DEVELOPMENT OF AN AUTOMATED REAL-TIME CREDIT CARD FRAUD DETECTION SYSTEM

DEVELOPMENT OF AN AUTOMATED REAL-TIME CREDIT CARD FRAUD DETECTION SYSTEM

ABSTRACT

Assume you have a credit card in your possession. Your previous spending patterns will be discovered. For example, how much money you spend, where you spend it, how often you spend it, and what you buy. If your current credit card transaction deviates from your previous spending habits, it will be suspected of fraud; otherwise, it will be treated as a legitimate transaction and fraud transactions will be alerted in the dashboard. Millions of transactions will be used to make such predictions. Distributed frameworks that can scale as the number of transactions increases are therefore employed. Spark Kafka and Cassandra are used to create this system for real-time credit card fraud detection. Preprocessing is done using Spark Machine Learning Pipeline Stages such String Indexer, Vector Slicer, Standard Scaler, and Vector Assembler. Vector Slicer, Standard Scaler and Vector Assembler is used for Preprocessing. Utilizing the Random Forest Algorithm, a Machine Learning model is produced. K-means Algorithm is used for data balancing. Automation of both Spark Machine Learning and Spark Streaming with Kafka and Cassandra is done using Apache Airflow.

 

 

CHAPTER 1  

1.1 INTRODUCTION

From the moment payment systems came to existence, there have always been people who will find new ways to access someone‘s finances illegally. This has become a major problem in the modern era, as all transactions can easily be completed online by only entering your credit card information. (Chuprina, 2021).

Fraud is as old as mankind itself and can take an unlimited variety of different forms (Lakshmi, Mettu ,& Hameed, 2020). As a result, fraud detection has become an important and urgent task for businesses. Necessary prevention measures can be taken to stop this abuse and the behavior of such fraudulent practices can be studied to minimize it and protect against similar occurrences in the future.

The solutions to the fraud can be categorized into prevention and detection. Fraud prevention involves preventing the fraud in the source itself. The technologies like the Address Verification System (AVS) and Card Verification System (CVM) are usually operated to prevent fraud. (Shakya, 2018)

When fraud cannot be prevented, it must be recognized as quickly as feasible and appropriate action taken. The action conducted following a fraudulent incident is known as fraud detection. It is the process of determining whether or not a transaction is valid. It entails tracking the behaviors of large groups of people in order to predict, detect, and avert unwanted conduct such as fraud, intrusion, and defaulting. These frauds are classified as:

• Credit Card Frauds: Online and Offline
• Card Theft
• Account Bankruptcy
• Device Intrusion
• Application Fraud
• Counterfeit Card
• Telecommunication Fraud

The current era’s rapid technical growth has fueled a demand for more innovative payment options.

Payment methods such as cheques and cash were previously used. Credit card usage has risen dramatically over the world, and many people now believe in going cashless and rely solely on internet transactions. Thanks to credit cards, digital transactions have become easier and more accessible. A payment card, such as a credit card or debit card, is used to perpetrate fraud, and this practice is referred to as credit card fraud. The goal could be to pay into another account that is under criminal control or to receive goods or services. Credit card fraud can be authorized—where a legitimate customer uses their own credit card to make a payment to an account that is managed by a criminal—or unauthorized—where the account holder does not give consent for the payment to occur and a third party completes the transaction.

There are two kinds of card fraud:

• card-present fraud
• card-not-present fraud

The compromise can occur in several ways and can usually occur without the knowledge of the cardholder. A compromised account’s credentials may be kept by a fraudster for months prior to any theft, making it challenging to pinpoint the point of penetration. Stolen cards can be reported swiftly by cardholders. Unauthorized use might not be discovered by the cardholder until they get a statement.

Until the cardholder contacts the issuing bank and the bank places a block on the account, the credit card can be used for unauthorized purchases. For early reporting, most banks offer free, 24-hour telephone numbers. Even so, before the card is cancelled, a thief may use the card to make illicit purchases.

