-entropy weight method (Ph. D. dissertation,

China University of Mining and

Technology).PhDhttps://link.cnki.net/doi/10.27623/d.c

nki.gzkyu.2022.000092doi:10.27623/d.cnki.gzkyu.202

2.000092.

Xu, G. L. & Huang, Z. Y., (2024). Typhoon disaster risk

assessment in Guangxi based on AHP-entropy

combination weighting. Journal of Nanning Normal

University (Natural Science Edition) (04),86-

96.doi:10.16601/j.cnki.issn2096-7330.2024.04.011.

Xu, L., (2023). Research on Key issues of SPID Integrated

Chinese and Western Medicine Diagnosis and

Treatment Guidelines based on Delphi Method and

GRADE System methodology (Doctoral dissertation,

Beijing University of Chinese Medicine).

PhDhttps://link.cnki.net/doi/10.26973/d.cnki.gbjzu.20

23.000037doi:10.26973/d.cnki.gbjzu.2023.000037.

Xu, Y. G., (2024). Application of big data and intelligent

algorithm in power grid fault prediction and diagnosis.

Integrated circuit application (11),192-

193.doi:10.19339/j.issn.1674-2583.2024.11.088.

Yan, C., (2024). Application of intelligent algorithm in

supply chain management optimization. Electronic

technology (10),268-269.

Yang, J. N. & Zhang, P. F., (2024). Construction risk

evaluation of intelligent renovation project of office

building based on AHP-entropy weight method and

grey clustering comprehensive evaluation. Project

management technique (03),19-26.

Yang, S. J., Wang, R. J., Zheng, J. H., Zhao, P. Y., Han, W.

Q., Mao, X. R. & Fan, H., (2024). Cotton growth

monitoring combined with coefficient of variation

method and machine learning model. Surveying and

mapping bulletin (07),111-116.doi:10.13474/j.cnki.11-

2246.2024.0720.

ICDSE 2025 - The International Conference on Data Science and Engineering

672