A REVIEW OF VARIOUS CREDIT CARD FRAUD DETECTION TECHNIQUES - PDF

Please download to get full document.

View again

of 5
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Information Report
Category:

Presentations

Published:

Views: 11 | Pages: 5

Extension: PDF | Download: 0

Share
Related documents
Description
A REVIEW OF VARIOUS CREDIT CARD FRAUD DETECTION TECHNIQUES Shraddha Ramesh Bhagwat ME Computer Science Yadavrao Tasgaonkar Institute of Engineering & Technology Bhivpuri Road, Karjat Prof. Vaishali Londhe
Transcript
A REVIEW OF VARIOUS CREDIT CARD FRAUD DETECTION TECHNIQUES Shraddha Ramesh Bhagwat ME Computer Science Yadavrao Tasgaonkar Institute of Engineering & Technology Bhivpuri Road, Karjat Prof. Vaishali Londhe (H.O.D) Dept. of Computer Science Yadavrao Tasgaonkar Institute of Engineering & Technology Bhivpuri Road,Karjat Abstract -In present scenario when the term fraud comes into a discussion, credit card fraud clicks to mind so far. With the great increase in credit card transactions, credit card fraud has increasing excessively in recent years. Fraud detection includes monitoring of the spending behavior of users/ customers in order to determination, detection, or avoidance of undesirable behavior. As credit card becomes the most prevailing mode of payment for both online as well as regular purchase, fraud relate with it are also accelerating. Fraud detection is concerned with not only capturing the fraudulent events, but also capturing of such activities as quickly as possible. The use of credit cards is common in modern day society. Fraud is a millions dollar business and it is rising every year. Fraud presents significant cost to our economy worldwide. Modern techniques based on Data mining, Machine learning, Sequence Alignment, Fuzzy Logic, Genetic Programming, Artificial Intelligence etc., has been introduced for detecting credit card fraudulent transactions. This paper shows how data mining techniques can be combined successfully to obtain a high fraud coverage combined with a low or high false alarm rate. [6] Keywords decision tree; fuzzy Darwinian system; genetic algorithm; artificial neural network;hidden markov model I. INTRODUCTION Financialfraud can be defined asan intentional act ofdeception. Intention could be anything like getting profit in terms of money, land etc. There are several frauds come under financial fraud like bank fraud, insurance fraud, securities and commodities fraud and other related financial fraud like massmarketing and corporate fraud [1]. Credit card fraud is a partof bank fraud. Credit card fraud can be defined as whensomeone uses credit card of another one for personal reasonswhile the owner of the card and the card issuer are not aware of the fact that the card is being used. Further, the individual using the card has no connection with the cardholder or issuer, and has no intention of either contacting the owner of the cardor making repayments for the purchases made [6]. There areseveral ways by which credit card fraud can take place for example counterfeiting which is related to online transaction in which customer and merchant are in different physical location and hence verification of signature cannot be done so it may lead to an unusual transaction because merchant does not know that whether the customer providing the credit card information is indeed the authorized cardholder or anfraudster. Another example is phishing which has been very prevalent method to steal information of a credit card holder that means fraudsters keep sending spam s to credit card holders by putting a bank link to make their mails look like a real bank mail and ask them to provide their information by giving one simple reason like database crash so they need information for recovery purpose. After getting the information fraudsters use it for buying some stuff. Likewise so many other techniques are there by which credit card fraud takes place. [11] Themain aims are, firstly, to identify the different types of credit card fraud, and, secondly, to review alternative techniques that have been used in fraud detection. The focus here is in Europe, and so ethical issues arising from other cultures are not taken into accountbut for a discussion of these the reader is referred to Chepaitis (1997) and Gichure (2000). Indeed, transaction products, including credit cards, are the most vulnerable to fraud[1][3]. In recent years so many countrieshave been affected by credit card frauds due to which not only common people but also financial institution and corporate sectors are getting affected by losing their money. A similar statistics of annual global fraud losses is given in Table I. It can be clearly observed that credit card frauds have been increasing year by year. Hence credit card frauds need to bedetected at reasonable time so that fraudsters could beprevented from committing illegal activities ISSN: Page 187 II. TABLE I Year Amount (in Billions) 2000 $ $ $ $ $ $ $ $ $ $ $ $ $11.2 ANNUAL GLOBAL FRAUD LOSSES (CREDIT & DEBIT CARDS) LITERATURE STUDY OF CREDIT CARD FRAUD Credit card fraud is an important and interesting work of research technology. Several techniques has been develop for credit card fraud detecting system in online transaction, such techniques base on artificial intelligence, fuzzy logic, data mining, machine learning, genetic algorithm, decision