









































Pa
ge

 
1



Pa
ge

 
17

7

American Journal of  Financial 
Technology and Innovation (AJFTI)

AI-Driven Fraud Detection in Digital Banking: Ml Approach for Secure and Transparent 
Financial Transactions

Oreoluwa Abimbola Serifat1*, Roseline C. Igah2, Kehinde M Balogun3, Gershom Randy Mensah4, Emmanuel Niiboye Odai4

Volume 3 Issue 1, Year 2025
ISSN: 2996-0975 (Online)

DOI: https://doi.org/10.54536/ajfti.v3i1.5168
https://journals.e-palli.com/home/index.php/ajfti

Article Information ABSTRACT

Received: August 25, 2025

Accepted: September 23, 2025

Published: October 25, 2025

The convenience of  digital banking services has transformed the global financial industry 
and is now available to consumers all over the world. As with any advancement, there’s 
an increase in associated risk. In this case, we have an upsurge in fraudulent activity, the 
mobility of  cybercriminals, and their more advanced technologies to breach vulnerabilities 
within digital infrastructures. Indeed, financial crimes are constantly evolving like the 
rest of  technology and society. Those who monitor and manually analyse systems are no 
match for the speed at which criminals can devise new rule-of-thumb schemes. This article 
examines how artificial intelligence and machine learning can reform the detection of  fraud 
within digital banking systems. The research analyses different techniques of  AI and ML, 
supervised learning, unsupervised learning, ensemble, and deep learning approaches, while 
also observing their uses in practical fraud detection systems. The paper also analyses the 
ethical and legal concerns involving the use of  AI within banking, considering data issues, 
algorithmic discrimination, and other contentious aspects of  legal compliance, including 
quasi-legal frameworks like GDPR and PCI-DSS. It also explores some of  the newer 
directions in AI, like quantum computing, explainable AI (XAI), and federated learning, 
and their potential implications to improving fraud detection systems performance. Finally, 
the focus of  this paper has been on a continuing effort and partnership across sectors in 
building resilient, secure, transparent financial systems. AI, ML, and blockchain technologies 
enhance the capability to prevent fraud in digital banking, while ensuring and maintaining 
customer trust and security in financial transactions.

Keywords
Artificial Intelligence (AI), Digital 
Banking, Fraud Detection, 
Machine Learning (ML)

INTRODUCTION
Digital banking represents a fundamental change in the 
finance by transforming the ways that people access and 
handle their money, for example (Mohmmed et al., 2024). 
Digital banking, through online and mobile banking, or 
use of  digital or e- wallets, has increased access to financial 
services by making them more accessible, convenient, and 
efficient (Barroso & Laborda, 2022; Oduro et al., 2025). 
Consumers can now execute various banking transactions 
from any location at any moment which leads to enhanced 
financial inclusion and equal access to banking services. 
Digital banking solutions enable businesses to enhance 
operational processes while cutting operational expenses 
and offering customized services to their clients (Bueno 
et al., 2024).
The expansion of  the digital landscape creates greater 
opportunities for financial fraud to occur. Online 
transaction growth leads to a highly susceptible 
financial environment for numerous criminal activities 
(Kipngetich, 2025). Fraudsters exploit digital platforms 
to target financial system vulnerabilities using advanced 
techniques to overcome conventional security barriers. 
As digital platforms become essential for daily banking 
operations, there has been a substantial rise in fraudulent 
activities including identity theft, card-not-present fraud, 
account takeovers and money laundering (Adeyeri et al., 

2023). Traditional fraud detection methods that depend 
on rule-based algorithms and human supervision fail to 
match the speed and complexity of  current fraudulent 
activities (Ismaeil, 2024).
As the fraud increases, the need for automated, 
innovative and real-time detection solutions is evident. 
Banking is realizing that traditional fraud detection has 
its weaknesses; high rates of  false positives, expensive 
manual review process, and lack of  ability to expose new, 
emerging fraud patterns. This realization has given rise 
to a more nuanced application of  techniques, especially 
utilizing Artificial Intelligence and Machine Learning 
(AI/ML) (Olowu et al., 2024). The need to adapt to new 
schemes and the capability of  AI technologies in large 
data analysis, pattern recognition, and adaptation to 
new threats make it an increasingly important resource 
for improving the fraud detection systems in use. By 
leveraging these technologies, banks can prevent and 
minimize fraud before it occurs, enabling secure, efficient 
and transparent financial transactions (Ismaeil, 2024). 
This evolution towards fraud detection via AI, addresses 
the need for better intelligence, scalability, and real time 
capabilities in a more digital and online oriented industry. 
What is certain is that as the digital banking system has 
brought much benefit, it needs to be secured against new 
digital fraud threats, and thus protected just as much.

1 Department of  Data Analytics, Nexford University, USA
2 Department of  Management Science And Information Systems, Oklahoma State University, USA
3 Department of  Mathematics, Austin Peay State University, Tennessee
4 Northeastern University, Massachusetts, USA
* Corresponding author’s e-mail: oreoluwaabimbolaserifat@gmail.com



Pa
ge

 
17

8

https://journals.e-palli.com/home/index.php/ajfti

Am. J. Financ. Technol. Innov. 3(1) 177-187, 2025

Hence the objective of  this paper is to discuss how AI and 
ML could be fruitfully employed to improve the detection 
of  fraud in digital banking as a novel and dynamic way to 
maintain security, efficiency, and transparency in a world 
where fraud in digital banking is becoming increasingly 
prevalent. The types of  fraud in digital banking, AI and 
ML solutions to detect and deter fraud, practical examples 
of  successful implementations that enhance the security 
and robustness of  digital banking against ever-changing 
fraud techniques, and bolstering consumer confidence in 
banking transactions.

LITERATURE REVIEW 
Digital banking has been a disruptive innovative force in 
financial services, changing how people and businesses 
interact with their finances in a big way (Bueno et al., 
2024). It is technology, internet access, and demand 
for reduced and efficient banking services that have 
motivated the transformation from physical to digital 
banking (Iwedi, 2024). For instance, with digital banking 
customers can check their account balances, transfer 
money, request loans, and pay bills online in their houses 
or anywhere by using their mobile phones (Javaid et 
al., 2022). Also, financial institutions have strategically 
leveraged this technology to provide a spectrum of  
e-payment opportunities and services to their clientele 
(Igah & Luse, 2024).
Among the essential services of  digital banking are online 
payments, which are transactions that help the transfer 
of  money between individuals and businesses through 
the internet (Abdelrhman, 2025; Windasari et al., 2022). 
The result has been that e-commerce has exploded as 
consumers now can purchase goods and services easily. 
Mobile banking applications take this convenience a 
step further, allowing individuals to view their accounts 
and execute transactions from their mobile phones and 
thus transforming banking into an even more accessible 
medium than before (Ezie et al., 2023; Rahman et al., 
2024). Wallet applications like Apple Pay and Google Pay 
or apps from individual banks help consumers save and 
manage their payments information digitally, enabling 
fast and secure transactions without requiring physical 
cards (Khando et al., 2022).
While digital banking has offered many advantages 
including financial inclusion for the unbanked, user-
friendliness, cost- efficiency, etc, it has also increased 
the attack surface available to cybercriminals. In other 
words, the advantages of  digital banking have also been 
problematic, as cybercriminals have taken advantage of  
online platform weaknesses to obtain personal information 
and carry out unauthorized illegal transactions, and 
therefore online banking has become subject to fraud 
and cybercrime activities (Aziz & Andriansyah, 2023; 
Mallesha & Hymavathi, 2024; Roszkowska, 2021).As a 
result, the financial sector faces increasing pressure to 
implement robust security measures to protect customers 
from the growing risk of  fraud.

