Credit Cards Frauds and Cybersecurity Threats Machine Learning Detection Algorithms As Countermeasures

International Journal of Trend in Scientific Research and Development (IJTSRD)
Volume 6 Issue 7, November-December 2022 Available Online: www.ijtsrd.com e-ISSN: 2456 – 6470
Credit Cards Frauds and Cybersecurity Threats: Machine

Learning Detection Algorithms as Countermeasures
Obodoeze Fidelis C.1, Oliver Ifeoma Catherine2, Onyemachi George Olisamaka2,
Udeh Ifeanyi Frank Gideon3, Obiokafor, Ifeyinwa Nkemdilim4
1
Department of Computer Engineering, Akanu Ibiam Federal Polytechnic Unwana, Ebonyi State, Nigeria
2
Department of Computer Science, Akanu Ibiam Federal Polytechnic Unwana, Ebonyi State, Nigeria
3
Department of Computer Science, Federal College of Agriculture, Ishiagu, Ebonyi State, Nigeria
4
Department of Computer Science, Anambra State Polytechnic, Mgbakwu, Anambra State, Nigeria
ABSTRACT How to cite this paper: Obodoeze

Credit and Debit cards have become the choice mode of payment Fidelis C. | Oliver Ifeoma Catherine |
online as a result of the proliferation of electronic transactions and Onyemachi George Olisamaka | Udeh
advancement in Information and Communication Technology (ICT). Ifeanyi Frank Gideon | Obiokafor,
Ifeyinwa Nkemdilim "Credit Cards
Because of the increased use of credit cards for payment online, the
Frauds and Cybersecurity Threats:
number of fraud cases associated with it has also increased;
Machine Learning Detection Algorithms
scammers and fraudsters are stealing credit card information of as Countermeasures" Published in
victims online and thereby stealing their monies. There is the need International
therefore to stop or abate these frauds using very powerful fraud Journal of Trend in
detection system that detects patterns of credit card frauds in order to Scientific Research
prevent it from occurring. In this paper we x-rayed the concept of and Development
credit card frauds and how they are carried out by fraudsters. Python (ijtsrd), ISSN:
3.7.6 programming language, Jupyter Notebook 6.0.3 and Anaconda 2456-6470,
Volume-6 | Issue-7, IJTSRD52440
Navigator 1.9.12 were used as experimental test bed. Also, we
implemented two different supervised machine learning algorithms December 2022,
pp.940-948, URL:
on an imbalanced dataset such as Decision Tree and Random forest
www.ijtsrd.com/papers/ijtsrd52440.pdf
techniques. A comparative analysis of the credit card detection
capabilities of these machine learning algorithms were carried out to Copyright © 2022 by author (s) and
ascertain the best detection algorithm using different performance International Journal of Trend in
evaluation metrics such as accuracy, precision, recall, f1 score, Scientific Research and Development
confusion matrix. Experimental results showed that Random Forest Journal. This is an
outperformed Decision Tree algorithm slightly in performance Open Access article
metrics used for performance evaluation. distributed under the
terms of the Creative Commons
KEYWORDS: Credit Card frauds, Accuracy, f1 score, precision, Attribution License (CC BY 4.0)
recall, support, fraud detection, fraud patterns, machine learning (http://creativecommons.org/licenses/by/4.0)
algorithms
1. INTRODUCTION:
Cybersecurity is becoming increasingly significant in Here, a system that can monitor the patterns of all
our daily lives. When addressing digital life security, transactions is required, and if any patterns are
the main issue is identifying anomalous behavior. abnormal, the transaction should be stopped or
When making purchases from online e-commerce terminated.
stores or conducting business online, many people Credit card information should always be kept private
frequently prefer using credit cards as well as debit or confidential. Information about credit card privacy
cards. Occasionally, we can make purchases even should not be compromised. Phishing websites, stolen
when we don't have the cash on hand thanks to credit or lost credit cards, fake credit cards, the theft of card
card credit limitations. information, intercepted cards, etc. are some
On the other hand, scammers and online attackers examples of ways to steal credit card information
