Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Journal of Intelligent & Fuzzy Systems 37 (2019) 6021–6028 6021

DOI:10.3233/JIFS-179184
IOS Press

Computer network security technology


based on artificial intelligence
Tao Hua∗ and Laixiang Li
Experiment Teaching and Network Information Center, Liaocheng University, Liaocheng, China

Abstract. With the rapid development of the Internet, the use of computer networks is becoming more and more frequent.
However, in the process of network use, various vulnerabilities and some other factors lead to network security problems
occur from time to time. How to make the network provide efficient and rapid services for people and also meet different
levels of security issues has become a problem. Based on this, the use of artificial intelligence technology has been studied,
and the system detection algorithm based on the concept of artificial intelligence has been proposed. Integrating artificial
intelligence into the field of network security can not only improve the overall performance of the network, but also effectively
and reliably guarantee the security. Firstly, according to the characteristics of computer network system, an immune detection
algorithm is proposed to evaluate the security of the system. Secondly, based on the analysis of computer network security
technology, a system detection algorithm based on artificial intelligence concept is further proposed. Then, data mining is
carried out based on Prefix, and the intrusion lines and their correlation are analyzed. Finally, the effectiveness of the algorithm
is analyzed experimentally. The results show that the proposed algorithm has good accuracy and adaptability, and can play a
good supporting role in the security detection of computer network system.

Keywords: Artificial intelligence, computer network, security technology

1. Introduction virus intrusion. In order to improve the security of


computer networks in, a new detection method must
With the continuous development of science be adopted [2]. The operation of traditional detection
and technology, computer network security is also methods is very difficult, and the whole computer
constantly improving. However, computer network cannot be evaluated. The evaluation accuracy is not
security threats are also increasing, such as a large high, and the whole computer cannot be accurately
number of Internet viruses, as well as a large number conceived [3].
of criminal and hacker attacks on computers, which With the improvement of computer network secu-
makes a large number of computer network security rity requirements in, an evaluation method based on
risks in. And illegal threats will do harm to the whole artificial intelligence technology–immune detection
Internet, causing economic losses and even casualties algorithm, has emerged [4]. Some intelligent algo-
[1]. From the traditional computer network security rithms are organically combined with the algorithm
detection methods in, simple linear standards cannot to establish new evolutionary theory and algorithm
meet the requirements of today’s computer industry to improve the overall performance of the algorithm.
Immune detection algorithm applies immune con-
cept and its theory to genetic algorithm. On the
∗ Corresponding author. Tao Hua, Experiment Teaching and premise of retaining the excellent characteristics of
Network Information Center, Liaocheng University, Liaocheng, the original algorithm, some feature information or
China. E-mail: huat@lcu.edu.cn. knowledge in the problem to be solved are selectively

ISSN 1064-1246/19/$35.00 © 2019 – IOS Press and the authors. All rights reserved
6022 T. Hua and L. Li / Computer network security technology based on artificial intelligence