Fraud in Digital Banking
Digital banking fraud is the unauthorized and illegal 
utilization of  digital banking systems to steal, manipulate 
and compromise financial data for an individual’s own 
benefit. As a result, more and more fraudsters are taking 
advantage of  online banking, utilizing multiple methods 
of  exploiting weaknesses inherent in digital banking 
systems. These include, among the more common, 
identity theft (Venigandla & Vemuri, 2022), where 
fraudsters obtain personal data to pose as clients and gain 
access to their accounts.
Phishing is another form of  scam in which customers 
are tricked into revealing sensitive information such 
as passwords and account numbers (Adaji et al., 2024). 
The second most frequent form of  fraud is transaction 
manipulation, where the fraudster is actually the one 
who initiates the transaction exploiting a loophole in the 
payment mechanisms (Adeyeri, 2024; Oduro et al., 2025). 
Another related issue of  grave concern is also money 
laundering, where criminals can disguise the source of  
illegal funds through sophisticated transactions on digital 
banking platforms that seem authentic (Bello & Olufemi, 
2024; Olowu et al., 2024; Khodabandehlou et al., 2024). 
Also, without adequate security measures, funds can 
be moved illegally across borders using digital payment 
systems, making it even more difficult to get any form of  
control or trace these funds.
Consumer trust in the banking system is thus diminished 
as these scams are expensive for banks and other financial 
institutions to deal with (Adeyeri, 2024). The increased 
use of  digital banking has made the implementation of  
advanced fraud detection systems capable of  addressing 
and recognizing such fraudulent behaviours a necessity.

Traditional Fraud Detection Methods 
Most banks use rule-based systems and human 
monitoring to catch fraud when it happens. In rule-based 
systems a predefined set of  rules is utilized to identify 
potentially suspicious behaviour, for example, these sorts 
of  rules could indicate abnormally large transactions 
or simultaneous multiple withdrawals from various 
locations. Goyal et al., 2025; Metha, 2025). These systems 
have the ability to detect patterns of  fraud that have 
been established, but they are limited in that they cannot 
evolve to detect emerging patterns of  fraud. Traditional 
rule systems tend to be rigid; they can only catch fraud 
patterns that are in line with existing parameters and 
are unsuccessful at recognizing new fraud strategies 
that might be unpredicted patterns of  fraud identified 
by existing parameters and would be ineffective in 
recognizing new fraud strategies that might not have been 
predictive of  fraud patterns by existing parameters and 
would be ineffective at recognizing new fraud strategies 
that might not have been anticipated (Bello et al., 2023; 
Ikemefuna et al., 2024).
Fraud monitoring is also completed by humans, who 
review flagged transactions that were detected by rules 
or other non-automatic fingerprints of  alleged fraudulent 



Pa
ge

 
17

9

https://journals.e-palli.com/home/index.php/ajfti

Am. J. Financ. Technol. Innov. 3(1) 177-187, 2025

activity (Hilal et al., 2021). But human oversight adds an 
even more rigid layer of  control, as it is highly laborious 
and also subject to mistakes given the enormous number 
of  daily transactions. Plus, humans reviewing cases can be 
overwhelmed by the volume and complexity of  rampant 
fraudulent activity, causing delays in the detection and 
subsequent action (Bello et al., 2022).
Combined, these traditional techniques can produce high 
false positive rates which burden banking staff  and lead to 
unneeded investigations. Just as fraudsters have continued 
to enhance their techniques, the conventional systems have 
become less effective to deal with the increasing complexity 
of  this digital fraud and this has opened the door to more 
sophisticated, AI-based solutions for fraud detection.

Machine Learning in Fraud Detection 
Fraud detection has changed with AI and ML as they 
help financial institutions, “analyse huge data sets in 
real-time and recognize intricate patterns that signal 
the possibility of  fraudulent behaviour” (Adhikari et al., 
2024; Odufisan et al., 2025). Unlike traditional rule-based 
systems, ML algorithms are trained on historical data and 
thus are adaptive in detecting patterns of  both historical 
and emerging fraud. Supervised learning, unsupervised 
learning, and reinforcement learning are now commonly 
used in fraud detection systems by machine learning 
techniques (Sarker, 2021; Hernandez Aros et al., 2024).
Supervised learning involves training a model using a 
labelled dataset with transactions already labelled as 
fraudulent or legitimate (Afriyie et al., 2023). The model 
will then learn the patterns that distinguish these two 
in order to predict the probability of  fraud in future 
transactions. Unlike, unsupervised learning is used for 
finding anomalies in unlabelled data (Venigandla & Vemuri, 
2022). This is particularly advantageous in identifying 
novel or unknown patterns of  fraud that are actually 
not found in the historical data. Also, reinforcement 
learning, where the model learns by interacting with the 
environment, becomes an effective fraud detection tool 
as it constantly improves the predictions of  fraudulent 

actions (Sharma, 2024).
AI and ML applications to fraud detection writ large has 
been the subject of  study in a couple of  papers. In 2020 Li 
et al. concluded that individual algorithms performed worse 
than random forests and gradient boosting in the detection 
of  credit card fraud, supporting the results of  this study. 
Another example of  research emphasizing the use of  RPA 
and AI in online banking fraud detection is by Wang et al. 
(2019), who recommended hybrid RPA and AI predictive 
analytics system to be used as a cutting-edge approach for 
online banking fraud detection. Also, Oduro et, al. (2025) 
pointed out the potential of  machine learning models to 
improve accuracy in fraud detection and decrease the false 
positive rate in digital banking systems.

MATERIALS AND METHODS
The study is based on secondary data concerning the use 
of  Artificial Intelligence and Machine Learning in Fraud 
Detection in Digital Banking from among a general 
pool of  published research articles. The required data 
was obtained by systematically identifying peer reviewed 
journal articles, conference proceedings, and industry 
reports retrieved through Google Scholar, IEEE Xplore, 
Elsevier, and other reputable journals. Selected articles 
were those that addressed digital fraud detection systems 
that implemented the use of  AI and ML, specifically in 
regards to the categories of  fraud detection systems of  
supervised learning, unsupervised learning, reinforcement 
learning, and ensemble methods. A method and finding 
section were extracted from each study of  interest that 
focused on fraud detection in banking. A focus was on 
these techniques as employed in actual banking scenarios, 
specifically data preprocessing, model selection, and 
algorithm performance. 