abuse these features. To solve this problem, there is (Anderson, 2007). The aforementioned activities
need for a system that can abort the transaction if it should be avoided for security reasons. Online fraud
finds anything fishy or anomalous patterns in the simply requires the card information and takes place
whole financial transaction.
@ IJTSRD | Unique Paper ID – IJTSRD52440 | Volume – 6 | Issue – 7 | November-December 2022 Page 940
International Journal of Trend in Scientific Research and Development @ www.ijtsrd.com eISSN: 2456-6470
remotely. At the time of purchase, neither a manual domain registration and hosting, mobile phones, fuel,
signature nor a PIN or card imprint are necessary. The groceries, air ticket booking, books hotels and other
legitimate cardholder is typically unaware that items. When using credit cards for various purchases,
someone else has seen or stolen his or her card they are most valuable because they offer numerous
information. rewards in terms of points. Making payments with
credit cards is equally easy and seamless. But the
The easiest way to spot this kind of fraud is to
examine each card's purchasing habits and look for challenge of using credit cards for online payments is
as a result of activities of hackers and scammers.
any deviations from the "normal" spending habits. The
greatest strategy to lower the number of successful According to ProjectPro (2022), today's credit card
credit card frauds is to detect fraud by examining the fraud falls into a number of different categories:
current cardholder data purchases. Since the outcomes Lost or stolen cards: Online shopping credit
are not made public and the data sets are unavailable. cards are stolen and used fraudulently on the
The logged data and user behavior are two data types owner's behalf. The process of canceling stolen
that can be used to identify fraud incidents. Currently, credit cards and reissuing them is difficult for
a variety of techniques, including data mining, both customers and credit card providers. Several
statistics, and artificial intelligence, are used to detect banking organizations limit the use of credit cards
fraud. until it is certain that the card's legitimate owner
has received it.
Currently, there are many supervised machine
Card Abuse: The customer uses a credit card to
learning approaches based on Artificial Intelligence
make purchases, but he or she has no intention of
(AI) that can classify odd or anomalous transactions.
The only prerequisites are historical data and an returning the money that the bank has charged for
algorithm that can more closely match the data. those purchases. When the due date for payment
approaches, some customers quit returning calls.
In order to identify frauds and potential cyber risks, They even occasionally file for bankruptcy; every
decision tree and random forest machine learning year, this kind of scam causes losses in the
(ML) algorithms were utilized in this paper to identify millions to banks.
unusual or odd and unexpected patterns in credit card Identity Theft: Customers submit false
transactions. The performance of the two ML information while applying for credit cards, and
algorithms is evaluated experimentally using a variety they may even steal the personal information of a
of performance evaluation criteria, including real client to do so. Even card blocking cannot
accuracy, precision, recall and F1 score. prevent the credit card from getting into the
1.1. How Credit Card Fraud works wrong hands in such circumstances.
Credit cards are very important in making purchases Merchant Abuse: Some online merchants
online especially when finances are not readily display fictitious unlawful transactions to
available. Credit cards are one of the most popular facilitate money laundering. Legal information of
financial tools used to make online purchases and legitimate credit card customers is stolen to create
payments for commodities such as electronic gadgets counterfeit cards and be used for these illegal
such TVs, computer hardware and software, website transactions.
Fig.1 shows a comprehensive list of credit card frauds that can compromise the cybersecurity of credit
cards and financial systems.
Fig.1. Types of Fraud (Anderson, 2007)

