Credit Card Fraud Detection Methods Information Technology Essay

The credit card is a small plastic card issued to users as a system of payment. It allows its cardholder to buy goods and services based on the cardholder’s promise to pay for these goods and services. Credit card security relies on the physical security of the plastic card as well as the privacy of the credit card number. CVV (Card Verification Value Code) is an anti-fraud security feature to help verify that you are in possession of your credit card. CVV is a new authentication procedure established by credit card companies to further efforts towards reducing fraud for internet transactions. Globalization and increased use of the Internet for online shopping has resulted in a considerable proliferation of credit card transactions throughout the world. Thus a rapid growth in the number of credit card transactions has led to a substantial rise in fraudulent activities. Occurrence of credit card fraud is increasing dramatically due to the exposure of security weaknesses in traditional credit card processing systems resulting in loss of billions of dollars every year. Credit card fraud is a wide-ranging term for theft and fraud committed using a credit card as a fraudulent source of funds in a given transaction. Credit card fraudsters employ a large number of techniques to commit fraud. To combat the credit card fraud effectively, it is important to first understand the mechanisms of identifying a credit card fraud. Over the years credit card fraud has stabilized much due to various credit card fraud detection and prevention measures.

Related Works

Fraud detection involves monitoring the behavior of users in order to estimate, detect, or avoid undesirable behavior. Credit card fraud detection has drawn quite a lot of interest from the research community and a number of techniques have been proposed to counter fraud. To counter the credit card fraud effectively, it is necessary to understand the technologies involved in detecting credit card frauds and to identify various types of credit card frauds [20] [21] [22] . Depending on the type of credit card fraud various measures and mechanisms can be adopted and implemented to counter those credit card frauds. There are multiple algorithms for credit card fraud detection [21] [29]. They are artificial neural-network models which are based upon artificial intelligence and machine learning approach [5] [7] [9] [10] [16], distributed data mining systems [17] [19], sequence alignment algorithm which is based upon the spending profile of the cardholder [1] [6], intelligent decision engines which is based on artificial intelligence [23], Meta learning Agents and Fuzzy based systems [4]. The other technologies involved in credit card fraud detection are Web Services-Based Collaborative Scheme for Credit Card Fraud Detection in which participant banks can share the knowledge about fraud patterns in a heterogeneous and distributed environment to enhance their fraud detection capability and reduce financial loss [8] [13], Credit Card Fraud Detection with Artificial Immune System [13] [26], CARDWATCH: A Neural Network Based Database Mining System for Credit Card Fraud Detection [18] which is bases upon data mining approach [17] and neural network models, the Bayesian Belief Networks [25] which is based upon artificial intelligence and reasoning under uncertainty will counter frauds in credit cards and also used in intrusion detection [26], case-based reasoning for credit card fraud detection [29], Adaptive Fraud Detection which is based on Data Mining and Knowledge Discovery [27], Real-time credit card fraud using computational intelligence [28], and Credit card fraud detection using self-organizing maps [30]. Most of the credit card fraud detection systems mentioned above are based on artificial intelligence, Meta learning and pattern matching.

This paper compares and analyzes some of the good techniques that have been used in detecting credit card fraud. It focuses on credit card fraud detection methods like Fusion of Dempster Shafer and Bayesian learning [2][5][12][15][25], Hidden Markov Model [3], Artificial neural networks and Bayesian Learning approach [5][25],BLAST and SSAHA Hybridization[1][6][11][14][24], Fuzzy Darwinian System[4]. Section II gives an overview about those techniques. Section III presents a comparative survey of those techniques and section IV summarizes the fraud detection techniques.

A fusion approach using Dempster-Shafer theory and Bayesian learning

FDS of Dempster-Shafer theory and Bayesian learning

Dempster-Shafer theory and Bayesian learning is a hybrid approach for credit card fraud detection [2][5][12][15] which combines evidences from current as well as past behavior. It is well known that every cardholder has a certain type of shopping behavior, which establishes an activity profile for them. This detection system learns the behavior of users dynamically so as to minimize its own loss. Thus, there is a need for developing fraud detection systems which can integrate multiple evidences including patterns of genuine cardholders as well as that of fraudsters. This paper develops a fraud detection system using information fusion and Bayesian learning of so as to counter credit card fraud. Number of rules is used to analyze the deviation of each incoming transaction from the normal profile of the cardholder by assigning initial beliefs to it.