RESULTS & DISCUSSION
AI And ML Techniques for Fraud Detection 
AI and ML thus represent an integral part of  fraud 
detection systems, particularly in the context of  online 
banking. These technologies allow banks to sift through 

Figure 1: Machine Learning (ML) in Fraud Detection



Pa
ge

 
18

0

https://journals.e-palli.com/home/index.php/ajfti

Am. J. Financ. Technol. Innov. 3(1) 177-187, 2025

massive volumes of  transaction data and detect minute 
patterns in transactions that may be suggestive of  fraud. 
While legacy fraud detection is primarily rule-based 
and human supervised, AI and ML algorithms apply 
knowledge to new data in real-time to adapt to shifting 
patterns of  fraud.
AI and ML’s main advantage in fraud detection is its capacity 
to handle, or “scale up”, large datasets and find patterns 
within them (Adeyeri, 2024). By detecting both established 
as well as new patterns of  fraud, these algorithms can aid 
in the prevention of  fraud before it occurs. AI technologies 
can identify patterns of  anomalies that would be hard to 
catch by human analysts or static rules-based systems. 
Rather, through their ability to use historical data and 
becoming better at predicting fraud over time, AI and ML 
“empower banks to detect new forms of  fraud even before 
they happen” (Odufisan, 2025).
As the advancement of  fraud schemes have also advanced 
detection methods are becoming obsolete. This ability to 
scale and adapt fraud is an important characteristic of  AI 
and ML models. Overall, these technologies help banks 
not only to identify fraud as it is happening but also to 
anticipate and stop fraud before it occurs, increasing the 
security of  the digital banking experience.

Common ML Techniques 
Supervised Learning
Afriyie et al., 2023 highlights that supervised learning is one 
of  the most commonly employed methods in machine 
learning to detect fraud. This includes training a model on a 
labelled dataset in which there is knowledge of  the outcome 
of  each transaction whether it was fraudulent or legitimate. 
The model then “learns” to correlate patterns or groupings 
of  input features, transaction amount, time, location, etc. 
with the target outcome fraud versus non-fraud.
Fraud detection often employs “Decision Trees” and 
“Random Forests” which are common fraud detection 
supervised learning algorithms (Salunke et al., 2025). It 
forms a model shaped like a tree by repeatedly partitioning 
the data based on values of  the features. The nodes are 
points at which a decision is made on a given feature and 
the leaves are the predicted classification as either fraud 
or non-fraud (Adeyeri, 2024; Afriyie et al., 2023; Johora, 
2024). Random forests are collections of  decision trees. 
They aggregate multiple decision trees to increase accuracy 
and overfitting, which is a flaw of  decision trees. Since it 
combines predictions from multiple trees, random forests 
are efficient tools, especially when dealing with large and 
feature-rich data (Ismaeil, 2024). Li et al. (2020) On top of  
that offered, as an example, logistic regression, decision 
trees and neural networks; who studied the efficiency 
of  different machine learning methodologies, including 
logistic regression, decision trees and neural networks, 
in detecting credit card fraud. With regard to accuracy 
and detection rate, the research found that ensemble 
methods, such as random forests and gradient boosting, 
outperformed single algorithms.
Logistic Regression is a well-known algorithm that 

applies to supervised learning approaches for the 
detection of  fraud (Adeyeri, 2024). Specifically, “it is a 
statistical method that “models the probability of  a 
binary dependent variable, for example, presence versus 
absence or fraud versus no fraud, as a function of  one or 
more independent variables”. Logistic regression can also 
be an efficient and interpretable approach when there is a 
linear relationship between the features and the outcome. 
It is also useful in interpreting the importance of  various 
features in the prediction of  fraudulent transactions 
(Venigandla & Vemuri, 2022).

Unsupervised Learning
This is particularly relevant in the context of  fraud 
detection in which obtaining label data is often hard but 
worse still it is impossible to obtain. On the other hand, 
supervised learning does not require labelled data to train 
the model. Instead, it seeks patterns and outliers in the 
data that do not adhere to normality (Khodabandehlou 
et al., 2023).
K-means and more generally “Clustering methods” 
are frequently included among the techniques used 
for unsupervised learning within the specific domain 
of  fraud detection (Huang et al., 2024). Clustering 
categorizes similar data points based on characteristics. 
For example, K-means partitions the data into K clusters 
of  like transactions. Most of  which are “anomalous” or 
“fraudulent” transactions. This would assist in detecting 
outliers in the data points which are often indications of  
a possibility of  fraud (Ali et al., 2021).
Another common approach used for fraud detection 
is the ‘Anomaly detection’ algorithms, Unsupervised 
(Venigandla & Vemuri, 2022; Rojan, 2024). These are 
algorithms based on rare or anomalous patterns within 
the data which are unlike normal behaviour. Anomaly 
detection has been done by means of  “Isolation Forests” 
and “One-Class SVM” (Support Vector Machines) 
amongst other methods. (Wei et al., 2023). These models 
learn what the normal behaviour of  the dataset is and 
treat anything that deviates from it as suspicious. In 
particular, anomaly detection could be very useful to 
detect a novel pattern for fraud that has never been 
previously experienced, and thus becomes an important 
technique to deploy against new fraud tactics.

Deep Learning
“Deep learning” is, in contrast to other traditional ML 
models, a more powerful approach to searching on 
massive datasets for more complex patterns of  fraud, as 
it uses artificial neural networks containing multiple layers 
to model very complex relations in the data instead of  
traditional ML models (Xuan et al., 2021; Bello & Olufemi, 
2024). On top of  that, large amounts of  unstructured 
data like images, or transaction texts/narratives, which 
are harder to deal with by classical models, are also easily 
accommodated by these models.
Two examples of  deep learning models that are used 
effectively for fraud detection are Convolutional Neural 



Pa
ge

 
18

1

https://journals.e-palli.com/home/index.php/ajfti

Am. J. Financ. Technol. Innov. 3(1) 177-187, 2025

Networks (CNNs) and Recurrent Neural Networks 
(CNNs) (Adeyeri, 2024). Convolutional neural networks, 
typically employed for image recognition, have the 
potential to be utilized for fraud detection as they excel 
at detecting patterns in an input of  sequences or time 
series data, such as sequences of  transactions (Bello et al. 
2022; Bello & Olufemi 2024; Chowdhury 2024). RNN’s, 
and specifically “Long Short-Term Memory (LSTM)” 
networks, are focused on spotting temporal dependencies 
within the data and therefore have good potential to be 
utilized for fraud detection in cases where the sequence 
of  the events in relevant (Bhuiyan et al., 2025; Mienye et 
al., 2024; Muthunambu et al., 2024). An example could 
be that an RNN can recognize fraud by the transactional 
pattern, or by the attempted login pattern recognition.
Deep learning methods have the advantage of  being 
able to detect more complex patterns of  fraud than 
what simpler machine learning methods would be able 
to detect. These types of  models are bolstered when 
exposed to more data and so are particularly suited to be 
used in environments such as digital banking where also 
the fraudsters are constantly evolving their methods. 

Ensemble Methods
The “Ensemble methods” combines the predictions of  
multiple models to improve accuracy and robustness 
(Olowu et al., 2024). The theory behind ensemble methods 
is that by ensembling many models that each have some 
strengths and weaknesses we can form a stronger, more 
accurate model. “Boosting”, which focuses on those 
training instances that are harder to classify by changing 
their weights iteratively, is one of  the most popular 
ensemble methods. “Gradient Boosting” and “XGBoost” 
are widely used boosting techniques which have proven 
effective when used for fraud detection tasks (Ganaie et 
al., 2022; Khan et al., 2023).
“Bagging” or Bootstrap Aggregating is another form 
of  ensemble method as described by Hernandez t al., 
2022; Vens, 2013. Bagging takes multiple models (usually 
decision trees) trained on multiples subsets of  data 
and averages they predictions. It is also a mechanism 
to reduce variance and avoid overfitting, especially for 
decision trees. An example of  bagging is the “Random 
Forest” algorithm, which aggregates many decision trees 
for better predictions. Detecting fraud using integrated 
systems based on NLP, anomaly detection, and supervised 
learning offers 15-25% higher detection levels than using 
any of  the three types of  models individually, as shown in 
a sample study of  25 large financial institutions (Olowu 
et al., 2024).
Ensemble methods have special applicability for fraud 
detection, since the trade-off  between false positives and 
false negatives in that context is very important. 