and purposefully utilized to restrain the degrada- electronic visa agents, user authentication products,
tion in the optimization process, so as to realize security management centers, intrusion detection sys-
the network security detection problem of computer tems, security databases, security operating systems,
systems. Therefore, the computer network security vulnerability scanning, anti-virus, identity authenti-
technology is studied by using immune detection cation and other security products [7]. Hashemi M
algorithm based on computer system security and R proposed an artificial intelligence model for trop-
system security detection algorithm based on artifi- ical storm surge prediction. In order to keep the
cial intelligence concept. meteorological data secret, the security device of the
The research is mainly divided into three parts: system has been designed, and good results have
The first part discusses the current situation of net- been achieved [8]. Lin Y simulated human thinking
work security in China, the advantages of artificial and behavior by using artificial intelligence technol-
intelligence and the application of artificial intelli- ogy to solve complex problems that require human
gence in network security in. It is a very novel idea experts to solve [9]. Feng X integrated artificial intel-
to apply artificial intelligence to computer network ligence into the field of network security. It has been
security technology, which will help to improve com- found that it cannot only improve the overall perfor-
puter network security technology. In the second part, mance of the network, but also effectively and reliably
combined with the concept of artificial intelligence, guarantee the security. Feng X believed that agent
immune detection algorithm and system detection technology plays an important role in artificial intel-
algorithm are proposed, and computer network secu- ligence and constantly promotes the improvement of
rity in is further optimized and analyzed. In order to network performance [10]. Xue B proposed a P2P
solve the problems of computer network security in trust management model based on improved evidence
an all-round way, so as to provide more secure net- combination rules, which solves the problems of P2P
work technology services. The third part carries out system that is difficult to deal with malicious node
experimental design and verification analysis of the attacks and uncertain information effectively [11].
proposed algorithm to verify the effectiveness of the Cui Y proposed a method of evidence combination
algorithm. rules based on the criterion of complement and share.
It is found that this method could solve the problems
of evidence combination conflicts and rules defects
2. Related work in information fusion [12]. Zhang G took the key
technology of network security management based
According to the related research results of com- on artificial intelligence theory as the research object,
puter network security technology at home and and anomaly detection method based on clonal selec-
abroad, the leading countries of information secu- tion fuzzy clustering algorithm was proposed. The
rity technology are mainly the developed western research proves that this algorithm can effectively
countries such as the United States, France, Israel, solve the problems of low efficiency of anomaly
the United Kingdom, Denmark, Switzerland and so detection and high false alarm rate [13]. Starting from
on. Zhou Y, a scholar, has studied the application of the current situation of network security manage-
artificial intelligence in mobile network technology, ment based on artificial intelligence theory, Wang F
and found that these countries have great advantages put forward anomaly detection method and P2P trust
in computer network hardware technology, espe- model, analyzing the operation process of modern
cially chip technology. China has some research and network security management technology, and found
products in this area, but the key technologies and that this method provided theoretical support for solv-
main equipment depend on foreign countries [5]. ing network security problems and hidden dangers
The on-line system of artificial intelligence technol- efficiently [14]. Masaki N analyzed the key tech-
ogy is designed by Chen Z. It uses cryptography nology of computer network security management
theory, security protocol theory, security architec- based on artificial intelligence theory and the spe-
ture theory and information confrontation theory to cific application of artificial intelligence in computer
explore the security of user information security [6]. network technology. The research considers that the
Yan D has found that the leading advantages of application of artificial intelligence in computer net-
computer network security technology in developed work technology provides a theoretical reference for
countries are mainly concentrated on firewalls, secu- the study of the application of artificial intelligence
rity routers, virtual private networks, security servers, [15].
T. Hua and L. Li / Computer network security technology based on artificial intelligence 6023

Generally speaking, although scholars have stud- and optimize system environment. In the narrow
ied secure operating system, secure database, sense, immune computing refers to various com-
multi-level security mechanism and anti-virus in puter immune algorithms with biological immune
recent years, the autonomous security kernel is con- algorithm, immune network model and evolutionary
trolled by human [16]. The solution of network model, such as Multi-Agent immune model, ARTIS
security is a comprehensive problem involving many distributed monitoring model, immune artificial neu-
factors, such as technology, products and manage- ral network and so on.
ment, which cannot be accomplished overnight [17]. Firstly, AA is an important component of multi-
Therefore, in order to build a hundred-year-old agent intrusion detection system based on resource
national information security system and effectively availability task allocation strategy. The load bal-
guarantee national security, social stability and eco- ancing ability of the whole system is regulated by
nomic development, public awareness of information AA. The main task of AA is to collect TZ data after
security and the quality and ability of informa- each DCA pretreatment, and distribute the TZ data to
tion system research, development, production, use, the appropriate MDA according to the current load
maintenance and management must be improved as situation and the selection strategy of each MDA.
soon as possible, establishing our own technical stan- According to the actual situation of the system, an
dards and systems, and constantly developing and improved resource availability scheduling strategy is
updating network security technologies and products proposed. In this system, whether the TZ data to be
to meet the emerging challenges that endanger net- processed is assigned to an MDA is determined by
work information security [18–20]. the comprehensive performance indicators of MDA,
including detection accuracy A and resource effi-
ciency E. Therefore,
3. Computer network security technology
and system detection algorithms Ui = Ai × Ei (1)
Among them, Ui is the resource availability of the
3.1. Computer security system immune detection ith MDA, Ai is the detection accuracy of the ith MDA,
algorithms and Ei is the work efficiency of the ith MDA.