There are various types of payment card fraud

• Application fraud: Application fraud takes place when a person uses stolen or fake documents to open an account in another person’s name. To create a personal profile, criminals may steal or fabricate papers like utility bills and bank accounts. The fraudster could take money out of the account or apply for credit in the victim’s name after opening the account with bogus or stolen documents.
• Account takeover: An account takeover refers to the act by which fraudsters will attempt to assume control of a customer’s account (i.e. credit cards, email, banks, SIM card and more).
• Due to its varied characteristics, including class imbalance, this issue is particularly difficult to solve from the standpoint of learning. Fraudulent transactions are vastly outnumbered by legitimate ones, and transaction patterns frequently shift over time.
• Skimming: Skimming is the stealing of private data that was used in a typically routine transaction. Employing simple techniques like duplicating receipts or more sophisticated ones like using a small electrical gadget (skimmer) to swipe and save several victims’ card numbers, the thief can obtain the victim’s card number.
• Unexpected repeat billing: Repeat billing, also referred to as “repeated bank charges,” is a result of online bill payment or online transactions made with a bank account. These are banker’s orders or standing orders from customers to honor and pay the payee a specific sum each month.

In Big computing communities such as machine learning and data science the solution to this problem can be automated. Due to its varied characteristics, including class imbalance, this issue is particularly difficult to solve from the standpoint of learning. Fraudulent transactions are vastly outnumbered by legitimate ones, and transaction patterns frequently shift over time. stical properties over the course of time, and considering the huge traffic of transaction data, and it is not possible for humans to check manually every transaction one by one if it is fraudulent or not.

Algorithms for machine learning are used to analyze all permitted transactions and flag any that seem suspect. Professionals look into these reports and get in touch with the cardholders to confirm whether the transaction was legitimate or fraudulent.

 

1.2 BACKGROUND OF STUDY

With the growth of online business around Nigeria, the number of credit card frauds has also increased drastically. The fraudulent credit card transactions result in yearly losses of a significant sum of money. Unauthorized financial fraud losses using credit cards and online banking in the UK reached £844.8 million in 2018. While in 2018, banks and credit card companies stopped £1.66 billion in unlawful fraud. This translates to the prevention of £2 out of every £3 of attempted fraud. In 2015, fraud losses on credit, debit, and prepaid cards issued worldwide reached $21.84 billion, according to a Bloomberg report. By 2020, Bloomberg predicts that this could grow by a rate of 45 percent. (Jesus, 2019)

Interestingly credit card fraud affects card owners the least because their liability is limited to the transactions made. The existing legislations and cardholder protection policies as well as insurance schemes in most countries protect the interests of the cardholders. However, the most affected are the merchants, who, in most situations, do not have any evidence (e.g. digital signature) to dispute the cardholders‘ claim of misused card information. Merchants end up bearing all the loses due to chargeback, shipping cost of goods, card issuer fees and charges as well as their own administrative costs. Numerous fraudulent incidents involving the same business can scare away customers, force banks that provide credit cards to stop accepting payments, and harm the business’ brand and goodwill.  (Amanze & Onukwugha, 2018)

Some famous credit card fraud attacks:

Between July 2005 and mid-January 2007, a breach of systems at TJX Companies exposed data from more than 45.6 million credit cards. Albert Gonzalez is accused of being the ringleader of the group responsible for the thefts. In August 2009 Gonzalez was also indicted for the biggest known credit card theft to date information from more than 130 million credit and debit cards was stolen at Heartland Payment Systems, retailers 7-Eleven and Hannaford Brothers, and two unidentified companies.

Approximately 40 million sets of credit card data were stolen from Adobe Systems in 2012 as a result of a cyberattack. According to Chief Security Officer Brad Arkin, the data exposed included client names, encrypted credit card numbers, expiration dates, and details on orders.

In July 2013, press reports indicated four Russians and a Ukrainian were indicted in the U.S. state of

New Jersey for what was called “the largest hacking and data breach scheme ever prosecuted in the United States.” Albert Gonzalez was named as a co-conspirator in the attack, which resulted in more than $300 million in losses and at least 160 million credit card losses. American and European businesses, such as Citigroup, Nasdaq OMX Group, and PNC Financial, were impacted by the attack financial Services Group, Visa licensee Visa Jordan, Carrefour, J. C. Penny and JetBlue Airways.

A Target Corporation system breach occurred between November 27 and December 15, 2013, exposing information from roughly 40 million payment cards. Names, account numbers, expiration dates, and card security codes were among the data taken.

A hacking attack that occurred between July 16 and October 30, 2013, exposed about a million sets of credit card data kept on Neiman-Marcus computers. Target’s systems were compromised by malware that was intended to hook into cash registers and monitor the credit card authorisation process (RAM-scraping malware), exposing data from as many as 110 million consumers.