Data Preprocessing 
High quality data is a requisite condition for effective 
performance of  machine learning models. The first and 
most important step when creating any fraud detection 

model, is data preprocessing; this step ensures that the 
data is cleaner and structured for ease of  use.
Data cleaning typically deals with issues of  missing, 
duplicate, or inconsistent data. Records may be incomplete 
for technological reasons, such as system errors, or 
human reasons, such as input error. Such gaps must be 
filled because they are sources of  bias or inaccuracy in 
predictions. Common methods include imputation by 
filling values with mean, median, or mode, or excluding 
records with missing values (Alam et al., 2023).
This is often called “feature extraction” the process of  
determining which variables (or features) found in the 
data are useful in order to improve the accuracy of  the 
model. For instance, the features of  interest in the case of  
fraud might be the transaction amount, time, location and 
number of  transactions. Identifying these characteristics 
in the input data is useful to provide lower dimensionality 
to the dataset and to allow the model focus on the 
important variables to make the prediction (Cherif  et al., 
2022; Islam et al., 2025).
These are known as “feature transformation” techniques 
that can also be used to normalize or standardize the 
features so that all variables are treated equally in the 
model. For example, the learning from data which 
contains large numerical values might be biased if  they 
are not normalized. Two common ways of  transforming 
data are known as “z-score normalization” and “min-max 
scaling”, which normalize the data to a common range or 
distribution (Bello et al., 2024).
Effective preprocessing is important for the training of  
accurate fraud detection models. Every properly cleaned 
and well-engineered data can significantly improve the 
performance of  machine learning algorithms, which can 
in turn lead to more reliable fraud detection.

Model Evaluation and Metrics 
Model evaluation is a crucial step in fraud detection 
modelling to ascertain that the models used are 
performing effectively and can be relied on. Several 
performance metrics are analysed in order to assess the 
model’s capability of  being not only effective at detecting 
fraud, but also being able to minimize false positives, as 
well as minimizing frauds that are missed.
A common metric to apply is “accuracy”, which may not be 
a sufficient measure in the case of  fraud detection as there 
is an imbalanced class problem in which the fraudulent 
transactions are a very small count as compared to the 
legitimate ones (Tejesh et al., 2025). Instead “precision” and 
“recall” are usually more useful. Precision is the percentage 
of  actual positive cases of  fraud that were predicted to be 
fraud out of  all cases that were predicted as fraud, and recall 
means the percentage of  actual positive cases of  fraud that 
were correctly predicted by the model (Dangsawang & 
Nuchitprasitchai, 2024).
The “F1- score”, being the harmonic mean of  precision 
and recall provides a single measure that is used to balance 
both and is particularly useful in cases of  class imbalance. 
Receiver Operating Characteristic - Area Under the Curve 



Pa
ge

 
18

2

https://journals.e-palli.com/home/index.php/ajfti

Am. J. Financ. Technol. Innov. 3(1) 177-187, 2025

(ROC-AUC) analysis is another commonly used metric, 
which examines the true-positive versus false-positive 
ratio across various thresholds (Trucco et al., 2019).
These measures help account for the fact that there is a 
trade-off  between detecting fraud and annoying a non-
fraudulent customer. These metrics are hence optimized 
since, in order to provide useful and accurate outputs, AI-
based fraud detection tools need to be able to detect fraud 
in real time, while not impinging unduly on legitimate users.

Applications of  Ai-Driven Fraud Detection In 
Digital Banking 
In the last few years, a number of  banks and financial 
institutions have adopted AI- ML systems successfully 
into their fraud detection processes and have seen major 
benefits in their fight against fraud. This is especially 
clear in the case of  “AI-based credit card fraud detection 
systems”. For example, big banking companies like 
American Express and Citibank already implement 
real-time AI solutions that mine massive datasets of  
transactions by identifying patterns within this data 
(Mejia, 2019; Owen, 2021). The systems flag these unusual 
behaviours, such as when international transactions of  
high dollar amount suddenly increase or spending habits 
shift rapidly, as potentially fraudulent. These AI models 
keep learning from every new transaction they analyse, 
thus getting better over time in identifying new types of  
fraudulent schemes and minimizing false positives.
The other is “anti-money laundering (AML) systems”, 
which also have been successful in deploying AI to detect 
fraud. These days, banks and financial institutions apply 
AI to detect suspicious transactions connected with 
money laundering processes (Oyedokun et al., 2024). 
For instance, systems can identify transaction networks 
in patterns typical of  money laundering, such as layering 
and integration stages. These examples capture the use 
of  AI for the augmentation of  fraud detection systems, 
demonstrating its capability as a solution to combat more 
complicated forms of  financial crime that cannot be 
addressed through conventional methods.

AI-Powered Fraud Prevention Systems 
AI-driven fraud prevention technology is now embedded 
in the core of  banking technology infrastructure and 
offers automated, real-time fraud prevention solutions. 
Rather, banks want to deploy machine learning models 
for security by means of  analyzing transactions on 
the fly. These can automatically identify cases where 
the behaviour of  a specific user deviates from what is 
considered normal – for instance, either an unusual 
transaction or a connection from an unusual place – and 
notify the authorities to investigate.
Working in real time is one of  the major benefits of  
AI for fraud prevention. While rule based systems are 
checked and worked on in batches or through review 
by a human, an AI system has the capacity of  running 
continuously and being reactive in real-time to flags of  
suspicious behaviour. Should the AI then determine 

a transaction to be suspicious, it may automatically 
block it from going through or request additional 
verification from the customer, preventing possibilities 
for a fraudulent transaction, while still allowing for a 
frictionless experience for valid customers.
Plus, AI models are very effective in minimizing false 
positives, or transactions wrongly identified as fraudulent. 
The high number of  false alerts in conventional fraud 
detection systems can inundate bank employees and result 
in customer dissatisfaction. On the contrary, machine 
learning algorithms can train based on historical data 
so to increase time by time their accuracy in classifying 
legitimate versus fraudulent transactions.
Also, AI fraud prevention systems are continuously updated 
to be ahead of  new fraud strategies. They also improve 
detection as the fraudsters get better by adapting to new 
schemes, all without the need for humans to intervene.

Integration with Existing Systems
Another important move towards securing transactions 
is the deploying of  AI fraud detection systems integrated 
with banks’ current infrastructures. Most banks already 
have their established processes handled through legacy 
systems and any AI implementation needs to blend into 
existing systems without hindering active operations. 
Therefore, the secret to achieving and maintaining this 
integration is to use APIs (Application Programming 
Interfaces) and data pipelines between the old and the 
new technology (Adeleke et al., 2024).
An example of  this type of  system are transaction 
monitoring systems, in which data from several banking 
services, like mobile banking, online payments, and 
ATMs, are collected in real time and introduced into 
machine learning models for fraud detection purposes. 
Banks are able to forward transactions data to AI, that 
in turn detects fraud patterns. It enables an easy flow 
of  information between platforms so that suspicious 
activities can be acted upon right away.
Also, AI systems can more readily and easily be 
connected to cloud computing platforms that provide 
the processing power to leverage use scale data analyses. 
Cloud computing offers the possibility of  storing big data 
sets and training and deploying machine learning models 
without the need for expensive local infrastructure. This 
is helpful for smaller financial institutions that may not 
have the means to develop and maintain such difficult 
fraud detection systems.
In addition, when AI is merged with other methods, 
it gives room for banks to enlarge their capabilities to 
prevent fraud. In line with digital banking, the number 
of  transactions is increasing, and these transactions can 
all be handled in real time through AI systems that are 
capable of  learning patterns of  fraud as the volume of  
such data increases, thereby, providing a strong level of  
security as banks go digital.