The FDS system consists of four components, namely, rule-based filter, Dempster-Shafer adder, transaction history database and Bayesian learner. This system combines multiple evidences including patterns of genuine cardholders as well as that of fraudsters. In the rule-based component, the suspicion level of each incoming transaction based on the extent of its deviation from good pattern is determined. Dempster-Shafer’s theory is used to combine multiple such evidences and an initial belief is computed. Then the initial belief values are combined to obtain an overall belief by applying Dempster- Shafer theory.

Fig. 1. Block diagram of the proposed fraud detection system

The transaction is classified as suspicious or suspicious depending on this initial belief. Once a transaction is found to be suspicious, belief is further strengthened or weakened according to its similarity with fraudulent or genuine transaction history using Bayesian learning. Thus the fusion approach using Dempster-Shafer theory and Bayesian learning has high Accuracy and high Processing Speed. It improves detection rate and reduces false alarms and also it is applicable in E-Commerce. But it is highly expensive and its processing Speed is low. It is not applicable in other transactions.

BLAST-SSAHA Hybridization for Credit Card Fraud Detection

BLAST-SSAHA in credit card fraud detection

The Hybridization of BLAST and SSAHA algorithm [1][6][14] is refereed as BLAH-FDS algorithm. BLAH-FDS is a two-stage sequence alignment algorithm in which a profile analyzer (PA) determines the similarity of an incoming sequence of transactions on a given credit card with the genuine cardholder’s past spending sequences. The unusual transactions traced by the profile analyzer are passed to a deviation analyzer (DA) for possible alignment with past fraudulent behavior. The final decision about the nature of a transaction is taken on the basis of the observations by these two analyzers.

Sequence Alignment

Sequence alignment becomes an efficient technique for analyzing the spending behavior of customers. Sequence alignment is quite commonly used in bioinformatics for finding similarity between genome sequences. It is broadly classified as local alignment and global alignment. Local alignment method finds related regions within sequences having significant similarity. Global alignment is an arrangement of sequences in which all the elements in the given sequences participate in the alignment process. The fraudsters are not expected to be fully familiar with the genuine cardholder’s purchase behavior. In credit card transaction processing, spending sequence containing information about the transaction amount, time, etc, is available to the card issuing bank. Any deviation from the existing sequences can be computed efficiently using sequence alignment.

BLAST-SSAHA Hybridization

When a transaction is carried out, the incoming sequence is merged into two sequences time-amount sequence TA. The TA is aligned with the sequences related to the credit card in CPD. This alignment process is done using BLAST. SSAHA algorithm [9] is used to improve the speed the alignment process. If TA contains genuine transaction, then it would align well with the sequences in CPD. If there is any fraudulent transactions in TP, mismatches can occur in the alignment process. This mismatch produces a deviated sequence D which is aligned with FHD. A high similarity between deviated sequence D and FHD confirms the presence of fraudulent transactions. PA evaluates a Profile score (PS) according to the similarity between TA and CPD. DA evaluates a deviation score (DS) according to the similarity between D and FHD. The FDM finally raises an alarm if the total score (PS – DS) is below the alarm threshold (AT).

Fig. 2. Architecture of BLAST and SSAHA Fraud Detection System

The performance of BLAHFDS is good and it results in high accuracy. At the same time, the processing speed is fast enough to enable on-line detection of credit card fraud. It Counter frauds in telecommunication and banking fraud detection. But it does not detect cloning of credit cards

Credit Card Fraud Detection using Hidden Markov Model

Hidden Markov Model

A Hidden Markov Model is a double embedded stochastic process with used to model much more complicated stochastic processes as compared to a traditional Markov model. Hidden Markov Model based applications are common in various areas such as speech recognition, bioinformatics and genomics. HMM is used to model human behavior. Once human behavior is correctly modeled, any detected deviation is a cause for concern since an attacker is not expected to have behavior similar to the genuine user. If an incoming credit card transaction is not accepted by the trained Hidden Markov Model with sufficiently high probability, it is considered to be fraudulent transactions.

Use Of HMM For Credit Card Fraud Detection

Fig. 3. Process Flow of the Proposed FDS

A Hidden Markov Model [3] is initially trained with the normal behavior of a cardholder. Each incoming transaction is submitted to the FDS for verification. FDS receives the card details and the value of purchase to verify whether the transaction is genuine or not. The types of goods that are bought in that transaction are not known to the FDS. It tries to find any anomaly in the transaction based on the spending profile of the cardholder, shipping address and billing address, etc. If the FDS confirms the transaction to be malicious, it raises an alarm and the issuing bank declines the transaction. The concerned cardholder may then be contacted and alerted about the possibility that the card is compromised.