Immune computing is the research field of mod- Ai = AlarmRate(x)i × wi +MisreportRate(x)×w2


ern computer science. It extensively introduces the (2)
immune mechanism and model of medical immunol- Among which, AlarmRate(x) is the alarm rate of
ogy into the research and engineering of computer the ith MDA at x time, MisreportRate(x) is the false
intelligence, network science, computer control and alarm rate of the ith MDA at x time, where w1 , w2
security. Modern immune computing is also called are weight coefficients.
computer immunology. According to its definition Mumworked (y)
and application, it can be divided into generalized Ei = (3)
Numwordek (y) + Numdeplyed (y)
immune computing and narrow immune computing.
The earliest development and application of computer In which, Numworked (y) is the number of
immunity came from the computer immune system TZData processed by the ith MDA at y time, and
for virus prevention developed by IBM in the 1980 s. Numdeplyed (y) is the number of TZData processed
In the 1990 s, American scholars first elaborated the by the ith MDA at y time. AA calculates the resource
phenomena and definitions of immunity in the field availability for each MDA according to the above
of computer, pointed out the problems and functions resource availability calculation method, and decides
of immunity in every field involving security and the distribution mode of TZData to be processed. The
control, and put forward the concepts of computer specific steps of the proposed immune detection algo-
immunology and immune computing. Immunity in rithm are as follows: on the one hand, the negative
the field of computer is a function to identify nor- selection algorithm eliminates the detector tolerance
mal and abnormal substances and information, delete, response step, avoiding the disadvantage that the gen-
isolate or modify redundant or harmful information, erated immune antibodies need too long due to a
so as to maintain the security and stability of the large number of network packets. On the other hand,
system environment. To achieve the purpose of secu- the boundary between normal overload and abnormal
rity protection system, improve system performance overload is blurred, so the results of FANAb antibody
6024 T. Hua and L. Li / Computer network security technology based on artificial intelligence

detection include three states: normal, suspicious accesses the host inside the system, the value of V
and abnormal. Therefore, UAb antibody detection increases by 1, and when the internal host accesses
method is added to the system, which promotes the the host outside the system, the value of V decreases
experience of MDA to learn from each other, achieves by 1. The definition of active entropy can be obtained
the effective allocation of ultra-complex intrusion by applying the theory of entropy value to the above
detection tasks, and improves the intelligence of the probability set as shown in formula (4).
system and the ability of early warning for complex
problems. 
k
Ha = − (pi) log(pi) (4)
i=1
3.2. Computer network security technology
The method designed in this paper needs to adjust
The application of artificial intelligence in com- the size of detection window according to the change
puter network security is manifested in many aspects, of network traffic. It is not difficult to see that net-
one of which is the application of artificial intelli- work traffic changes with time, but from the overall
gence in computer network security management in. point of view and long-term perspective, there are
At present, a very important technology of computer certain rules of traffic change, it is obviously inappro-
network security management is firewall technol- priate to adopt a unified approach to different traffic.
ogy, and the core technology of firewall technology In order to give full play to the advantages of active
is intrusion detection technology. This technology entropy in network anomaly traffic monitoring, the
ensures that our internal information and personal following assumptions can be made on the basis of
information are not harmed by virus and Trojan horse selecting the size of detection window. If there are n
from outside, which ensures the safety of our Internet non-repetitive elements {a1 , a2 , . . . , an } in the set of
access. It is in this intrusion detection system that sev- length T, the set of occurrence times of each different
eral artificial intelligence technologies play their own value is {d1 , d2 , . . . , dn }. At this time, the formula for
unique roles, such as rule production expert system, calculating the entropy is as follows:
artificial neural network system, artificial immune n    
technology and so on. These technologies support  di di
H(x) = − log (5)
intrusion detection technology and the core of com- T T
i=1
puter network security management. Expert system
is the most widely used artificial intelligence technol- In this case, the number of times when the entropy
ogy. The administrator of this technology codes the value of small scale calculation reaches the limit is
intrusion characteristics into rules according to his more than that of large scale calculation. In order to
own experience, and then formulates corresponding prove the above hypothesis, the following deductions
prevention procedures according to these rules to pre- are made. When T is taken as the unit length, the
vent the intrusion. Artificial neural network system entropy value can be obtained as follows:
imitates human brain. It uses its own learning ability
n    
and adaptability to identify intrusions to ensure net- di di
H1 (x) = − log (6)
work security. Artificial immune technology imitates T T
i=1
the self-defense of the human body when it is injured,
and maintains the network completely according to When T/2 is taken as the unit length, the original
the same principle. set is divided into two parts. At this time, the entropy
values of the two parts are calculated as follows:
3.3. Computer network system security detection n/2    
 di di
algorithms based on artificial intelligence H2 (x) = − log (7)
T/2 T/2
i=1
Assuming that a network containing n hosts exists,
n    
the network is regarded as a whole system. Such a set
 di di
can be obtained, that is S = {s1 , s2 , . . . , sn }. Then, H3 (x) = − log T (8)
in this system, every host’s action can be recorded T/2 /2
n i= /2
as an action unit, i. e. v. When the host inside the
system interacts with the host outside the system, the After deforming H1 (x), the following results can
action unit begins to count. When the external host be obtained:
T. Hua and L. Li / Computer network security technology based on artificial intelligence 6025