Challenges and Limitations 
Using AI for detecting fraud in the banking sector include 



Pa
ge

 
18

3

https://journals.e-palli.com/home/index.php/ajfti

Am. J. Financ. Technol. Innov. 3(1) 177-187, 2025

to anonymising data employed in training these AIs and 
customers would have to provide their consent.
Plus, independently of  GDPR, “Payment Card Industry 
Data Security Standards (PCI-DSS)” is another relevant 
mandate that governs AI systems when payment data is 
involved. PCI-DSS mandates that all financial institutions 
and any third-party vendors engage in rigorous security 
practices to protect cardholder information. Specifically, 
regarding sensitive payment data, it is critical that AI 
systems designed to address fraudulent credit card 
transactions are constructed with a view not to violate 
these standards (Onyekwuluje et al., 2025; Shaul & 
Ingram, 2007).
Also, “Anti-Money Laundering (AML)” laws also define 
the role of  banks and financial institutions in detecting 
money laundering. It is, therefore, very important that 
banks adhere to regulations regarding money laundering 
and that AI systems are designed in such a way that they 
can effectively recognize patterns of  money laundering in 
order to safeguard financial transactions and keep them 
secure and intact (Oztas et al., 2024).

Data Privacy 
Keeping user data private without compromising on 
effective fraud detection is among the biggest issues for 
AI systems in the digital bank arena. But, as financial 
institutions increasingly turn to AI models to sift through 
large pools of  sensitive customer data to identify potential 
indicators of  fraud, the response to this risk cannot 
compromise individuals’ privacy to achieve security. 
Most machine learning and deep learning applications of  
AI must be fed large sets of  data, often including past 
transactions, customer information, and even biometric 
data. These processes and storage also create privacy 
concerns about the data since if  a breach occurred an 
individual’s financial information could be made public.
To alleviate the aforementioned issues, banks should adopt 
international data protection measures. This includes 
practices like encrypting and anonymizing data, securely 
storing data to prevent unauthorized access, and other 
measures to avoid data leaks. Plus, also in line with data 
minimization principles, their collection and processing 
should not exceed what is necessary to detect fraud.
Also, AI systems should necessarily be developed in line 
with privacy laws like GDPR, which grants individuals 
access and control rights over their data, including the 
ability to consult, amend, or delete it. Financial institutions 
should also ensure explainability of  the AI models that 
detect fraud so that customers are aware of  and can 
question data-usage by AI systems that might negatively 
impact them. Finding the right equilibrium between 
effective fraud detection and data privacy requirements is 
crucial to ensure trust and legal compliance in the context 
of  digital banking practices.

Future Directions And Innovations 
Promising technologies such as federated learning, 
explainable AI (XAI) and quantum computing, are 

several benefits, although, it is not without challenges. 
Privacy is one, as AI often needs access to highly-sensitive 
data on customers in order to operate, increasing the 
chance of  breaches. Likewise, the expensive financial 
cost of  executing more complex AI innovation, like 
deep learning networks, may be an obstacle for some 
banks. Moreso, low prevalence of  fraudulent transactions 
makes it difficult to obtain high-quality labelled data to 
train the models, resulting in imbalanced datasets. These 
challenges need to be overcome to use the potential of  AI 
to help in fraud prevention.

Ethical and Regulatory Considerations
The employment of  Artificial Intelligence and Machine 
Learning technologies specifically in fraud detection has 
brought about considerable ethical issues, specifically 
regarding “bias”, “fairness”, and “transparency” 
(Adhikari et al., 2024). Among these is machine learning 
bias. These models are usually trained on past data that 
can include societal biases. If  biases in the datasets used 
to train the AI models are not carefully filtered out, the 
AI systems can reproduce these biases with unsound and 
unmeritocratic results. An AI trained on biased historical 
data could, for example, identify some population group 
as more likely to commit fraud, despite being actually no 
more at risk than others.
AI fairness is yet another primary issue. Artificial 
Intelligence fraud detection must be fair in that there is 
no abusing of  a human being; no one is discriminated 
against, all customers are treated equitably, regardless 
of  race, gender, socio-economic status, etc. This is 
especially problematic within financial services, where 
discriminatory conduct can lead to financial harm, loss 
of  banking access, or being wrongfully accused of  fraud.
“Transparency” is also of  great importance in the context 
of  using AI for fraud detection. Banks should only 
use AI-based decision models that are transparent and 
the reasons for decisions can be explained. Particularly 
when the ramifications are significant, such as blocking 
a transaction or freezing an account, it is important 
for both customers and regulators to comprehend the 
reasoning behind the outcomes of  AI models. The 
absence of  transparency can promote “black boxes” in 
AI systems, raising issues of  trust concerning fairness 
and accountability (Ismaeil, 2024).

Regulatory Compliance 
AI technology fraud detection programs are covered by 
much of  the same financial regulations, as any acceptable 
program must use in order to be legally practiced. This 
is particularly true for sectors that involve sensitive 
financial data like banking. Perhaps more importantly, 
the most pertinent of  these frameworks is the “General 
Data Protection Regulation (GDPR)” which sets 
stringent protocols around personal data collection, 
processing, and storage (Adaji et al., 2025). AI systems 
for fraud detection would also need to be regulated under 
GDPR to safeguard customer information, translating 



Pa
ge

 
18

4

https://journals.e-palli.com/home/index.php/ajfti

Am. J. Financ. Technol. Innov. 3(1) 177-187, 2025

poised to take the future of  AI and ML in fraud detection 
to thrilling new levels (Odeyemi et al., 2024). Because 
quantum computing can process large sets of  data with far 
greater speeds than classical computers, this technology 
can revolutionize fraud detection. These systems could 
be orders of  magnitude faster and efficient in the search 
for fraudulent patterns, patterns that could be analysed 
even in real time, patterns that previously could not be.
A rising second area of  interest is Explainable Artificial 
Intelligence (XAI), which aims at increasing transparency 
and interpretability of  AI models. Since AI is now moving 
into more critical domains such as fraud detection, we 
need to ensure that the decision-making process is 
interpretable to humans. Because XAI provides more 
transparency into how models come to conclusions, it 
can help foster trust amongst customers and regulators.
One new technique, called federated learning, is a new 
way of  training machine learning models in which the 
modelling occurs on distributed devices or servers rather 
than a centralized server, helping financial institutions 
to cooperate in building a better fraud detection system 
without the need to share sensitive customer data. The 
benefits include improved privacy and security as well as, 
more effective training of  models across institutions.