HMM never check the original user as it maintains a log. The log which is maintained will also be a proof for the bank for the transaction made. HMM reduces the tedious work of an employee in bank since it maintains a log. HMM produces high false alarm as well as high false positive.

Fuzzy Darwinian Detection of Credit Card Fraud

The Evolutionary-Fuzzy System

Looking at credit card transactions alone, with millions of purchases every month, it is simply not humanly possible to check every one and when many purchases are made with stolen credit cards, this inevitably results in losses of significant sums. Fuzzy Darwinian Detection system [4] uses genetic programming to evolve fuzzy logic rules capable of classifying credit card transactions into “suspicious” and “non-suspicious” classes. It describes the use of an evolutionary-fuzzy system capable of classifying suspicious and non-suspicious credit card transactions.The system comprises of a Genetic Programming (GP) search algorithm and a fuzzy expert system.

Data is provided to the FDS system. The system first clusters the data into three groups namely low, medium and high. The GPThe genotypes and phenotypes of the GP System consist of rules which match the incoming sequence with the past sequence. Genetic Programming is used to evolve a series of variable-length fuzzy rules which characterize the differences between classes of data held in a database.

Fig. 4. Block diagram of the Evolutionary-fuzzy system

The system is being developed with the specific aim of insurance-fraud detection which involves the challenging task of classifying data into the categories: “safe” and “suspicious”. When the customer’s payment is not overdue or the number of overdue payment is less than three months, the transaction is considered as “non-suspicious”, otherwise it is considered as “suspicious”.

The Fuzzy Darwinian detects suspicious and non -suspicious data and it easily detects stolen credit card Frauds. The complete system is capable of attaining good accuracy and intelligibility levels for real data. It has very high accuracy and produces a low false alarm, but it is not applicable in online transactions and it is highly expensive. The processing speed of the system is low.

Credit Card Fraud Detection Using Bayesian and Neural Networks

The credit card fraud detection using Bayesian and Neural Networks are automatic credit card fraud detection system by means of machine learning approach. This system identifies and detects the fraudulent behavior in credit card transactions. These two machine learning approaches are appropriate for reasoning under uncertainty.

An artificial neural network [5][7][9][10][16] consists of an interconnected group of artificial neurons and processes information using a connectionist approach to computation. They are usually used to model complex relationships between inputs and outputs or to find patterns in data. It is used in applications, such as Pattern recognition or data classification, through a learning process. The most commonly used neural networks for pattern classification is the feed-forward network. A feed forward neural network is an artificial neural network where connections between the units do not form a directed cycle. In this network, the signals are propagated in forward as well as in backward direction. Perceptrons can be trained by a simple learning algorithm. It consist of three layers namely input, hidden and output layers. The incoming sequence of transactions passes from input layer through hidden layer to the output layer. This is known as forward propagation. The ANN consists of training data which is compared with the incoming sequence of transactions .The neural network is initially trained with the normal behavior of a cardholder. The suspicious transactions are then propagated backwards through the neural network and classify the suspicious and non-suspicious transactions.

Bayesian networks are also known as belief networks and it is a type of artificial intelligence programming that uses a variety of methods, including machine learning algorithms and data mining, to create layers of data, or belief. Bayesian learning combines evidences from current as well as past behavior using supervised learning. Number of rules is used to analyze the deviation of each incoming transaction from the normal profile of the cardholder by assigning initial beliefs to it. By using supervised learning, Bayesian networks are able to process data as needed, without experimentation. Bayesian belief networks are very effective for modeling situations where some information is already known and incoming data is uncertain or partially unavailable. This information or belief is used for pattern identification and data classification.

A neural network learns and does not need to be reprogrammed. It can be implemented in any application without any problem. Its processing speed is higher than BNN. Neural network needs training to operate and requires high processing time for large neural networks. Bayesian networks are supervised algorithms and they provide a good accuracy, but it needs training of data to operate and requires a high processing speed. The accuracy in fraud detection of ANN is low compared to BNN.

Comparison of various Fraud Detection Systems

Parameters Used For Comparison

The Parameters used for comparison of various Fraud Detection Systems are Accuracy, Fraud Detection Rate in terms of True Positive and false positive, cost and training required, Supervised Learning. The comparison performed is shown in Table 1.

Accuracy: It represents the fraction of total number of transactions (both genuine and fraudulent) that have been detected correctly.

Method: It describes the methodology used to counter the credit card fraud. The various efficient methods like sequence alignment, machine learning, neural networks, artificial intelligence, fuzzy logic are used to detect and counter frauds in credit card transactions.

