INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 2280
www.rsisinternational.org
Intelligent Detection Approaches for Securing E-Banking Platforms
*
*
Corresponding Author
DOI:
https://doi.org/10.47772/IJRISS.2025.910000188
Received: 02 October 2025; Accepted: 08 October 2025; Published: 07 November 2025
ABSTRACT
E-banking has introduced fresh innovations in finances because of the new conveniences, efficiencies, and global
access. Conversely, the fast spread of e-banking has increased phishing attacks, which are a cyber-attack on
banking sites. This paper concentrates on the intelligent detection of phishing websites to protect e-banking
systems, particularly the assessment and comparison of algorithms on intelligent detection. The UCI Phishing
Dataset URL, content, and behavioral features were diverse and applied in training and testing different machine-
learning models. Support Vector Machines, Random Forest, Neural Networks, XGBoost, and Hybrid
performance models were evaluated in terms of accuracy, precision, recall, F1-score, and AUC. Although SVM
showed a mediocre detection risk, the Random Forest and Neural Networks proved to be significantly more
reliable, whereas XGBoost exhibited the highest performance due to its accuracy and scalability of performance.
All tests yielded the most reliable results, and hybrid systems achieved the highest metrics and performance.
This is revealed the most reliable detection and control systems of phishing threats in e-banking system since
phishing frauds hugely make use of vulnerabilities in e-banking. In online banking systems, phishing frauds are
best controlled by hybrid systems and intelligent detection systems. Some of the practical refinements provided
by this study are improving fraud prevention, customer confidence, and regulatory compliance within financial
organizations. The investigations of the future must focus on creating mechanisms of fraud detecting in real-
time, using larger and more diverse data sets, and more adaptable and adaptable learning systems that can evolve
with phishing attacks to sustain the digital banking protection mechanisms.
Keywords: Phishing Detection, E-Banking Security, Machine Learning Algorithms, Hybrid Models.
INTRODUCTION
The emergence of new technologies has significant consequences bringing even bigger changes in financial
services across the globe and the spread of electronic banking systems. Customers can easily and conveniently
access banking services and complete numerous financial operations in any location where they transact, pay
their bills, transfer funds, and manage their accounts through e-banking systems (Kumari & Nagarjan, 2022).
These services give the customers convenience and this led to the actualization of the e-banking systems in the
modern automated financial systems. Above all, e- banking systems facilitate financial inclusion to the emerging
economies. Nevertheless, convenience and ease of use of e-banking services has also led to jeopardy of the safety
of these systems as has become a key priority to consumers and a growing number of financial service providers
(Money & Iyoha, 2025). A plethora of threats attacks e-banking systems, and phishing can be considered one of
the most common and the most advanced. The fraudsters use online banking portals and masquerading as clients
to cheat people and persuade them to provide them with sensitive banking details. The phishing attacks that
cybercriminals use are dynamic, advanced, and entrenched in the online exploitation of the human factor, which
is hard to detect or counter with the normal and available security (Pinjarkar et al., 2024). Successful phishing
attacks cause customers to lose money and tarnishes the reputation of affected financial institutions, which also
loses credibility in the eyes of the general population and regulatory oversight. The current strategies that are
being used to deal with phishing attacks (blacklisting, heuristic systems and manual reporting) are ineffective
when it comes to countering the increased and more complex ways of attacking that are collectively being used
against Phishing Websites
Okonkwo Chisom Michael., Ngene Chidiebere David., Onyedeke, Obinna Cyril
Enugu State University of Science and Technology, ESUT. Department of Computer Science, Nigeria
1
1
1
Department of Computer Science, University of Nigeria, Nsukka, Nigeria
2
2
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 2281
www.rsisinternational.org
by the attackers (Jabir et al., 2025). Blacklists, such as those, are not aware of phishing sites until they are
reported which leaves users exposed to zero-day attacks. Similarly, phishing methods change rapidly, which is
too fast to be matched by heuristic mechanisms. It is because of them that I feel that the necessity to have more
intelligent systems that can effectively identify phishing attacks real-time has been long overdue (Li et al., 2025).