Collaboration and Cross-Industry Solutions 
A rise in the complexity of  fraud as well as reliance 
on digital banking explains the growing collaboration 
between industries to produce more efficient fraud 
detection systems. Collaborative effort between banks, 
tech companies and regulators is needed to combat the 
ongoing evolution of  digital fraud. Partnerships between 
the banks and tech organizations could provide the 
opportunity to create the best and newest technologies 
based on current trends including artificial intelligence, 
machine learning and cyber security.
Regulators also have an important role in enforcing legal 
and ethical issues with regards to fraud detection through 
the use of  AI. International cooperation is also necessary 
since fraudulent schemes are frequently international. 
Collaborating with foreign institutions can provide the ability 
to exchange information, deal with international fraud and 
consolidate financial security methods and protocols.

The Role of  Blockchain 
Another technology that could meaningfully complement 
AI- powered fraud detection is blockchain, which has 
been described as a way to increase transaction security 
and transparency. As a distributed and tamper-proof  
record of  transactions, blockchain technology makes it 
possible for transactions to be public and permanent.
Concerning fraud detection, blockchain technology can 
minimize the ability to commit fraudulent transactions 
such as manipulating transaction information, stealing 
identities, and laundering money. One illustration of  this is 
the transparency of  blockchain where once a transaction 
is imprinted it cannot be changed creating a secure audit 
record. It becomes very difficult to alter transaction data 

or be fraudulent without detection.
In addition, AI systems for fraud detection can take 
advantage of  the decentralized structure of  blockchain 
to improve accuracy of  fraud detection by comparing 
transaction data across different networks. AI and 
blockchain combined can contribute towards a more 
robust and trustworthy financial system.

CONCLUSION
This study looked at how Artificial Intelligence or, AI and 
Machine Learning or ML have transformed the detection 
and prevention of  fraud in digital banking. Digital banking 
contributes to accessibility, efficiency, and financial 
inclusion. The downside to this is also a rise in more 
complex forms of  fraud, requiring advanced technology 
to combat it . Existing traditional fraud detection practices 
which are mostly rule-based, can no longer meet the 
complexity and changing nature of  the digital fraud. AI 
and ML provide a more flexible and anticipatory mode 
to the detection of  fraud. These technologies are able 
to recognize intricate patterns as well as alerts to fraud 
instantaneously which hugely increase the capability 
of  detecting fraud with minimal false positives. Several 
ML methods such as supervised learning, unsupervised 
learning, deep learning, and ensemble methods were 
reviewed and have each been successfully used to fight 
fraud. On top of  that, AI systems make fraud prevention 
more efficient because they continuously learn and 
adapt as new threats arise. The ethics and regulation of  
deploying AI for fraud detection is extremely important, 
such as aspects of  bias, transparency, and privacy of  data. 
Financial institutions are obliged to abide by regulations 
like GDPR, PCI-DSS, and AML laws, while at the same 
time maintain effective fraud detection mechanisms and 
the need to protect customer data. As digital banking is 
new, so must the security mechanism. While the use of  AI 
and ML in fraud detection is a game changer, the constant 
need for development and research is imperative to keep 
up with more complex and advanced forms of  fraud. 
Only through the cooperation between banks, technology 
companies, and regulators can we come up with complete 
and effective fraud prevention solutions. In the future, AI, 
machine learning and blockchain technology integrated 
can have the potential to build a safe, secure, transparent 
and resilient digital banking environment that will build 
trust and protect customer assets in digital economy and 
digital banking.

REFERENCES
Abdelrhman, A. B. (2025). Popularity of mobile 

transaction services in the banking sector. American 
Journal of Financial Technology and Innovation, 3(1), 23–31. 
https://doi.org/10.54536/ajfti.v3i1.3807

Adaji, C. C., Bello, A. A., Ukatu, C. E., Okika, N., 
Agboola, O. K., & Amomo, C. G. (2025). AI-powered 
cybersecurity governance: The role of business 
analysts in ethical AI deployment. International Journal 
of Innovative Science and Research Technology, 10(3), 1384-



Pa
ge

 
18

5

https://journals.e-palli.com/home/index.php/ajfti

Am. J. Financ. Technol. Innov. 3(1) 177-187, 2025

1396. https://doi.org/10.38124/ijisrt/25mar924
Adeleke, A. G., Sanyaolu, T. O., Efunniyi, C. P., Akwawa, 

L. A., & Azubuko, C. F. (2024). API integration in 
FinTech: Challenges and best practices. Finance & 
Accounting Research Journal, 6(8), 1531-1554. https://
doi.org/10.51594/farj.v6i8.1506

Adeyeri, T. B. (2024). AI-based fraud detection in banking 
and financial services. International Journal of Enhanced 
Research in Science, Technology & Engineering, 13(7).

Adhikari, P., Hamal, P., & Baidoo, F. J. (2024). Artificial 
intelligence in fraud detection: Revolutionizing 
financial security. International Journal of Science and 
Research Archive, 13(01), 1457–1472. https://doi.
org/10.30574/ijsra.2024.13.1.1860

Afriyie, J. K., Tawiah, K., Pels, W. A., Addai-Henne, 
S., Dwamena, H. A., Owiredu, E. O., Ayeh, S. A., 
& Eshun, J. (2023). A supervised machine learning 
algorithm for detecting and predicting fraud in credit 
card transactions. Decision Analytics Journal, 6, 100163. 
https://doi.org/10.1016/j.dajour.2023.100163

Alam, S., Ayub, M. S., Arora, S., & Khan, M. A. (2023). 
An investigation of the imputation techniques for 
missing values in ordinal data enhancing clustering 
and classification analysis validity. Decision Analytics 
Journal, 9, 100341. https://doi.org/10.1016/j.
dajour.2023.100341

Ali, A., Abd Razak, S., Othman, S. H., Eisa, T. A., Nasser, 
M., Elhassan, T., Elshafie, H., & Saif, A. (2021). 
Financial fraud detection based on machine learning: 
A systematic literature review. Applied Sciences, 12(19), 
9637. https://doi.org/10.3390/app12199637

Aziz, L. A. R., & Andriansyah, Y. (2023). The role 
of artificial intelligence in modern banking: An 
exploration of AI-driven approaches for enhanced 
fraud prevention, risk management, and regulatory 
compliance. Reviews of Contemporary Business Analytics, 
6(1), 110-132.

Barroso, M., & Laborda, J. (2022). Digital transformation 
and the emergence of the Fintech sector: Systematic 
literature review. Digital Business, 2(2), 100028. https://
doi.org/10.1016/j.digbus.2022.100028

Bello, H. O., Ige, A. B., & Ameyaw, M. N. (2024). 
Adaptive machine learning models: Concepts for 
real-time financial fraud prevention in dynamic 
environments. World Journal of Advanced Engineering 
Technology and Sciences, 12(02), 021–034. https://doi.
org/10.30574/wjaets.2024.12.2.0266

Bello, O. A., & Olufemi, K. (2024). Artificial intelligence 
in fraud prevention: Exploring techniques and 
applications, challenges, and opportunities. Computer 
Science & IT Research Journal, 5(6), 1505–1520.

Bello, O. A., Folorunso, A., Ejiofor, O. E., Budale, F. Z., 
Adebayo, K., & Babatunde, O. A. (2023). Machine 
learning approaches for enhancing fraud prevention 
in financial transactions. International Journal of 
Management Technology, 10(1), 85-109.

Bello, O. A., Folorunso, A., Ogundipe, A., Ajani, O. K., 
Budale, F. Z., & Ejiofor, O. E. (2022). Enhancing 

cyber financial fraud detection using deep learning 
techniques: A study on neural networks and 
anomaly detection. International Journal of Network and 
Communication Research, 7(1), 90-113.

