See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/377399143

Machine Learning-Based Detection and Mitigation of XML SQL Injection

Attacks

Conference Paper · December 2023

DOI: 10.1109/GCITC60406.2023.10426458

CITATIONS READS

2 268

3 authors, including:

Rajat Dubey Ishu Sharma

Allianz Commercial Chitkara University
14 PUBLICATIONS 140 CITATIONS 102 PUBLICATIONS 585 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Rajat Dubey on 15 January 2024.

The user has requested enhancement of the downloaded file.

 Machine Learning-Based Detection and Mitigation
of XML SQL Injection Attacks
Vanshika Pahuja Rajat Dubey Ishu Sharma
Chitkara University Institute of Cybersecurity Expert, Chitkara Universiry institure of
Engineering and Technology, Allianz Commercial, Engineering and Technology,
Chitkara University, Austin, United States Chitkara University,
Punjab, India [email protected] Punjab, India
[email protected] [email protected]

Abstract— XML and SQL injection attacks are occurring vector attack for the purpose of exploiting the vulnerability of
very frequently nowadays as developers lack major security the document and database by injecting malicious SQL code
measures and awareness for the purpose of securing web at the backend database which will access all the important
applications and documents. Many factors are responsible for information of the database that was not supposed to or
these types of attacks and vulnerabilities are the main reasons intended as all this disruption occurs for some important
behind all these types of attacks like urbanisation, intention. Many crucial information may be lost due to such
environmental degradation and some of the market conditions type for injection attacks in every sector like finance,
are responsible for these types of vulnerabilities in the market education, management, and banking sector. Sensitive
and lead to such attacks. These may also allow the attackers to
company information and other important credentials may be
gain advantage to overlook the front end of the application by
taking advantage of all the vulnerabilities of input fields like
lost like salary details of employees, user lists and private
usernames, passwords, and important credentials. This customer details. Example of SQL injection attacks occur by
research paper covers the detailed study of the occurrence of inserting, updating and deletion of the data at the backend
XML injection attacks on front end development applications, database and attack can also be used as the main tool for
also the detection and mitigation policies employed in the changing the cookies and harming the data at the important
literature. The machine learning based model is presented in database queries of web application. SQL injection attacks
this case study for detecting XML SQL injection attacks and the pose the great threat to web applications and the backend
mitigation policies are also discussed. The classification machine databases as all the core data and code will destroy and all
learning employed in the research paper are also compared potential sensitive information also get corrupted by using
using metrics based on accuracy, precision, Specificity and F-1 these data. Web applications which are vulnerable to the SQL
Score. attacks may allow and permit all hackers and attackers to
completely access the databases and steal all the important
Keywords—Extensible Markup Language, SQL Injectio, information from the databases including sensitive
Network Security, Machine Learning, Classification, Secure information and other mandatory stuff [1].
Communication.
These web applications and XML documents lack all the
I. INTRODUCTION automatic privacy and security measures and highly efficient
XML injection attack is the major technique used for the and equipped testing methods and hence they allow all the
manipulation of actual XML document or application so that injection attacks whether in the form of XML injection attack
all the data of the document will get affected due to which or SQL injection attack. Precise detection of attacks might
functionalities will get affected. It mainly occurs due to the occur and these may be allowed without using the false
injection of malicious code especially by the malicious positives. XML injection attacks enable all the unvalidated
hackers for the fulfilment of intention to break down their and other data to construct malicious queries into database that
privacy. User supplied input to the code is not treated allow all the attackers to modify and read the documents and
accordingly for the cleaning and sanitization of the data and disrupt them whenever they want or required. Many front end
hence all the malicious code will be erased from the actual and backend may be destroyed and vulnerabilities can be
document and application. Sanitization of the code is found in the applications which contain and rely on some of
necessary before adding it into XML document and hence, all the XML based fields and database or flat files. XML based
the code will get affected and breaches its security. This type application programming interfaces will also get affected by
of attack usually works by the XML entity containing works using the XML injection attacks. Vulnerabilities are the major
for the external entity which will get processed for the external reason behind their increasing rate of these attacks nowadays
configured parser. XML injection attack mainly works by as developers lack knowledge of security in the web
modifying the schema definition of the document and applications due to which many database queries and XML
application for the purpose of processing of XML data. These documents will get affected and leads to the ruin of all
types of attacks would result in the loss of important data, database. All these attacks are mainly done for manipulating
important format credentials and ultimately lead to the logic the web server of the application that can compromise the
errors. One of the most important examples of the XML database. Subverting application logic, retrieving hidden data
injection attacks would be seen when some of the attacker of are some of the examples of SQL injection attacks.