Specifically, sophisticated intelligent detection methods informed by artificial intelligence and machine learning
technology are to be used to address this requirement. These systems examine the characteristics of websites,
user behavior and contextual indicators by using data-driven models to detect phishing attacks with extreme
precision and flexibility (Shahbazi et al., 2025). The smart systems will be dynamic and will adjust and evolve
faster than the phishing techniques and will offer the e-banking services with great security. This study explores
intelligent ways of guarding e-banking systems against phishing websites. It examines various algorithms and
evaluates the best performance and strategies that are likely to enhance security of online banking systems and
consequently curb cybercrime in the online banking environment.
Problem statement
The existence of e-banking platforms is subject to threats that have never been seen to consumer confidence and
financial institutions exposed to potentially harmful reputational effects due to the continued and constantly
developing sophistication of phishing attacks. Financial phishing attacks still present technological and human
weaknesses that criminals can use to access confidential financial data directly. Their social engineering attacks
employ AI and dynamically spoof sites. Ex use phishing technology and human weaknesses to steal confidential
financial information. Fraudulently making counterfeit websites or messages to masquerade as genuine banking
services with the view of stealing user passwords, account numbers and credit card details. Even though
automated phishing intercepting or stealing credentials is a simple affair, the sharing of fake websites or messages
are complex highly sympathetic phishing attacks, and as a result, the imitations of banking websites are
extremely difficult to execute. Specifically, these attacks are particularly efficient at evading the industry
dependence on blacklists and shorthand detection of threats. This issue would need balanced smart solutions that
can precisely and quickly locate and adjust to the detection of phishing attacks to safeguard e-banking
successfully.
LITERATURE REVIEW
Overview of phishing in the context of e-banking.
Phishing has become one of the most long-term and harmful threats to electronic banking (e-banking), which
erodes the security and trust upon which digital financial services rely. It can be described as a trick in which
cybercriminals masquerade as trusted institutions, frequently by using fake websites, emails, or messages to lure
users into sharing sensitive information like passwords, credit cards or personal identification information
(Nadeem et al., 2023). Phishing plays a significant role because it can use technological weaknesses and human
faithfulness as an effective means of attack in the setting of e-banking where remote and often unchecked
financial transactions are carried out (Pinjarkar et al., 2024). Phishing in online banking is also encouraged by
the growing digitalization of financial services and the convenience that customers demand online platforms to
provide. Attackers build counterfeit banking portals that look and feel like legitimate websites; send believable
emails containing harmful links or social engineering techniques that make it appear that there is a sense of
urgency and that an user must make a decision without hesitation (Yuspin et al., 2024). More sophisticated
methods, including spear phishing, pharming, malware-aided phishing, have further complicated such attacks
and can now be harder to be detected with traditional methods. In the case of financial institutions, the
ramifications of phishing are varied (Nadeem et al., 2023). In addition to direct financial losses, a successful
phishing attack undermines customer trust, destroys institutional reputations, and sets banks up against legal and
regulatory consequences (Nadeem et al., 2023). Instead, customers expose themselves to identity theft, financial
fraud and emotional trauma, which, in many cases, have long-term effects. Phishing, therefore, not only
interferes with individual users but also other more valuable attempts to create secure, inclusive, and resilient
digital banking systems. Conventional anti-phishing techniques including blacklists, heuristic-based systems,
and manual reporting are not very effective especially in dealing with zero-day phishing sites, which develop at
a high rate. These shortcomings indicate the urgency of smarter and more adaptive methods (Aldakheel et al.,
2023). By relying on artificial intelligence (AI), machine learning (ML) and a combination of both, intelligent
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 2282
www.rsisinternational.org
detection systems would be able to process large volumes of data, discovering previously undetected patterns
and detecting phishing attempts in real-time with even higher precision. Such strategies are becoming widely
regarded in the context of e-banking as a way to enhance the security, safeguard the users and maintain their
confidence in online financial systems.
Intelligent detection approaches
The intelligent detection methods are more sophisticated, dynamic strategies employing computational
intelligence, machine learning and artificial intelligence to detect and curb phishing attacks more efficiently than
conventional security mechanisms. In contrast to blacklist and heuristic-based systems that use set rules,
intelligent approaches are databased and thus they can identify new, emerging phishing patterns and act in real-
time. This flexibility is especially useful in the environment of e-banking, where stakes of phishing attacks are
great and threats are developing at an extremely fast pace.