Bhuiyan, M. S. M., Rafi, M. A., Rodrigues, G. N., Mir, 
M. N. H., Ishraq, A., Mridha, M., & Shin, J. (2025). 
Deep learning for algorithmic trading: A systematic 
review of predictive models and optimization 
strategies. Array, 100390. https://doi.org/10.1016/j.
array.2025.100390

Bueno, L. A., Sigahi, T. F., Rampasso, I. S., Filho, W. 
L., & Anholon, R. (2024). Impacts of digitization 
on operational efficiency in the banking sector: 
Thematic analysis and research agenda proposal. 
International Journal of Information Management Data 
Insights, 4(1), 100230. https://doi.org/10.1016/j.
jjimei.2024.100230

Cherif, A., Badhib, A., Ammar, H., Alshehri, S., Kalkatawi, 
M., & Imine, A. (2022). Credit card fraud detection 
in the era of disruptive technologies: A systematic 
review. Journal of King Saud University - Computer and 
Information Sciences, 35(1), 145-174. https://doi.
org/10.1016/j.jksuci.2022.11.008

Chowdhury, R. H. (2024). Advancing fraud detection 
through deep learning: A comprehensive review. 
World Journal of Advanced Engineering Technology and 
Sciences, 12(02), 606–613. https://doi.org/10.30574/
wjaets.2024.12.2.0332

Dangsawang, B., & Nuchitprasitchai, S. (2024). A machine 
learning approach for detecting customs fraud 
through unstructured data analysis in social media. 
Decision Analytics Journal, 10, 100408. https://doi.
org/10.1016/j.dajour.2024.100408

Ezie, O., Oniore, J., & Ajaegbu, P. C. (2023). Financial 
technology and economic growth in Nigeria: 2012Q1-
2022Q4. American Journal of Financial Technology and 
Innovation, 1(1), 35–45. https://doi.org/10.54536/
ajfti.v1i1.2325

Ganaie, M., Hu, M., Malik, A., Tanveer, M., & Suganthan, P. 
(2022). Ensemble deep learning: A review. Engineering 
Applications of Artificial Intelligence, 115, 105151. https://
doi.org/10.1016/j.engappai.2022.105151

Goyal, K., Garg, M., & Malik, S. (2025). Adoption of 
artificial intelligence-based credit risk assessment 
and fraud detection in banking services: A hybrid 
approach (SEM-ANN). Future Business Journal, 11(1). 
https://doi.org/10.1186/s43093-025-00464-3

Hernandez Aros, L., Bustamante Molano, L. X., Moreno 
Hernandez, J. J., & Rodríguez Barrero, M. S. (2024). 
Financial fraud detection through the application 
of machine learning techniques: A literature review. 
Humanities and Social Sciences Communications, 11(1), 
1-22. https://doi.org/10.1057/s41599-024-03606-0

Hernandez, M., Epelde, G., Alberdi, A., Cilla, R., & Rankin, 
D. (2022). Synthetic data generation for tabular health 
records: A systematic review. Neurocomputing, 493, 28–
45. https://doi.org/10.1016/j.neucom.2022.04.053

Hilal, W., Gadsden, S. A., & Yawney, J. (2021). Financial 



Pa
ge

 
18

6

https://journals.e-palli.com/home/index.php/ajfti

Am. J. Financ. Technol. Innov. 3(1) 177-187, 2025

fraud: A review of anomaly detection techniques and 
recent advances. Expert Systems with Applications, 193, 
116429. https://doi.org/10.1016/j.eswa.2021.116429

Huang, Z., Zheng, H., Li, C., & Che, C. (2024). 
Application of machine learning-based K-means 
clustering for financial fraud detection. Academic 
Journal of Science and Technology, 10(1), 33-39. https://
doi.org/10.54097/74414c90

Igah, R., & Luse, A. (2024). Unravelling Biased Blocking 
in The Adoption of Payattitude NFC Electronic 
Payment in Nigeria: An Exploratory Analysis. 
MWAIS 2024 Proceedings. 7. https://aisel.aisnet.org/
mwais2024/7

Ikemefuna, C. D., Okusi, O., Iwuh, A. C., & Yusuf, S. 
(2024). Adaptive fraud detection systems: Using 
ML to identify and respond to evolving financial 
threats. International Research Journal of Modernization 
in Engineering Technology and Science, 6(9). https://doi.
org/10.56726/IRJMETS61738

Islam, S., Gupta, G. R., Chakraborty, A., Singh, S., Soni, A., 
& Patle, C. (2025). Detecting fraudulent transactions 
for different patterns in financial networks using 
layer weighted GCN. Human-Centric Intelligent Systems. 
https://doi.org/10.1007/s44230-025-00097-3

Ismaeil, M. K. A. (2024). Harnessing AI for next-
generation financial fraud detection: A data-driven 
revolution. Journal of Ecohumanism, 3(7), 811–821. 
https://doi.org/10.62754/joe.v3i7.4248

Iwedi, M. (2024). Digital finance infrastructure and 
growth of commercial banking firms in Nigeria. 
Discover Analytics, 2(1). https://doi.org/10.1007/
s44257-024-00022-1

Javaid, M., Haleem, A., Singh, R. P., Suman, R., & Khan, S. 
(2022). A review of blockchain technology applications 
for financial services. BenchCouncil Transactions on 
Benchmarks, Standards and Evaluations, 2(3), 100073. 
https://doi.org/10.1016/j.tbench.2022.100073

Johora, F. T., Hasan, R., Farabi, S. F., Akter, J., & Mahmud, 
M. A. A. (2024). AI-powered fraud detection in 
banking: Safeguarding financial transactions. The 
American Journal of Management and Economics Innovations, 
6(06), 8-22.

Khan, A. A., Chaudhari, O., & Chandra, R. (2023). A 
review of ensemble learning and data augmentation 
models for class imbalanced problems: Combination, 
implementation and evaluation. Expert Systems with 
Applications, 244, 122778. https://doi.org/10.1016/j.
eswa.2023.122778

Khando, K., Islam, M. S., & Gao, S. (2022). The 
emerging technologies of digital payments and 
associated challenges: A systematic literature review. 
Future Internet, 15(1), 21. https://doi.org/10.3390/
fi15010021

Khodabandehlou, S., Golpayegani, A. H., & FiFrau, D. 
(2024). Unsupervised financial fraud detection in 
dynamic graph streams. ACM Transactions on Knowledge 
Discovery from Data, 27(5), 1-29.

Kipngetich, A. (2025). A review of online scams and 

financial frauds in the digital age. GSC Advanced 
Research and Reviews, 22(01), 302-329.

Li, X., Huang, J., Chen, C., & Zhang, Y. (2020). Credit card 
fraud detection using machine learning: A systematic 
literature review. Expert Systems with Applications, 164, 
113909.

Mallesha, C., & Hymavathi, M. (2024). A review on AI 
and fraud detection in accounting: Reducing risks and 
enhancing financial security. Academy of Accounting and 
Financial Studies Journal, 28(2), 1-18.