True Positive (TP): It represents the fraction of fraudulent transactions correctly identified as fraudulent and genuine transactions correctly identified as genuine.

False Positive (FP): It represents fraction of genuine transactions identified as fraudulent and fraudulent transactions identified as genuine.

Training data: It consists of a set of training examples. The fraud detection systems are initially trained with the normal behavior of a cardholder.

Supervised Learning: It is the machine learning task of inferring a function from supervised training data.

Comparison Results

The Comparison table was prepared in order to compare various Fraud Detection mechanisms that were used in identifying various credit card frauds. All the techniques of credit card fraud detection described in the table 1 have its own strengths and weaknesses.

Results show that the fraud detection systems such as Fuzzy Darwinian Detection, Dempster Shafer and Bayesian theory have very high accuracy in terms of TP and FP. At the same time, the processing speed is fast enough to enable on-line detection of credit card fraud in case of BLAH-FDS and ANN. BLAST-SSAHA hybridization approach can be effectively used to counter frauds in telecommunication and banking industry. The Fraud detection rate of Fuzzy Darwinian detection system in terms of true positive values is higher than other methods. The HMM is semi-supervised but it shows high false alarm. BLAHFDS takes less than 50 ms for sequence alignment and it is inexpensive than others. The Neural Networks, Bayesian Belief Networks are artificial intelligence based systems. Dempster Shafer and Bayesian theory is based on Machine Learning approach. The Fuzzy Darwinian system is based on genetic programming and fuzzy logic. The Artificial Neural Networks and Bayesian Networks are used to detect cellular phone fraud, Calling card fraud, Computer Network Intrusion. The above all fraud detection systems are scalable for handling large volumes of transactions.

Table 1 Comparison of various fraud detection methods

Parameter

Fusion of Dempster-Shafer

theory and

Bayesian learning

Hybridization

of BLAST

and SSAHA

HMM

Artificial Neural Networks

and

Bayesian Neural Networks

ANN

BNN

Study

Amlan Kundu, Suvasini Panigrahi, Shamik Sural and Arun K. Majumdar (2009)

Abhinav Srivastava, Amlan Kundu, Shamik Sural, Arun K. Majumdar (2008)

Choi Sam Maes,

Karl Tuyls ,

Bram Vanschoenwinkel,

Bernard Manderick

(1993)

Method

Machine Learning

Sequence Alignment

Hidden Markov

Model

Artificial Intelligence, Machine Learning

Fraud

Detection

TP%

98%

86%

70%

74%

FP%

10%

20%

10%

Processing Speed

Medium

Very High

High

Low

Training required

Yes

Supervised Learning

Supervised

Unsupervised

Semi-supervised

Supervised

Cost

Implementation is expensive

Inexpensive

Quite expensive

Expensive

Accuracy

High

Medium

Research issues addressed

Intrusion detection in many database applications Applicable in

E-Commerce

Applicable in telecommunication and banking fraud detection

Online detection, cost is inexpensive

Applicable in online detection of credit card fraud.

No need to check the original user as it maintains a log

Cellular phone fraud, Calling card fraud, Computer Network Intrusion Applicable in E-Commerce

Research Challenges

Processing speed is very low

Cannot detect cloning of credit card fraud

High false alarm,

False Positive is high

Needs training to operate and requires high processing time for large neural networks and BNN

Conclusion

Efficient credit card fraud detection system is an utmost requirement for any card issuing bank. Credit card fraud detection has drawn quite a lot of interest from the research community and a number of techniques have been proposed to counter credit fraud. The Fuzzy Darwinian fraud detection systems improve the system accuracy. Since The Fraud detection rate of Fuzzy Darwinian fraud detection systems in terms of true positive is 100% and shows good results in detecting fraudulent transactions. The neural network based CARDWATCH shows good accuracy in fraud detection and Processing Speed is also high, but it is limited to one-network per customer. The Fraud detection rate of Hidden Markov model is very low compare to other methods. The hybridized algorithm named BLAH-FDS identifies and detects fraudulent transactions using sequence alignment tool. The processing speed of BLAST-SSAHA is fast enough to enable on-line detection of credit card fraud. BLAH-FDS can be effectively used to counter frauds in other domains such as telecommunication and banking fraud detection. The ANN and BNN are used to detect cellular phone fraud, Network Intrusion. All the techniques of credit card fraud detection discussed in this survey paper have its own strengths and weaknesses. Such a survey will enable us to build a hybrid approach for identifying fraudulent credit card transactions.

Order Now