1. Machine Learning (ML) Algorithms: The intelligent detection systems rely on the use of Machine
Learning (ML) Algorithms. The most popular supervised learning systems, which include Support
Vector Machines (SVM), Random Forests, Decision Trees, Logistic Regression, and k-Nearest
Neighbors (k-NN) are typically used to classify websites as either legitimate or phishing, based on
features extracted (Tufail et al., 2023). Such characteristics can be URL structure, HTML content, age of
the domain, details of the Jessica Stewart Lawrence (SSL) certificate and behavioral indicators. Such
ensemble techniques as Gradient Boosting and XGBoost provide an extra benefit to predictive accuracy
by incorporating the merits of a variety of classifiers.
2. Deep Learning Models such as Convolutional Neural Networks (CNNs) and Recurrent Neural Networks
(RNNs) are becoming more commonly used to detect phishing using non-linear relationships that are
complex and require learning large amounts of data. CNNs can efficiently analyze visual similarities
between phishing and legitimate sites whereas the RNNs can successfully extract sequential information
in the URL strings as well as user interactions (Alshingiti et al., 2023). These models have increased
detection precision, but this comes at the cost of increased datasets and more computation.
3. Hybrid Models integrate a variety of intelligent methods, combining machine learning, deep learning and
heuristic methods to enhance resilience and minimize false positives. As an example, hybrid systems can
apply ML algorithms at the first classification step and deep learning at the second step to guarantee
speed and accuracy (Ibrahim et al., 2025).
4. Natural Language Processing (NLP) is used to scan phishing emails and web content, detecting
suspicious patterns of language, misspellings, or irregularities in the style of communication. In addition,
adaptive learning models enable detection systems to keep up to date with the latest phishing tricks and
thereby maintain effectiveness in the long term (Saias, 2025).
METHODOLOGY
This paper adopts a comparative experimental research design that seeks to assess the effectiveness of various
intelligent detection algorithms in detecting phishing websites in the e-banking systems. Experimental approach
is the right approach because it enables systematic testing, benchmarking, and comparison of various models
under controlled conditions in order to establish their strengths and weaknesses vis-a-vis each other. Data set
collection entail publicly accessible data sets like the UCI Phishing Website Dataset that offer an annotated
instance of both legitimate and phishing websites. Available real e-banking data will also be used to add variety
to these datasets to improve validity of findings. To select the features, three groups of will be taken: URL-based
features (length, use of special characters, and depth of subdomain), content-based features (HTML tags, scripts,
and details of the certificate), and behavioral features (redirection, pop-ups, and response time). These
capabilities can record technical and contextual pointers to phishing attacks. The algorithms that are taken into
account are Support Vector Machines (SVM), Random Forests, Neural Networks, XGBoost, and Hybrid models
that are a combination of several classifiers. This choice represents a compromise of classical machine learning
and those of the state-of-the-art ensemble or deep learning approaches. Detection effectiveness will be measured
using performance measures like accuracy, precision, recall, F1-score, and Area under the Curve (AUC). Such
metrics give us a comprehensive picture of the model performance, and the correctness, as well as the robustness,
are both considered. The experiments are going to be run in Python as the main programming language; libraries
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 2283
www.rsisinternational.org
like Scikit-learn will be used to run classical ML models, TensorFlow to run deep learning models, and WEKA
to run comparative validation. The selected methodology is reasonable because it combines various datasets,
efficient feature engineering, and a range of smart algorithms, which offer a solid foundation to determine the
most efficient way to detect phishing in the e-banking setting.
RESULTS
Table 1: Performance metrics
Algorithm
Accuracy (%)
Precision (%)
Recall (%)
F1-Score (%)
AUC (%)
SVM
92.4
91.8
90.6
91.2
93.1
Random Forest
95.7
95.3
94.8
95.0
96.5
Neural Network
96.3
96.0
95.7
95.8
97.0
XGBoost
97.1
96.8
96.5
96.6
97.8
Hybrid Model
98.0
97.7
97.3
97.5
98.6
Table 1 indicates that all the algorithms were effective, although there are significant differences. SVM had the
lowest accuracy (92.4%), which suggests that it is not as effective when it comes to processing complicated
phishing characteristics. Random Forest and Neural Networks showed better outputs, with 95.7 and 96.3%
accuracy respectively, which shows their strength in pattern recognition. XGBoost performed better than these
models in 97.1 percent correctness and equal precision, recall, F1-scores. The Hybrid Model was the most
successful, with the highest accuracy (98.0 percent) and AUC (98.6 percent) showing that this type of classifier
is more adaptable and reliable in identifying phishing websites than single classifiers are.