To identify credit card fraud, several traditional, obsolete techniques have been used since the dawn of time,
including CVV verification, geolocation tracking, IP address verification, etc. However, since thieves improve
their skills with time, it is challenging to put an end to them all using only traditional methods. In today's
environment, when millions of transactions occur every second, human intelligence is incapable of processing
all the data and identifying the patterns of behavior of fraudsters. Machine learning-based credit card fraud
detection is essential in this situation.
Financial organizations are increasingly depending on automated machine learning algorithms to make informed
judgments and prevent large losses. These safety measures greatly reduce the risk involved in conducting online
transactions. Machine learning algorithms use historical transaction data to appraise present and future
transactions in a manner similar to how people do. Despite the fact that computers may not be as intelligent as
people and may need some additional supervision, the speed at which data is processed and calculations are
performed is an advantage. Machines are also more adept than people at recognizing and recalling patterns in
vast volumes of data. Anomaly detection is the term most often used to describe these methods.
1.2. Credit Card Fraud Detection Techniques using Machine Learning
In the past, credit card fraud has been identified using the following machine learning techniques:
Unsupervised Learning - Algorithms for machine learning, such as Isolation Forest, One-class SVM, LOF,
etc., train models without the need of labeled training data. They look for patterns in the data and attempt to
group the data points based on apparent similarities in patterns.
Supervised Learning – These machine Learning Algorithms include KNN, Artificial Neural Networks
(ANNs), Auto encoders, Random Forest, XGBoost, LightGBM, etc. These algorithms learn to anticipate the
labels or patterns for the unobserved data by training on labeled data. It can be costly and difficult to collect
labeled data for machine learning modeling.
2. Materials and Methods
In this paper, we used Supervised Learning ML algorithms such as Decision Tree, Random Forest and XGBoost
to perform modeling experiments in Python 3.7.6 and Jupyter Notebook 6.0.3 in order to detect credit card
patterns that are abnormal and unusual that may contain some frauds in the credit card dataset.
The historical credit card dataset used in the experiments contains transactions made by credit cards in
September 2013 by European cardholders.
This dataset presents transactions that occurred in two days, where we have 492 frauds out of 284,807
transactions. The dataset was downloaded online at https://www.kaggle.com/datasets/mlg-ulb/creditcardfraud.
Fig. 2 shows the methodology adopted in the experiments.
The system architecture of the proposed machine learning method is shown in Fig.3.
Import the credit card historical dataset
to Python testbed
Perform Exploratory Data Analysis

(EDA) on the dataset
Clean up the dataset and prepare it for

ML modeling
Fit the ML algorithms on the dataset

and predict credit card fraud patterns
Train and evaluate the models on the

dataset
Evaluate the fraud detection

performances of the ML algorithms
Select the better model for credit card

fraud detection
Fig. 2: Credit card fraud detection methodology using Machine Learning Algorithms
Credit card Exploratory Data

dataset Analysis and sampling
Clean up the dataset
Divide the dataset into Training set

(70%) and Testing set (30%)
Decision Tree Random Forest XGBoost
Test Credit card fraud prediction and testing

Data
Performance Evaluation of ML models
Fig. 3: Credit card fraud detection system architecture

The following tools were used for the programming and modeling of the credit card detection models using
machine learning algorithms:
Python – 3.7.6
Numpy – 1.19.2
Scikit-learn – 0.24.1
Matplotlib – 3.3.4
Anaconda Navigator 1.9.12
Jupyter Notebook 6.0.3
Algorithm steps for finding the Best algorithm for Credit fraud detection:
Step1: Import the credit card dataset into Pandas data frame
Step2: Convert the data into data frames format suitable for machine learning modeling
Step3: Do random sampling
Step4: Split dataset into training set (70%) and testing set (30%)
Step5: Fit the training dataset to the machine learning algorithms
Step6: Apply the algorithms to the training dataset and create the prediction models
Step7: Make Credit card fraud prediction for test dataset for each algorithm
Step8: Compute the performance metrics for each machine learning algorithm
Random Forest ML algorithm:
Random forest (one of the most popular algorithms) is a supervised machine learning algorithm. It creates a
“forest” out of an ensemble of “decision trees”, which are normally trained using the “bagging” technique. The
bagging method’s basic principle is that combining different learning models improves the prediction outcome.
To get a more precise and reliable forecast, random forest creates several decision trees and merges them. Fig.4
shows the Random Forest algorithm’s anomaly detection technique using majority class rule.
Fig 4: Random Forest anomaly detection using Majority class technique