hacker inject some of the malicious XML payload into the Vulnerabilities in the applications are the major reasons for
XML document or application that will retrieve all the account the unauthorised access to the important information to the
details and important data including all data like usernames websites which lead to the passing of unexpected code
and passwords. Exploitation of the vulnerability occurs using statements through the backend database exploitation. Major
such type of mechanisms and XML injection attacks which focus is on the vulnerabilities issues due to which violation of
will read out all the sensitive information from the database the database occurs and leads to the whole disruption [2].
that will create chaos and major loss of organisation. SQL
injection attack is also same technique used as the common
 The impact of SQL injection attacks and XML injection XML and SQL injection. Governmental agencies are
attack is totally devastating and far-reaching impact on the traditional in terms of security and due to which many attacks
business due to which it has many disadvantages and results happen at their agencies for deriving crucial information and
produced are wrong due to the use of such attacks. This data [4]. [5].
successful attack has some of the major disadvantages such as
unauthorised viewing of user data and entire tables of the
database and hence the attacker gains administrative rights to
the backend database. These results produced with the help of
attacks are harmful and detrimental for business. SQL
injection attacks can lead to the modification, destruction and
theft of the sensitive information such as personally available
data, information of the user which are usernames and
passwords. In section II, background of the SQL and XML
injection attacks is being discussed giving overview of all
attacks in some of the important sectors like finance, banking,
and organisational sectors. In section III, literature review is
given giving in depth detail of these attacks nowadays.
II. BACKGROUND OF SQL XML INJECTION ATTACKS
SQL injection attacks including retrieving hidden data
which can modify all the important data and modify any of
the SQL query to return the essential results from the attacks.
Another example of these types of attacks includes subverting
the application logic where any of the person can make
amendments and change the query whenever we wanted and
this can also interfere in between the main logic of the Fig. 1. Code for XML Injection Testing
application leading to total disruption and criteria. UNION
attacks are the common example of such types of attacks In figure 1, code is described depicting malicious code
where it is of the utmost importance to the attackers and testing in which tester usually injects the XML document into
hackers to find the main data and code from the database and the application. XML injection points are identified using this
ruin them for their ow purpose. Attackers find some of the code and injected into application using an XML style
place where they can easily penetrate in the corporate systems communication. People from outside of the organization or
and breach into some important corporate systems such as within the organization find the vulnerabilities of the
main example taken as Eleven Breach which is the main organization and trying to deploy them and extract the
attack used in SQL injection attack used for penetration into confidential information of the government organization.
corporate systems in one of the major companies mainly used More than 80% of the government sectors and organizations
for stealing the information about the credit card details and are prone to these types of attacks due to use of traditional
this can be done nearly to the 130 million credit card details.. methodologies of security and preventing such type of
Undetected code of the SQL injection may be the main storm attacks. These sectors are more prone to common web
for creating the strong damage. US government recently application attacks such as SQL injection attacks.
detected SQL injection attack during on of the data review Industrialized countries are even more vulnerable to such find
session occurring in their agency. There are some unusual of attacks as they have more data and less security measures
indicators occurred which shows the injection attacks and to prevent their organization from these cyber threats and
government agency comes to know about this. Hackers criminals. Malicious data possibilities have led to great
mostly used these types of injection attacks to target the exposure to attacks on Norwegian government agencies and
government data and steal their information which might be even more promising agencies and industries [6]. All
sued for their future purposes. They usually used the SQL organizations can sustain without much communication and
injection to derive the information from the backend SQL security measures which has been linking to the large sources
databases which contain vulnerabilities and they are exposed of essential data over globe due to which technology experts
to threats. One such main attack occurred in European state are struggling for the high performance of the machines
agency in which some of the anonymous hackers target the which has catered the needs of high-tech firms. Banking
main agency while having the suspicious code of SQL and sectors in India have been targeted by many cyber criminals
hence due to this vulnerability of European agency and and they adopt many methodologies for ruining their
attacks, hackers gained access to the ES database which is operations. Automated Teller Machine (ATM) or credit card
most important database and therefore leaked main data frauds are common by using injection attacks [7].
which contains 8000 names of the employees and others, their
III. LITERATURE REVIEW
emails, and passwords. This attack at the European agency is
too devastating and harmful which has led to the main loss to Security flaws in the web applications of organizations are
the agency. the main reason of the cyber threats and prone to injection
Web application occurs as the main interface between attacks. Vulnerabilities in the web applications of documents
users and data and database driven applications are more are more prone to cyber attacks like XML injection attacks