Figure 1: Accuracy Comparison of Algorithms
The accuracy comparison graph shows that there is an apparent difference in performance of the algorithms.
SVM has the worst accuracy of approximately 92.3% implying the least usefulness when it comes to complex
phishing detection. Random Forest shows a significant improvement of almost 95.7% accuracy, which shows a
better classification ability. Neural Networks and XGBoost are more effective with 96.3 and 97.1 respectively,
which indicates that they are effective in pattern recognition. The Hybrid model is the most accurate at 98.0, as
it is a combination of all the others. Overall, the graph shows that there is a tendency according to which advanced
ensemble and hybrid methods will greatly improve the accurate detection in comparison to classical machine
learning methods.
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 2284
www.rsisinternational.org
Figure 2: Performance Comparison of Algorithms
The figure 2 shows comparisons of Precision, Recall and F1-Score among five algorithms. SVM has been the
poorest performer and its precision, recall and F1-Score have been mostly 91-92, indicating a weakness in
phishing detection. Random Forest performs moderately at 95% which is reliable but a little lower than more
advanced models. Both of the models are more stable, with 96 and 97 percent of each, indicating balanced
predictability. All models are almost identical to each other with a score of around 97.5-97.7% with all metrics,
which demonstrates its strength and consistency. As a whole, the trend suggests that ensemble and hybrid models
are superior in detection efficiency compared to traditional models although they offer a baseline.
Figure 3: AUC Comparison of Algorithms
The AUC comparison plot indicates that algorithms have a discriminative effect in phishing detection. SVM has
the lowest AUC of around 93 rounding off to show a poor performance in distinguishing between phishing and
legitimate emails. Random Forest improves it 96.5 and Neural Networks to 97%. XGBoost demonstrates high
performance of around 97.8 per cent, which indicates that it learns complex patterns better. The Hybrid model
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 2285
www.rsisinternational.org
outperforms the rest with AUC of about 98.5, which proves to be very effective in classifying and with a small
overlap between false positive and false negative. Overall, ensemble and hybrid models have always been shown
to be better in detection accuracy compared with traditional models.
Comparative analysis of algorithms’ strengths and weaknesses.
Algorithm
Strengths
Weaknesses
SVM
Effective with small to medium datasets; strong in
high-dimensional spaces; good generalization.
Struggles with large datasets; sensitive
to kernel choice; lower scalability.
Random Forest
Handles large datasets well; resistant to overfitting;
interpretable feature importance.
Computationally expensive with many
trees; less effective for highly
imbalanced data.
Neural Network
Strong in capturing complex, non-linear patterns;
adaptable to large datasets; high predictive power.
Requires large training data; prone to
overfitting; high computational cost.
XGBoost
High accuracy; efficient handling of missing
values; fast training; robust against overfitting.
Requires careful parameter tuning; less
interpretable than simpler models.
Hybrid Model
Combines strengths of multiple algorithms;
maximizes accuracy, recall, and adaptability;
superior generalization.
Increased complexity; higher training
and computational costs; harder to
implement in real-time systems.
Table 2 shows that the different algorithms have unique strengths and weaknesses in detecting phishing. SVM
scales well with small data but does not scale and whereas Random Forest is interpretable and robust but
computationally expensive. Neural Networks are very effective in the modeling of complex patterns but are
resource-intensive and require huge data. XGBoost is highly accurate and efficient, but needs a fine tune. The
Hybrid Model is superior to others combining the strengths, superior accuracy, and flexibility, but it is more
complex and expensive. In general, advanced models have superior performance but the practical
implementation of e-banking should compromise between accuracy, scalability and computational efficiency.