Decision Tree ML algorithm:
A decision tree is a non-parametric supervised learning algorithm, which is utilized for both classification and
regression tasks. It has a hierarchical, tree structure, which consists of a root node, branches, internal nodes and
leaf nodes (IBM, 2022).
The experiments were carried out in Anaconda Navigator running Python 3.0 and Tensorflow environment as
shown in Fig. 5.
Fig.5: Anaconda Navigator Development Environment for Python
2.1. Exploratory Data Analysis
Exploratory data analysis (EDA) was carried out in Python 3, Jypyter Notebook and several statistical data
analysis tool. Figs. 6 and 7 show the credit card data analysis carried out in Pandas dataframe.
Fig. 6: Data Analysis showing credit card dataframe in Pandas
Fig.7: Data Analysis showing data types in Pandas

Fig.8 shows the printout of the percentage of credit card frauds in the database used for the experiments.
Fig.8: Data Analysis showing percentage of credit card incidences in the dataset
Fig.9 shows the printout of the dataset ratio split used for the experiments. The dataset was divided into two sets,
training set was allocated 70% while testing and evaluation set was allocated 30%.
Fig.9: The Dataset split into Training and Testing ratio

Performance Evaluation Tools:
The following performance evaluation metrics were used to evaluate the credit card prediction performances of
the machine learning algorithms used in this paper. These metrics (Accuracy, Precision, Recall and F1- Score)
are calculated using True Positive (TP), True Negative (TN), False Positive (FP), and False Negative (FN)
parameters.
When TP is likely to cause most kidney issues in a kidney patient; FPs will likely find that the rate at which
kidney disease is detected is the rate at which a healthy person is found. TN hopes to reveal that a person with
kidney disease is healthy. FN is prescribed when a healthy man or woman has kidney disease.
Accuracy:
Accuracy is the proportion of instances that are accurately classified. It is one of the most popular performance
measures for machine learning categorization.
Accuracy = ……………………… (1)
Where TP, TN, FP and FN are referred to as True Positives, True Negatives, False Positives and False Negatives
respectively as used in binary classification machine learning tasks.
Precision:
Precision is the number of classified Positive or fraudulent instances that actually are positive instances.
Precision = ……………………………………………………………… (2)
Recall:
Recall is a metric that measures the proportion of accurate positive predictions among all possible positive
predictions. Recall gives an indicator of missed positive predictions, unlike precision, which only comments on
the accurate positive predictions out of all positive predictions. The number of true positives divided by the sum
of true positives and false negatives is used to determine recall.
Recall = ………………………………………………….………………(3)
F1-Score:
F1 Score is the weighted average of Precision and Recall. Therefore, this score takes both false positives and
false negatives into account.
F1-Score = ………………………………….………………(4)
3. Results and Discussions

This section describes the results obtained from several experiments as carried out in the testbed. Fig. 9 shows
the modeling prediction of credit card fraud by Decision Tree and Random Forest machine learning algorithms.
Fig. 10: Modeling prediction results for credit card frauds by Decision Tree and Random Forest
Algorithms
From Fig.10, it is clear that Random Forest algorithm outperformed the Decision Tree Algorithm by Accuracy
Score of 99.96% while Decision Tree scored 99.93%.
We also carried out another experiment to comprehensively evaluate the performance of the Decision Tree and
Random Forest algorithms using standard performance evaluation metrics such as Accuracy, precision, Recall
and f1-score. Fig.11 and Table 1 shows the performance evaluation results obtained for Decision Tree algorithm.
Decision Tree performed very well in credit card pattern detection with Accuracy score of 0.99932 or 99.93%
and Precision of 0.76712, Recall=0.82353 and F1-score of 0.79433.
Fig. 11: Modeling prediction performance evaluation for Decision Tree Algorithm
Table 1: Decision Tree credit card fraud detection results
Performance Metric Score
Accuracy 0.99932
Precision 0.76712
Recall 0.82353
F1-score 0.79433
Fig.12 and Table 2 show the performance evaluation results obtained for Random Forest algorithm. Random
Forest performed very well (better than Decision Tree algorithm) in credit card pattern detection with Accuracy
score of 0.99961or 99.96% and Precision of 0.94783, Recall=0.80147 and F1-score of 0.86853.
Fig. 12: Modeling prediction performance evaluation for Random Forest Algorithm
Table 2: Random Forest credit card fraud detection results
Performance Metric Score
Accuracy 0.99961
Precision 0.94783
Recall 0.80147
F1-score 0.86853
Table 3 shows the comparative prediction performances between Decision Tree and Random Forest algorithms.
It is very clear here that Random Forest outperformed Decision Tree algorithm in credit card fraud detection
with higher Accuracy, Precision and F1-score and lower Recall score.
Table 3: Comparison of credit card frauds prediction performances between Decision Tree and
Random Forest algorithms
Performance Metric Score ML Algorithm
0.99932 Decision Tree
Accuracy
0.99961 Random Forest
Precision
Recall
F1-score
4. Summary and Conclusion 4.1. Future Scope