prone to the attacks. As web applications are assumed as the and SQL injection attacks. Various measures should be taken
good target to the attacks, many attacks happen by using to prevent these attacks and prevent all the organisations
whether private or public from these attacks. More measures attacks. Cyber security poses the great importance in
should be taken to prevent these attacks by taking accurate preventing the data and information from these attacks in this
measures and methodologies so that further attacks and threats world of major attacks. It is most dangerous vulnerability due
should be prevented at the earliest. This is of utmost to its multiple applications and rapid changes over the world
importance that cyber-attacks should be prevented prior leading to the financial loss, data leakage and other effects
before loss occurs. In paper [8], Structured Query Language making the web applications more vulnerable to these attacks.
(SQL) injection attacks are discussed and new framework is In paper [13], black box testing is used for the prevention of
being designed for the prevention of such types of attacks. SQL injection attacks. SQL injection attacks are diverse and
These injection attacks pose the great risk and they are most having fast variations over the other making it more dramatic
challenging attacks on the security of the databases leading to and dangerous view. These attacks allowed the hackers and
the disruption of normal operations and other important attackers to gain unauthorized access to the web application
information which is more sensitive and confidential. SQL database. Automatic black box testing program has been
attacks provide ample number of opportunities to the launched for the automatically testing the task of SQL
malicious hackers and attackers to exploit the data and other injection to eradicate and remove completely the SQL
databases particularly the important information like client injection attacks.
data and personal data. Incorrect use of the dynamic SQL and
lack of input validation led to the inconsistent error handling In paper [14], detection of SQL injection attacks is done
mechanisms. Multi-phase algorithmic framework is being using more promising approach which is by using
proposed for the prevention of these type of attacks. In this probabilistic neural networks. Deep learning is used
framework, machine learning approach is sued for the everywhere nowadays for more accurate detection and
enhancement of the database security in the real time at initial improving its performance, and for this probabilistic neural
level. This proposed method is used to prevent the SQL network (PNN) is used for detecting the SQL injection attacks.
attacks, classification of the types of attacks at each step for Features are extracted by using the algorithms of neural
the purpose of detection methodology. In paper [9], detailed networks and detection of the SQL injection attack is done
explanation of the SQL injection attacks is given with making using more proposed model. This model has higher accuracy,
it one of the top 10 vulnerabilities of the web systems. It precision and recall which has employed cross-validation
mainly works and penetrates the database and for the purpose more effective than other models. In paper [15], more
of injecting the malicious code into the database. It has been proposed model is proposed for the purpose of SQL injection
one of the top vulnerabilities by exploiting the security flaws attack, this designed model is used for two purposes such as
of the system and ruin the normal operations. Machine signatures and regular expressions. This model is
learning algorithms are nowadays working for the detection of implemented and used ModSecurity which has more
such attacks so that immediate measures should be taken in enhanced version and these have some results which reveal
this regard for the quick exposure. Best strategy that is used that it is highly used and efficient than other methods and
for detection of these types of attacks is Naïve bayes, which is having more latency. Hence, it is more effective promised
mostly recommended algorithm nowadays with having more approach used for the detection of injection attacks and tighten
accuracy than before and mostly engaged for detecting attack the security so that no other attacks would happen from the
which has destroyed the database. Multiple machine learning next time. In paper [16], code injection attacks in the form of
methods used for detecting SQL injection defaults like logistic XML injection attacks are depicted in wireless sensor
regression, naïve bayes, extreme gradient boosting (XGBoost) networks and internet of things (IoT). With the more enhanced
and Random Forest. version and more updating by using these techniques,
developers can develop more optimised solution and internet
Deep learning architecture is used for the main purpose of of things systems which can be better protected against some
detecting SQL injection attacks which is based on the RNN of the emerging threats. In paper [17], vulnerability
autoencoder model and this architecture is deeply explained in assessment is required in web applications by using enhancing
[10]. These types of attacks exploit the vulnerabilities of the the burp suite with machine learning extension. These
system basically ruining the web application and other algorithms are used for scanning injection attacks in all web
documents allowing all attackers and hackers to execute the applications. It has proven to be the best approach for the long
unauthorised access to the commands and SQL queries of the short-term memory and safe to use.
backend database. Many machine learning models like
decision tree, random forest and logistic regression models are IV. RESEARCH METHODOLOGY
used for showing accuracy of the model with full effectiveness Attack classification is done using transfer learning and for
in comparison to other models. In paper [11], main technology classifying attacks which as XML SQL injection attacks. The
used for the detection of SQL injection attacks is being dataset used in the attack classification uses examples of SQL
introduced which is based on the semantic learning-based injection, XML External Entity (XXE) attacks and Denial of