n  
   middle management personnel, general management
di di
H1 (x) = − log personnel and first-line workers. A sample of 495
T T
i=1 employees is analyzed. The overall situation is shown

n   in Table 1 below.
1 di
=− T/2
(logdi − log T/2 + 1)
2
i=1
1 4.2. System security detection algorithms
= [H2 (x) + H3 (x)] − 1 (9) experiments and results
2
It is not difficult to see that if the number of times The research on the characteristics of the entropy
the entropy reaches the limit is 2Y when T/2 is value shows that when abnormal traffic occurs in the
taken as the unit length, then according to formula network, there are great differences or unique repre-
(6), the number of times that the entropy reaches sentations of different characteristic entropy values.
the limit when T is taken as the unit length will be In the process of research, it is also found that depend-
less than y. From the above analysis, it is not dif- ing on the change of one characteristic entropy value
ficult to draw the following conclusion: When the alone sometimes cannot distinguish distinct network
inter-cell scale is selected to reach the limit (early traffic. Therefore, more in-depth research (using five-
warning), the large-interval scale will not necessarily element entropy) has been carried out. The statistical
reach the limit (early warning). Therefore, different rules of different flow characteristics are found as
flow window intervals should be selected for different shown in Table 2. The table shows the change rule of
network characteristics (traffic distribution character- the characteristic entropy of different categories when
istics) in different time periods. A relatively small this kind of traffic anomaly occurs. Different anoma-
flow window interval is selected for large flow, while lous network traffic can be treated differently by the
a relatively large flow window interval is selected for different performance of the characteristic entropy
small flow. of different anomalous events, so as to realize the
monitoring of the network.
As shown in Fig. 1, when DDoS attacks occur,
4. Result analysis and discussion it is obvious that the source IP curve hovers around
a value above zero and fluctuates in a small range.
4.1. Experimental environment and data sources The target IP curve is close to zero. In the follow-up
study, it is found that when the number of attack hosts
In the experiment section, the development envi- increased, the source IP curve would hover around a
ronment, development tools and performance testing higher value or float up and down at a higher value.
software are introduced firstly. Then the performance The target IP curve keeps close to zero or even keeps
test of mobile payment system solution based on straight at zero. Therefore, the entropy value can not
encrypted short message authentication code is car- only be used as a criterion to judge the attack, but also
ried out. Firstly, the time of encryption and decryption be used to estimate the size of the botnet, such as the
is compared with that of RSA encrypted fixed-length source IP entropy value, the larger the value is.
data with different length keys. Taking the computer Through the analysis, after the botnet is detected,
network security of human resources information of we can mine and analyze the data flow such as Net-
ZD Company in as the research object, the com- flow that attacks the host in the area, and find the
pany’s existing human resources are divided into host information that has communicated with the
three categories according to the talent information of attacking host. After trusted hosts are removed, the
its company network: senior management personnel, remaining hosts are listed as suspicious controllers.

Table 1
Age structure
Age group <25 26–30 31–35 36–40 41–45 46–50 51–55 56–60 Total
Number of people 29 33 71 157 140 33 19 13 495
Proportion 6% 7% 14% 32% 28% 7% 4% 2% 100%
6026 T. Hua and L. Li / Computer network security technology based on artificial intelligence

Table 2
Reference table for change of characteristic entropy
Exception name H(X(SIP)) H(X(DIP)) H(X(SPT)) H(X(DPT)) H(X(PKT)) Explain
Download Smooth and Fluctuating, Fluctuating, Fluctuating, Stable
low low low low
Video player Stable Stable Stable Stable Stable Five kinds of entropy
wave type similar
Webpage video Volatility and Volatility and Volatility and Volatility and Volatility and Five types of entropy
stability stability stability stability stability wave type and Volatility
File transfer Very low, Very low, Very low, Very low, Very low, Five kinds of entropy
smooth smooth smooth smooth smooth value almost overlap
DoS Very low and Very low and Moderate and Very low and Very low and
smooth smooth smooth smooth smooth
DDoS Stable Very low and Moderate and Very low and Very low and The height of SIP is
smooth smooth smooth smooth closely related to the
size of the botnet