In this paper, we built a binary classifier using three Future experiments will be carried to determine the
machine learning algorithms such as Random Forest performances of hybrid ensemble machine learning
and Decision Trees classifiers to detect credit card algorithms by combining ensemble algorithms such
fraud transactions. Through this project, we as XGBoost, Random Forest with Decision Trees
understood and applied techniques to address the algorithm in carrying out fraud detection in credit
class imbalance issues and achieved an accuracy of cards.
more than 99%.
References:
We applied different performance evaluation metrics [1] Anderson, R. 2007. The Credit Scoring Toolkit:
such as Precision, Recall, f1-score, support and theory and practice for retail credit risk
accuracy to evaluate the performance of the two management and decision automation. New
Machine learning algorithms. Our comparative York: Oxford University Press.
performance evaluation shows that Random Forest [2] IBM (2022).” What is a decision Tree?”.
outperformed the Decision Tree algorithm in credit Retrieved online at
card fraud patterns detection. https://www.ibm.com/topics/decision-trees
The results obtained from the experiments show that [3] ProjectPro (2022).”How does Credit Card
Fraud work?”. Accessed online at
Machine learning classification algorithms such as
Decision Tree and Random Forest are well suited to https://www.projectpro.io/article/credit-card-
fraud-detection-project-with-source-code-in-
detect any credit card fraud before it occurs.
python/568

Credit Cards Frauds and Cybersecurity Threats Machine Learning Detection Algorithms As Countermeasures

Uploaded by

Copyright:

Available Formats

Credit Cards Frauds and Cybersecurity Threats Machine Learning Detection Algorithms As Countermeasures

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Credit Cards Frauds and Cybersecurity Threats Machine Learning Detection Algorithms As Countermeasures

Uploaded by

Copyright:

Available Formats

International Journal of Trend in Scientific Research and Development (IJTSRD)

Credit Cards Frauds and Cybersecurity Threats: Machine

ABSTRACT How to cite this paper: Obodoeze

Fig.1. Types of Fraud (Anderson, 2007)

Perform Exploratory Data Analysis

Clean up the dataset and prepare it for

Fit the ML algorithms on the dataset

Train and evaluate the models on the

Evaluate the fraud detection

Select the better model for credit card

Credit card Exploratory Data

Clean up the dataset

Divide the dataset into Training set

Decision Tree Random Forest XGBoost

Test Credit card fraud prediction and testing

Performance Evaluation of ML models

Fig. 3: Credit card fraud detection system architecture

Fig 4: Random Forest anomaly detection using Majority class technique

Fig.5: Anaconda Navigator Development Environment for Python

Fig. 6: Data Analysis showing credit card dataframe in Pandas

Fig.7: Data Analysis showing data types in Pandas

Fig.9: The Dataset split into Training and Testing ratio

3. Results and Discussions

4. Summary and Conclusion 4.1. Future Scope

You might also like