attack detection. Injection vulnerabilities are at the top of Service (DoS) attacks for purpose of attack classification. This
many vulnerabilities of the web applications due to which it dataset also contains non-category for non-attacks so that non
has wide range of attacks. It has posed more challenging attacks should also be classified. These are the files included
problem nowadays having variety of attack patterns. in the dataset and loading of custom dataset is done by
synBERT is being introduced in this paper which is semantic defining the class and labelling them for data processing and
based learning model that embed sentence level semantic further classification. Category of names is defined as ‘SQL
information from all the SQL statements into one of the Injection’, ‘XXE’, ‘None’ and ‘DoS’ [18]. These all are used
embarking vectors. Classification performance of this model for training and loading of data and predicting the output
is remarkable. which is of attack classification. Training and validation
In paper [12], detection of the injection attacks is depicted accuracy acquired from the training of dataset will depict that
to enhance the security of the website from these client-side how accurate our model works. Correct images are predicted
and incorrect also based on the random prediction which will steps of AI training and building model for attack
bifurcate the classification into right and wrong classification. classification. Various visualizations are retrieved from the
This model will be able to do all the classification and separate model such as accuracy, loss, precision, and recall. Validation
cyber-attacks based on some features in XML files. Transfer and training accuracy are obtained from the model and
learning model is used which is very powerful natural confusion matrix retrieved to access different model’s
language model used especially for classification of attacks. It performance and other important credentials. Natural
is the small and cheap model used which is based on some language processing is technique used in getting the results
architecture. This type of distillation used in the model has which is a part of artificial intelligence and training of the NLP
performed during the initial phase for higher efficiency and model which classify cyber-attacks and make applicable in
reducing the model by less rate [19]. This model is used in real world. Figure 2 depicts the proposed model for evaluation
tasks where speed and efficiency matter a lot and it is an and attack detection of SQL injection attacks using transfer
excellent choice for sentiment analysis tasks. This model learning techniques so that categorization occurs which
retains all the features and more enhanced updates of transfer bifurcates malicious and non-malicious code. Data pre-
learning model and it has ability to understand meaning of all processing is done on dataset which includes all processes like
words in accordance to other fewer layers and other reduction feature extraction, normalization, standardisation and data
of size. It is the method which includes the transferring of encoding. Encoding of the dataset is done using parameters
knowledge by training with larger and accurate models. This due to which all data is recognised and encoded. Transfer
efficiency is achieved with the training of model with higher learning mechanism is used which has detected the attack as
accuracy. SQL injection attack occurs by injecting the malicious or non-malicious. Hence, machine learning
malicious code into application such that it can be related to techniques help to identify the malcious code and prevent the
the database and run as the main SQL query for performing database.
all operations at database with the bad intent. This has been
launched for the purpose of gaining unauthorized access to the V. RESULTS AND DISCUSSION
malicious hackers and attackers due to which all data will be SQL injection dataset was done from the Kaggle and
lost and corrupted and even confidential information will be attack detection is done on it so that malicious and harmful
stolen. Some of the important information like credit card code should be detected as soon as possible which will not
details, bank details, usernames, and passwords. These are not affect other codes and documents. Transfer learning
only common for the people but they are very dangerous and methodologies are used which has detected the attack and
harmful as essential data will be lost in these attacks. XML many parameters are obtained by using these parameters. It is
entities are vulnerabilities which enable all attackers for the
the comprehensive approach to fulfill our task otherwise
purpose of viewing files and application sever system, this
will also lead to the interaction with backend and systems. many files and documents will be affected. It is defense
Denial of service attacks were used by the attackers for approach for safeguarding the servers and maintaining
making the resource as well as machine unavailable so that no critical data and other applications.
other work would be done at that and service remains A. Precision
unavailable. These all types of files are included in the dataset
so that attack classification takes place according to the dataset Precision is the parameter obtained by the model which
and attacks. will assess the ability of the model to make it more reliable
and give positive predictions. It works by comparison and this
follows the cognitive approach to divide the correctly
predicted positive outcomes by total of correctly predicted
positive and negative outcomes. Accessibility of model will
be defined by using this approach, and a great precision value
implies more defining approach which says that a model with
great precision value give the model accurate model
predictions and giving great percentage of model positive
predictions. Precision is mainly used in the situations where
minimizing false positives is necessary and crucial part. This
metric is of great importance and it is combined with other
metrics for making it more promising approach. In equation
1, TPXSIA shows the true positive value and FPNBD shows the
false positive approach of model of attack detection.