Table 3
RSA selects different key length encryption and decryption times
Clear text Key The encryption Decoding
size length algorithm algorithm
(bit) (bit) takes time (ms) takes time (ms)
128 128 4.4 4.9
256 11.0 10.4
512 24.9 22.1
1024 46.4 49.7

Table 4
Mobile client - authentication scheme and digital authentication
Fig. 1. Comparison of DDoS attacks of different sizes. (A. small response time comparison table
scale DDoS/DoS attack B. large-scale DDoS attack).
Views 10 20 30 40 50 60 70 80
Program
4.3. Expansion of computer system security 3DES 93.4 230.6 232.5 242.9 242.9 248.3 273.5 284.3
immune detection algorithms Program
AES 93.7 223.3 237.4 233.6 242.2 242.7 252.4 274.3
Program
The data in Tables 3 and 4 show how long it
takes to encrypt and decrypt plaintext of the same
size using different key lengths. The 128bit length using mobile terminals, while vertical coordinates
of encrypted plaintext used in experiment is mainly represent the average reaction time spent on RSA
considered that in practical applications, the length decryption in experiments.
of short message verification code is usually 6 digits, From the data Fig. 3 of the experimental results, it
so it is not necessary to encrypt too long plaintext. can be seen that the time of these numbers increases
Although encrypting and decrypting plaintext with linearly with the increase of concurrent requests, but
1024 bit length keys takes a little longer than other because the information capacity of the encrypted
length keys, it can ensure the security of authentica- files is relatively small, the overall trend of the average
tion codes when it cannot be violently cracked under encryption time is not obvious. This shows that the
the existing computing capacity conditions. And in use of the selected length of the key for the immune
practical application, the short message verification detection algorithm encryption used in short message
code will be invalid or retransmitted after 1 minute, so encryption will not affect the efficiency of the system.
it can ensure the security of transmission information. In order to validate the effectiveness of the algo-
The four curves in Fig. 2 represent authentication rithm, the model uses sequential allocation method,
codes encrypted in different digits. According to the GA allocation method and GASA allocation method
daily habits, the experiment used two digits, four dig- to plan the allocation order of mobile agents for net-
its, six digits and eight digits. Horizontal coordinates work storage resources, and the impact of mobile
represent the number of concurrent visits by users agent storage resources allocation method on network
T. Hua and L. Li / Computer network security technology based on artificial intelligence 6027

GA allocation method is between the two. It can be


concluded that: First, the optimized GASA alloca-
tion method can effectively plan the dispatch order of
mobile agents, effectively reduce the IO frequency,
and continuously improve the network storage per-
formance. Second, the global optimization capability
of GASA allocation method is much better than that
of GA allocation method. Resource allocation model
based on GASA allocation method can promote
better performance of network storage. Therefore,
GASA allocation method has better robustness and
Fig. 2. Comparison of DDoS attacks of different sizes. optimization effect than GA allocation method and
sequential allocation method.

5. Conclusion

With the continuous development of the computer


field, various technologies are constantly improv-
ing and upgrading. The use of artificial intelligence
as a new measurement means can make a correct
evaluation of the overall network level security of
the computer. Therefore, in this paper, the coastal
Fig. 3. Average time of authentication response using different
algorithm schemes for digital authentication.
computer network security technology based on arti-
ficial intelligence concept is studied. Based on the
computer network security technology, an immune
detection algorithm is proposed and a system security
detection algorithm is constructed. And it is applied
to the adaptive intrusion detection system. Finally, the
algorithm is tested experimentally. The experimental
results show that the algorithm designed has good
accuracy and adaptability, and the computer secu-
rity network system is effective. In sequential pattern
mining, the efficiency of the algorithm has been
greatly improved, especially for long sequential pat-
terns. After the improvement, the mining efficiency of
the algorithm is improved more significantly, the effi-
Fig. 4. Comparison of 10 frequency of three algorithms.
ciency of the intrusion detection system is improved,
and the optimization of the encryption protocol in the
storage performance is analyzed. In order to compare network security protocol is further improved, which
the changes and reasons of storage performance of lays a solid foundation for the development of net-
different algorithms in detail, the performance indi- work security. However, due to the limitations of time,
cators of IO frequency and delay are extracted and equipment, test means and other conditions, as well
quantified separately, as shown in Fig. 4. as the uncertainties of attacks on network information
From a global perspective, Fig. 4 shows that the security and the emergence of new threats over time,
IO frequency of network storage increases irregularly there are still many deficiencies in this study.
and linearly with time in the process of mobile agent
allocation under three different allocation methods.
However, from a vertical point of view, at the same Acknowledgments
time point, the IO frequency of GASA allocation
method is the lowest, and that of sequential allo- The study was supported by “CERNET Innovation
cation method is the highest. The IO frequency of Project, China (Grant No. NGII20170604)”.
6028 T. Hua and L. Li / Computer network security technology based on artificial intelligence