DISCUSSION
The results of the study offer a good understanding of how intelligent detection models can be used to fight
phishing attacks on e-banking websites. Phishing is one of the most common and harmful types of cyberattacks
that use human vulnerability and system flaws to damage sensitive financial information. Comparative analysis
of machine learning algorithms (SVM, Random Forest, Neural Networks, XGBoost, and Hybrid models)
demonstrate a definite hierarchy of the detection performance. Though SVM can only provide the simplest form
of protection since its accuracy and AUC scores are lower, ensemble-based models like the Random Forest and
Neural Networks are more resilient. However, XGBoost and Hybrid models perform better than others perform,
as they are recall that is more precise, better and have higher F1-scores, which means that they are more capable
of the balance between false positives and false negatives. These findings affirm that better detection is provided
by sophisticated ensemble and hybrid techniques that are more consistent and reliable in detecting phishing and
can effectively adapt to the changing strategies of attackers. The paper identifies hybrid intelligent systems as
the best modes of protection of e-banking environments. They are scalable and flexible, which makes them
applicable in the real-life context, where phishing strategies are changing fast. In addition to an algorithmic
performance, the findings also shed light on the consequences of cybersecurity management in general. These
advanced detection systems applied to e-banking platforms can help tremendously lower successful phishing
attacks, safeguard consumer information and increase digital trust, which is a key element to user adoption and
continued use of online banking. Moreover, high-quality detection will decrease the operational and financial
risk to banks, decreasing the costs of recovery and restitution and boosting adherence to regulatory cybersecurity
practices. Cost sensitive learning needs to be explored in the future to reduce the risks of false detection that can
still be a major operational issue. In addition, adversarial resilience would be improved to make the system more
robust in the face of changing and misleading phishing attacks. An implementation of federated learning and
blockchain would also help additional protection of data privacy and integrity, and facilitate secure collaborative
training without centralized exposure of sensitive banking data. Lastly, the incorporation of user-oriented
assessments (perceived trust, usability, customer satisfaction) in conjunction with technical performance
indicators would provide a better comprehensive picture of the effects of phishing prevention on digital banking
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 2286
www.rsisinternational.org
ecosystems. On balance, it is possible to state that intelligent phishing detection is not only a technological dump
but also a corporate necessity that will underpin proactive, adaptive, and customer-centric cybersecurity
approaches to the sustainable development of digital banking.
CONCLUSION
This paper analyzed intelligent methods of protecting e-banking platforms against phishing sites, and the paper
has particularly compared the performance of the algorithms in various measures. The key conclusions
demonstrate that phishing is a significant threat to cybersecurity of digital banking, and it is necessary to use
sophisticated detection tools to protect valuable financial data. The findings revealed that Support Vector
Machines (SVM) offered moderate performance in detection and thus not suitable in the complex phishing
patterns. Random Forest and Neural Networks have better results and they are more accurate and reliable.
XGBoost showed high detection power, in terms of efficiency and scalability. However, most importantly,
Hybrid models were consistently able to achieve better results than all the other metrics, which include accuracy,
precision, recall, F1-score, and AUC, which means that the combination of a set of algorithms provides a greater
degree of robustness and flexibility to adapt to changing phishing methods. The result of these findings is that
smart methods of detection, especially the ensemble and the hybrid-based systems, are very effective to deal
with the phishing threats in e-banking. They can trade false positives and false negatives and therefore are viable
in real world banking applications where reliability and user confidence are of the essence. With sophisticated
machine learning methods, banks will be in a position to attain proactive defenses against phishing, and, thus,
diminish risk of fraud, decrease cost, and strengthen consumer trust in online services. Based on these
observations, a number of recommendations can be given. In a bid to be fully protected against phishing, banks
ought to invest in the deployment of intelligent hybrid detection systems in their security systems. Regulatory
frameworks need to be designed to promote the use of AI-driven cybersecurity, and policymakers need to ensure
that the requirements of detecting data are standardized in all financial institutions to promote the wider digital
economy. It is recommended that software developers should focus on adaptive and scalable models capable of
learning in real time the new emerging patterns of phishing behaviors to be resistant to the new cyber threats.