= (1)

Fig. 2. Proposed Model for XML SQL attack detection using transfer Figure 3 shows the comparison of accuracy scores of all three
learning techniques
machine learning algorithms which are decision tree
Artificial intelligence (AI) is used for the purpose of classifier, linear support vector classification and naïve
attack classification which includes various steps so that work bayes, for more efficient results. The main purpose and goal
should be accomplished and attack classification occurs. AI of this algorithm is to maintain the algorithm and locate it at
includes various steps so that classification occurs with an its position. Accuracy scores are maintained under many
intention for retrieving the results from it. Data processing, datasets where x-axis represents the various scenarios and y-
model training, evaluation and saving of the model are few axis depicts the precision score having range from 0 to 1.
When examined based on precision, it is found that linear classifier, support vector classification and naïve bayes,
support vector machine classification has the higher precision linear support vector classification has more F1 Score
than other algorithms. However, other aspects are also
responsible for the attack detection and mechanisms like C. Accuracy
efficiency and scalability are more considerable approach Accuracy is the main parameter of the machine learning
which should be considered at any cost and will not be where it will access overall performance and proficiency of
neglected. the categorization model. It is expressed in the form of
percentage calculated by dividing all the accurate projections
by the number of false projections. When higher accuracy is
predicted then it will indicate higher indications and lower
accuracy will give the false predictions and give the incorrect
proportion of false predictions. It is highly valued, statistical
with more balanced class distributions.
In equation 4, equation of accuracy is defined by using
below defined equation.
,
) * + = (3)
, ,