References coupled simulations, International Journal of Applied Earth


Observation & Geoinformation 6(9) (2018), 45–46.
[11] B. Xue and T. Liu, Research on emotional model of sports
[1] H.Q. Huang, L.I. Wei-Min and X. University, Research on arena based on artificial intelligence emotion calculation,
the structure of cognitive optical network based on artificial Cluster Computing 7(8) (2018), 1–7.
intelligence technology, Optical Communication Technol- [12] Y. Cui, Y. Ma and Z. Zhao, Research on data fusion algo-
ogy 11(4) (2016), 9–15. rithm and anti-collision algorithm based on internet of
[2] J. Xi, Research on IETM service rules based on artificial things, Future Generation Computer Systems 55(7) (2018),
intelligence technology, Computer Measurement & Control 34–35.
67(24) (2017), 39–45. [13] G. Zhang, R. Wang and G. Zhao, Extraction of vegeta-
[3] Y. Tang, L. Guo and Y. Xie, Research on STEM interdis- tion information in coastal ecological vulnerable areas from
ciplinary integration model based on educational artificial remote sensing data based on phenology parameters and
intelligence, China Educational Technology 7(23) (2017), object-oriented method, Transactions of the Chinese Society
718–731. of Agricultural Engineering 9(11) (2018), 5–6.
[4] J. Chen and E. Polytechnic, Analysis on the development [14] F. Wang, Computer art design based on artificial intelli-
status of agent technology based on artificial intelligence, gence, Cluster Computing 7(2) (2018), 1–7.
Computer Knowledge & Technology 56(8) (2016), 34–45. [15] N. Masaki, S. Hashimoto and M. Kawado, The number of
[5] Y. Zhou, Research on the application of artificial intelligence deaths by suicide after the Great East Japan Earthquake
in mobile network technology, Journal of Zhangjiakou based on demographic statistics in the coastal and non-of
Vocational & Technical College 7(1) (2016) 89–101. Iwate: Miyagi, and Fukushima prefectures, Nihon Koshusei
[6] Z. Chen, Design of online learning system based on artificial Zasshi 65(4) (2018), 164–169.
intelligence technology, China Computer & Communica- [16] H. Ning and H. Liu, Cyber-physical-social based security
tion 89(9) (2018), 102–110. architecture for future internet of things, Advances in Inter-
[7] D. Yan, Q. Zhou and J. Wang, Bayesian regularisation neu- net of Things 2(1) (2012), 1.
ral network based on artificial intelligence optimization, [17] R. Kling, Learning about information technologies and
International Journal of Production Research 55(8) (2016), social change: The contribution of social informatics, The
2266–2287. Information Society 16(3) (2000), 217–232.
[8] M.R. Hashemi, M.L. Spaulding and A. Shaw, An efficient [18] F.Y. Wang, Toward a paradigm shift in social computing:
artificial intelligence model for prediction of tropical storm The ACP approach, IEEE Intelligent Systems 22(5) (2007).
surge, Natural Hazards 82(1) (2016), 471–491. [19] I. Foster, C. Kesselman and S. Tuecke, The anatomy
[9] Y. Lin, Research on application of artificial intelligence of the grid: Enabling scalable virtual organizations, The
based on big data background in computer network tech- International Journal of High Performance Computing
nology, Journal of Jiujiang Vocational & Technical College Applications 15(3) (2001), 200–222.
77(3) (2018), 56–57. [20] A.K. Jain, A. Ross and S. Pankanti, Biometrics: A tool
[10] X. Feng, M. Li and B. Yin, Study of storm surge trends for information security, IEEE Transactions on Information
in typhoon-prone based on observations and surge-wave Forensics and Security 1(2) (2006), 125–143.

You might also like