Cooperation among regulators, developers and banks will play a fundamental role in enhancing a safe and
reliable digital banking experience. In the future, it is possible to focus on a variety of directions to enhance
phishing detection. To begin with, real time detection features must be highlighted to stop the phishing attempts
prior to their accomplishment. Second, generalizability of models among global banking platforms can be
enhanced by using bigger and more varied datasets. Finally, yet importantly, adaptive learning models evolving
with phishing tricks will play a significant role in the long-term security. The combination of technological and
behavioral defenses in terms of combining intelligent detection with user education should also be explored by
research. Intelligent detection strategies are a revolutionary step in the security of e-banking. Through adoption
of hybrid models and adaptive approaches, banks will be able to build resilient systems that help not only
safeguard financial resources but also place digital banking on the road to the future, in a world of growing cyber
threats.
REFERENCE
1. Aldakheel, E. A., Zakariah, M., Gashgari, G. A., Almarshad, F. A., & Alzahrani, A. I. A. (2023). A Deep
Learning-Based Innovative Technique for Phishing Detection in Modern Security with Uniform
Resource Locators. Sensors, 23(9), 4403. https://doi.org/10.3390/s23094403
2. Alshingiti, Z., Alaqel, R., Al-Muhtadi, J., Haq, Q. E. U., Saleem, K., & Faheem, M. H. (2023). A Deep
Learning-Based Phishing Detection System Using CNN, LSTM, and LSTM-CNN. Electronics, 12(1),
232. https://doi.org/10.3390/electronics12010232
3. Ibrahim, N., Rajalakshmi, N.R., Sivakumar, V. et al. (2025). An optimized hybrid ensemble machine
learning model combining multiple classifiers for detecting advanced persistent threats in networks. J
Big Data 12, 212. https://doi.org/10.1186/s40537-025-01272-w
4. Jabir, R., Le, J., & Nguyen, C. (2025). Phishing Attacks in the Age of Generative Artificial Intelligence:
A Systematic Review of Human Factors. AI, 6(8), 174. https://doi.org/10.3390/ai6080174.
5. Kumari, A., & Nagarjan, C. (2022). The Impact of FinTech and Blockchain Technologies on Banking
and Financial Services. Technology Innovation Management Review, 12, 753–767.
https://doi.org/10.22215/timreview/1481.
INTERNATIONAL JOURNAL OF RESEARCH AND INNOVATION IN SOCIAL SCIENCE (IJRISS)
ISSN No. 2454-6186 | DOI: 10.47772/IJRISS | Volume IX Issue X October 2025
Page 2287
www.rsisinternational.org
6. Li, W., Manickam, S., Chong, Y.w., Leng, W., & Nanda, P. (2024). A State-of-the-Art Review on Phishing
Website Detection Techniques. IEEE Access, 12(1), 1–21.
https://doi.org/10.1109/ACCESS.2024.3514972.
7. Money, U., & Iyoha, A. (2025). ELECTRONIC BANKING CHANNELS AND FINANCIAL
PERFORMANCE IN THE NIGERIAN BANKING INDUSTRY. Journal of Accounting, Finance and
Risk Management, 8(1), 193–209. https://doi.org/10.61143/umyu-jafr.8(1)2025.013.
8. Nadeem, M., Zahra, S., Abbasi, M.N., Arshad, A., Riaz, S., & Ahmed, W. (2023). Phishing Attack, Its
Detections and Prevention Techniques. International Journal of Wireless Information Networks, 12(1),
13–25. https://doi.org/10.37591/IJWSN.
9. Pinjarkar, L., Hete, P., Mattada, M., Nejakar, S., Agrawal, P., & Kaur, G. (2024). An Examination of
Prevalent Online Scams: Phishing Attacks, Banking Frauds, and E-Commerce Deceptions. In
Proceedings of the 6th International Conference on Advanced Information Technology (ICAIT), 1–6.
https://doi.org/10.1109/ICAIT61638.2024.10690377.
10. Saias, J. (2025). Advances in NLP Techniques for Detection of Message-Based Threats in Digital
Platforms: A Systematic Review. Electronics, 14(13), 2551. https://doi.org/10.3390/electronics14132551
11. Shahbazi, Z., Jalali, R., & Molaeevand, M. (2025). AI-Based Phishing Detection and Student
Cybersecurity Awareness in the Digital Age. Big Data and Cognitive Computing, 9(8), 210.
https://doi.org/10.3390/bdcc9080210.