Figure 5 shows the accuracy results of all machine learning

Fig. 3. Comparative analysis of precision of decision tree classifer, linear algorithms showing accurate measures and prescribing it. As
support vector machine classification and Naïve Bayes. seen in the recall, precision and F1-score, accuracy is also
more in the case of linear support vector classification.
B. F1 Score Hence, accuracy is the main measure defined making it more
Recall stands for the machine learning statistical model algorithmic choice for prediction and its computing
and indicator of the model at the rate of positive outcomes of efficiency can be increased with including of more measures
model, it is assumed as the statistical learning within the like F1-score, recall and sensitivity.
machine learning which will factorize based on performance.
F1 score assigns the single number which is numerical
number ranging from 0 to 1 and it is employed when there is
large imbalance. It is assigned value for assigning proper
value for identifying positive cases precision and
This test measures will explain that how well model behaves
and assumes for identifying positive examples. In depth
assessment of the model is done by using F1 score with other
favorable measures like accuracy and precision.

Fig. 5. Comparative analysis of accuracy of decision tree classifer, linear

support vector machine classification and Naïve Bayes.

D. Sensitivity
It is the important measure which will give the possibility
of the true positive rate. It is the main and essential
informative approach which will give the probability of the
positive test result and conditioning is done on some of the
Fig. 4. Comparative analysis of F1- score of decision tree classifer, linear individuals of negative test results. Sensitivity will quantify
support vector machine classification and Naïve Bayes. and predicts the true positive results by giving it more
classifying approach. It is main approach for having ability
In equation 2, two terminologies are prescribed which tell for determining the cases in the correct manner. It is
about precision and recall used in the equation. calculated based on proportion of correctly positive values.
shows Precision and shows Figure 6 depicts the sensitivity of all machine learning
Recall of attack detection of SQL injection attacks. algorithms which will depict the main measure and it will be
!"#"$% ∗ & !'((
1 = 2∗
!"#"$% & !'((
(2) represented by the graphs giving the exact details