The Home Depot acknowledged that their payment systems had been breached on September 8th, 2014. Later, they issued a statement claiming that the incident led to the theft of 56 million credit card details by hackers.

In a planned theft on May 15, 2016, a gang of about 100 people stole $12,7 million from 1400 convenience stores in Tokyo over the course of three hours using the information from 1600 South African credit cards. They are thought to have gained enough time to escape Japan before the robbery was uncovered by operating on a Sunday and in a nation other than the bank that issued the cards.

1.3 STATEMENT OF PROBLEM

The ease of payment credit cards has brought to the marketing world is immeasurable. Transactions can now be made without any hassle. The world at large is taking advantage of this technology and now, almost all transaction are made with credit cards. This plight has brought about a need for more a secure way to handle these transactions. Fraudsters have been taking advantage of the technology to rob people of their money and they need to be stopped.

When a credit card is copied or stolen, the transactions made by them are labeled as fraudulent. These fraudulent transactions should be prevented or detected in a timely manner otherwise the resulting losses can be huge.

There has been a growing amount of financial losses due to credit card frauds as the usage of the credit cards become more and more common. As a result, numerous articles [2, 20] detailed significant losses in various nations. 64 I. Elikucuk and E. Duman There are numerous ways to commit credit card fraud, including straightforward theft, application fraud, using fake cards, never receiving issue (NRI), and online fraud (where the card holder is not present). In online fraud, the transaction is made remotely and only the card‘s details are needed. A manual signature, a PIN or a card imprint are not required at the time of purchase. Though prevention mechanisms like CHIP&PIN decrease the fraudulent activities through simple theft, counterfeit cards and NRI; online frauds (internet and mail order frauds) are still increasing in both amount and number of transactions. Online scams accounted for around 50% of all credit card fraud losses in 2008, according to Visa reports about European countries. When the fraudsters obtain a card, they usually use (spend) all of its available (unused) limit. Statistics show that people accomplish this in four to five transactions on average [18]. Thus, although the standard predictive modeling performance metrics are extremely essential for the fraud detection problem, as specified by bank authorities, a performance criterion, assessing the loss that can be avoided on the cards whose operations are identified as fraudulent, is more prominent. In other words, identifying fraud on a bigger available limit card is more beneficial than detecting fraud on a smaller available limit card.

1.4 AIMS AND OBJECTIVES

The aim of this project is to develop and automate a real-time credit card fraud detection system so as to prevent the continuity of credit card fraud by detecting fraudulent online transactions

The specific objectives are:

1. to create a credit card fraud detection system
2. to enable the credit card fraud detection system to detect fraud in real-time transaction
3. to automate both Machine Learning and Real-time Streaming Job using Apache Airflow
4. to create a fraud alert dashboard and display all the predicted fraudulent transactions

1.5 SIGNIFICANCE OF THE STUDY

The process of searching for fraud is lengthy due to the amount of data involved. The credit card dataset is classified using the random forest method in the suggested system. An approach for classification and regression is called Random Forest. In a nutshell, it is a group of decision tree classifiers. Random forest has advantage over decision tree as it corrects the habit of overfitting to their training set. A subset of the training set is sampled randomly so that to train each individual tree and then a decision tree is built, each node then splits on a feature selected from a random subset of the full feature set. Even for large data sets with many features and data instances training is extremely fast in random forest and because each tree is trained independently of the others. The Random Forest algorithm has been found to provide a good estimate of the generalization error and to be resistant to overfitting. (Lakshmi, Mettu ,& Hameed, 2020)

1.6 LIMITATIONS OF STUDY

Several challenges are associated with credit card fraud detection, some of these challenges are: determining which learning strategy to use (e.g., supervised learning or unsupervised learning), which algorithms to use (e.g., Logistic regression, decision trees, etc.), which features to use, how to deal with the class imbalance problem (fraudulent cases are extremely sparse as compared to the legitimate cases). (Shakya, 2018) The profile of fraudulent conduct is dynamic, meaning that fraudulent transactions frequently resemble genuine ones; credit card transaction databases are infrequent and severely biased; selection of features (variables) for the models that is optimal; appropriate metric to assess the effectiveness of strategies on distorted credit card fraud data.

DEVELOPMENT OF AN AUTOMATED REAL-TIME CREDIT CARD FRAUD DETECTION SYSTEM. GET MORE COMPUTER SCIENCE PROJECT TOPICS AND MATERIALS

Sharing is caring!