Mejia, N. (2019). Artificial intelligence at Citibank - 
Current initiatives. EmerJ Artificial Intelligence Research. 
https://emerj.com/ai-at-citi/

Metha, S. (2025). AI-driven fraud detection: A risk scoring 
model for enhanced security in banking. Journal of 
Engineering Research and Reports, 27(3), 23–34. https://
doi.org/10.9734/jerr/2025/v27i31415

Mienye, I. D., Swart, T. G., & Obaido, G. (2024). 
Recurrent neural networks: A comprehensive review 
of architectures, variants, and applications. Information, 
15(9), 517. https://doi.org/10.3390/info15090517

Mohmmed, A. A., Rahma, A. M. S., & AbdulWahab, 
H. B. (2024). Digital wallets evolution: Navigating 
challenges, innovation and the future landscape. 
Al-Qadisiyah Journal of Pure Science, 29(1), Article 4. 
https://doi.org/10.29350/2411-3514.1248

Muthunambu, N. K., Prabakaran, S., Kavin, B. P., 
Siruvangur, K. S., Chinnadurai, K., & Ali, J. (2024). 
A novel eccentric intrusion detection model based 
on recurrent neural networks with leveraging LSTM. 
Computers, Materials & Continua/Computers, 
Materials & Continua (Print), 78(3), 3089–3127. 
https://doi.org/10.32604/cmc.2023.043172

Odeyemi, O., Mhlongo, N. Z. M., Nwankwo, E. E., & 
Soyombo, O. T. (2024). Reviewing the role of AI in 
fraud detection and prevention in financial services. 
International Journal of Science and Research Archive, 
11(01), 2101–2110. https://doi.org/10.30574/
ijsra.2024.11.1.0279

Odufisan, O. I., Abhulimen, O. V., & Ogunti, E. O. 
(2025). Harnessing artificial intelligence and machine 
learning for fraud detection and prevention in Nigeria. 
Journal of Economic Criminology, 100127. https://doi.
org/10.1016/j.jeconc.2025.100127

Oduro, N. D. A., Okolo, N. J. N., Bello, N. A. D., 
Ajibade, N. A. T., Fatomi, N. A. M., Oyekola, N. T. 
S., & Owoo-Adebayo, N. S. F. (2025). AI-powered 
fraud detection in digital banking: Enhancing security 
through machine learning. International Journal of Science 
and Research Archive, 14(3), 1412–1420. https://doi.
org/10.30574/ijsra.2025.14.3.0854

Olowu, N. O., Adeleye, N. A. O., Omokanye, N. A. O., 
Ajayi, N. A. M., Adepoju, N. A. O., Omole, N. O. 
M., & Chianumba, N. E. C. (2024). AI-driven fraud 
detection in banking: A systematic review of data 
science approaches to enhancing cybersecurity. GSC 
Advanced Research and Reviews, 21(2), 227–237. https://
doi.org/10.30574/gscarr.2024.21.2.0418



Pa
ge

 
18

7

https://journals.e-palli.com/home/index.php/ajfti

Am. J. Financ. Technol. Innov. 3(1) 177-187, 2025

Onyekwuluje, T. P., Kpakpa, C. T., Panful, B., Apaflo, B. 
N., & Donkor, A. A. (2025). The role of IT compliance 
in enhancing cybersecurity measures for U.S. financial 
institutions. International Journal of Research Publication 
and Reviews, 6(1), 2600-2606.

Owen, R. (2021). Artificial intelligence at American 
Express - Two current use cases. EmerJ Artificial 
Intelligence Research. https://emerj.com/artificial-
intelligence-at-american-express/

Oyedokun, O., Ewim, S. E., & Oyeyemi, O. P. (2024). 
A comprehensive review of machine learning 
applications in AML transaction monitoring. 
International Journal of Engineering Research and 
Development, 20(11), 730-743. https://www.ijerd.com

Oztas, B., Cetinkaya, D., Adedoyin, F., Budka, M., Aksu, 
G., & Dogan, H. (2024). Transaction monitoring in 
anti-money laundering: A qualitative analysis and 
points of view from industry. Future Generation Computer 
Systems, 159, 161–171. https://doi.org/10.1016/j.
future.2024.05.027

Rahman, M., Yee, H. P., Masud, M. A. K., & Uzir, M. U. 
H. (2024). Examining the dynamics of mobile banking 
app adoption during the COVID-19 pandemic: A 
digital shift in the crisis. Digital Business, 4(2), 100088. 
https://doi.org/10.1016/j.digbus.2024.100088

Rojan, Z. (2024). Financial fraud detection based on 
machine and deep learning: A review. The Indonesian 
Journal of Computer Science, 13(3).

Roszkowska, P. (2021). Fintech in financial reporting and 
audit for fraud prevention and safeguarding equity 
investments. Journal of Accounting & Organizational 
Change, 17(2), 164-196.

Salunke, Y., Phalke, S., Madavi, M., Kumre, P., Bobhate, 
G. (2025). Fraud detection: A hybrid approach with 
logistic regression, decision tree, and random forest. 
Cureus Journal of Computational Science, 2, es44389-
024-02350-5. https://doi.org/10.7759/s44389-024-
02350-5

Sarker, I. H. (2021). Machine learning: Algorithms, 
real-world applications and research directions. SN 

Computer Science, 2(3). https://doi.org/10.1007/
s42979-021-00592-x

Sharma, P. (2024). How reinforcement learning keeps 
fraud detection smart and quick: Adapting to new 
fraud tricks. International Journal for Multidisciplinary 
Research (IJFMR), 6(2). https://www.ijfmr.com

Shaul, J., & Ingram, A. (2007). Database activity 
monitoring. In Elsevier eBooks (pp. 201–224). https://
doi.org/10.1016/b978-159749198-3.50010-x

Tejesh, P., Sai, G. P., Naveen, D., Khaleel, V. S., & Rao, V. 
H. (2025). Detection of fraud in banking transactions 
by machine learning algorithm. International Journal of 
Research Publication and Reviews, 6(4), 1775–1778.

Trucco, E., McNeil, A., McGrory, S., Ballerini, L., 
Mookiah, M. R. K., Hogg, S., Doney, A., & 
MacGillivray, T. (2019). Validation. In Elsevier eBooks 
(pp. 157–170). https://doi.org/10.1016/b978-0-08-
102816-2.00009-5

Venigandla, K., & Vemuri, N. (2022). RPA and AI-driven 
predictive analytics in banking for fraud detection. 
Tuijin Jishu/Journal of Propulsion Technology, 43(4).

Vens, C. (2013). Bagging. In Springer eBooks (pp. 68–69). 
https://doi.org/10.1007/978-1-4419-9863-7_602

Wang, X., Liu, H., & Yu, Z. (2019). A hybrid approach 
for fraud detection in online banking transactions. 
IEEE Access, 7, 64101-64113.

Wei, C., Xie, G., & Diao, Z. (2023). A lightweight deep 
learning framework for botnet detecting at the IoT 
edge. Computers & Security, 129, 103195. https://doi.
org/10.1016/j.cose.2023.103195

Windasari, N. A., Kusumawati, N., Larasati, N., & 
Amelia, R. P. (2022). Digital-only banking experience: 
Insights from Gen Y and Gen Z. Journal of Innovation 
& Knowledge, 7(2), 100170. https://doi.org/10.1016/j.
jik.2022.100170

Xuan, C. D., Duong, D., & Dau, H. X. (2021). A 
multi-layer approach for advanced persistent threat 
detection using machine learning based on network 
traffic. Journal of Intelligent & Fuzzy Systems, 40(6), 
11311-11329.