Figure 4 represents the comparison of recall of three machine

learning algorithms which are prescribed as decision tree
 be taken in this regard. The changing environment and more
changes into the security measures needs some innovation for
improving the security measures and more secure documents
by preventing injection attacks.
REFERENCES
[1] Halfond, William G., Jeremy Viegas, and Alessandro Orso. "A
classification of SQL-injection attacks and countermeasures."
In Proceedings of the IEEE international symposium on secure software
engineering, vol. 1, pp. 13-15. IEEE, 2006.
[2] Kemalis, Konstantinos, and Theodores Tzouramanis. "SQL-IDS: a
specification-based approach for SQL-injection detection." In Proceedings
of the 2008 ACM symposium on Applied computing, pp. 2153-2158. 2008.
[3] Abdullayev, Vugar, and Alok Singh Chauhan. "SQL Injection Attack:
Quick View." Mesopotamian Journal of CyberSecurity 2023 (2023): 30-34.
Fig. 6. Comparative analysis of sensitivity of decision tree classifer, [4] A. Maraj, E. Rogova, G. Jakupi and X. Grajqevci, "Testing techniques
linear support vector machine classification and Naïve Bayes. and analysis of SQL injection attacks," 2017 2nd International Conference
on Knowledge Engineering and Applications (ICKEA), London, UK, 2017,
E. Specificity pp. 55-59, doi: 10.1109/ICKEA.2017.8169902.
[5] A. Pomeroy and Q. Tan, "Effective SQL Injection Attack Reconstruction
Specificity is an important measure which is the ratio of Using Network Recording," 2011 IEEE 11th International Conference on
true negative outcomes for all the negative outcomes. It is of Computer and Information Technology, Paphos, Cyprus, 2011, pp. 552-556,
main interest which deal with the accuracy of negative rate doi: 10.1109/CIT.2011.103.
and its higher rate of outcome will give description about [6] Moen, V., Klingsheim, A.N., Simonsen, K.I.F. and Hole, K.J., 2007.
Vulnerabilities in e-governments. International Journal of Electronic
positive and negative outcome. Figure 7 explains specificity Security and Digital Forensics, 1(1), pp.89-100.
measure of all three machine learning techniques and it shows [7] Bamrara, Dr Atul, Gajendra Singh, and Mamta Bhatt. "Cyber attacks and
that linear support vector classification has the higher defense strategies in India: An empirical assessment of banking
measure than other classifiers which are decision tree and sector." Available at SSRN 2488413 (2013).
[8] A. A. Ashlam, A. Badii and F. Stahl, "Multi-Phase Algorithmic
naïve bayes. Framework to Prevent SQL Injection Attacks using Improved Machine
learning and Deep learning to Enhance Database security in Real-time," 2022
15th International Conference on Security of Information and Networks
(SIN), Sousse, Tunisia, 2022, pp. 01-04, doi:
10.1109/SIN56466.2022.9970504.
[9] P. Roy, R. Kumar and P. Rani, "SQL Injection Attack Detection by
Machine Learning Classifier," 2022 International Conference on Applied
Artificial Intelligence and Computing (ICAAIC), Salem, India, 2022, pp.
394-400, doi: 10.1109/ICAAIC53929.2022.9792964.
[10] Alghawazi, Maha, Daniyal Alghazzawi, and Suaad Alarifi. "Deep
Learning Architecture for Detecting SQL Injection Attacks Based on RNN
Autoencoder Model." Mathematics 11, no. 15 (2023): 3286.
[11] Lu, Dongzhe, Jinlong Fei, and Long Liu. "A Semantic Learning-Based
SQL Injection Attack Detection Technology." Electronics 12, no. 6 (2023):
1344.
[12] AL-Maliki, Manar Hasan Ali, and Mahdi Nsaif Jasim. "Review of SQL
injection attacks: Detection, to enhance the security of the website from
client-side attacks." International Journal of Nonlinear Analysis and
Applications 13, no. 1 (2022): 3773-3782.
Fig. 7. Comparative analysis of specificity of decision tree classifer, [13] Thombare, Bhakti M., and Dr Rashmi Soni. "Prevention of SQL
linear support vector machine classification and Naïve Bayes. injection attack by using black box testing." In Proceedings of the 23rd
International Conference on Distributed Computing and Networking, pp.
VI. CONCLUSION 266-272. 2022.
[14] Alarfaj, Fawaz Khaled, and Nayeem Ahmad Khan. "Enhancing the
In conclusion, SQL XML injection attacks can be detected Performance of SQL Injection Attack Detection through Probabilistic Neural
with the use of machine learning techniques which are Networks." Applied Sciences 13, no. 7 (2023): 4365.
decision tree classifier, support vector machine classification [15] Alotaibi, Fahad M., and Vassilios G. Vassilakis. "Toward an SDN-
Based Web Application Firewall: Defending against SQL Injection
and naïve bayes. SQL injection attacks are done by injecting Attacks." Future Internet 15, no. 5 (2023): 170.
wrong code and malicious file into it which has destroyed the [16] Noman, Haitham Ameen, and Osama MF Abu-Sharkh. "Code Injection
file and ruin its functionalities. While all the algorithms dive Attacks in Wireless-Based Internet of Things (IoT): A Comprehensive
into the problem and wants to prove the results better than Review and Practical Implementations." Sensors 23, no. 13 (2023): 6067.
[17] Thaqi, Rrezearta, Kamer Vishi, and Blerim Rexha. "Enhancing Burp
other, but linear support vector machine classification has Suite with Machine Learning Extension for Vulnerability Assessment of
more scope than others and it has been proved with the Web Applications." Journal of Applied Security Research (2022): 1-19.
measures like precision, F1score and accuracy. Many [18] S. Rani, D. Koundal, fnm Kavita, M. F. Ijaz, M. Elhoseny, and M. I.
advance feature machine learning techniques and main Alghamdi, “An optimized framework for WSN routing in the context of
industry 4.0,” Sensors, vol. 21, no. 19, p. 6474, 2021.
corporate domain knowledge have make some changes and [19] P. S. Rathore, J. M. Chatterjee, A. Kumar, and R. Sujatha, “Energy-
improve the domain knowledge but these injection attacks efficient cluster head selection through relay approach for WSN,” J
need to be continued and some immediate measures need to Supercomput, vol. 77, pp. 7649–7675, 2021.

View publication stats