tree, Bayesian network, neural network, clustering algorithm, etc., that evaluate of various credit card fraud transaction. Ghosh and Reilly [2] have proposed a neural network method to detect credit card fraud transaction. They have built a detection system, which is trained on a large sample of labeled credit card account transactions. These sample contain example fraud cases due to lost cards, stealing cards, application fraud, stolen card details, counterfeit fraud etc. Kokkinaki and other have proposed the technics of decision tree. This technics of decision tree are simple and easy to the implementation, decision trees is reduces misclassification of incoming transaction of data, but this is not for usedynamically adaptive of online transaction. [13] Meas, Suggest of fraud detection technics using the bayesian network, in this technics, improving the fraud detecting byremoving highly correlated attribute, ANN was found thecredit card fraud predication faster of the testing phase, atusing transaction profile. Bayesian algorithm is performedbetter result of fraud detection only on neural network. [6][9] Chan and Stalfo, have proposed the a technics of multiclassifier meta learning issues of credit card transaction, itdetecting the fraud detection 46 % improving of overall fraud. Kim, method improving number fraud detection classifier and compare only on the neural network by using the unsupervised algorithm of data mining. Bolton and Hand et al. [12] it has proposed credit card detection using unsupervised method by frequency of transactions and observing abnormal spending behavior. HiddenMarkov Model is one of the best methods for observationspending profile generate at the state transaction. HMM isstatically model for best engineering practice. Hidden markovmodel is best for using the FDS (fraud Detection system). III. VARIOUS TECHNIQUES OF CREDIT CARD FRAUD IN DETAILED A. Neural networks: Neural network is defined as a set of interconnected nodes designed to represent functioning of the human brain. Each node has a weighted connection to several other linked nodes in adjacent layers. Single node take input received from linked nodes and use the weights of the connected nodes together with easy function for computation of output values. Neural networks can be created for supervised and/or unsupervised learning. The user specifies the number of hidden layers along with the number of nodes within a specific hidden layer. The output layer of the neural network may contain one or several nodes depending upon the application. Recently, neural network researchers have several associated methods from statistics and numerical analysis into their networks.[9][10] Neural networks can learn and summarizes the internal assumptions of data even without knowledge of the potential data principles in advance. According to Rumelhart, (1986), Neural networks topologies, or architectures, formed by organizing nodes into layers and attach layers of neurons with modified weighted interconnections And it can match its own behavior to the new environment along with the results of formation of evolution capability from present environment to the new possible situation. Statistical methods are sometime unusual in the practice research even though the common advantages of the neural networks in application of credit card fraud detection. On the other side, there are still many disadvantages for the neural networks, such as (1) Difficulty to confirm the structure, (2) Excessive training, (3) Efficiency of training and so on. ISSN: Page 188 B. Decision Tree: After introducing the concept of learning system, decision tree method has been developed, that can deals with continuous data. The decision tree is a table of tree shape with connecting lines to available nodes. Each node is either a branch node followed with more nodes or only one leaf node assigned by classification. With this strategic approach of separating and resolving, decision tree usually detach the complex problem into many simple ones and resolves the sub-problems through repeatedly using, data mining method to discover training various kinds of classifying knowledge by constructing decision tree. The basis of decision tree model is how to construct a decision tree with high precision and small scale.[13] There are many advantages of Decision tree method. 1. High flexibility 2. Good haleness 3. It is explainable, which is also the reason of its varied utilization. Its disadvantages is that, it requires to check each transaction one by one. C. Fuzzy Darwinian System : This technique [12] 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 developed comprises two main elements: a Genetic Programming (GP) search algorithm and a fuzzy expert system. When the data is provided to the FDS system, the system first clusters the data into three groups namely low, medium and high (fuzzy clustering). The 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. 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. For classification of transactions, 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. This system has very high accuracy and produces a low false alarm in comparison with other techniques, but it is highly [1]. The speed of the system islow. D. Hidden Markov Model : A Hidden Markov Model is a double embedded stochastic process which is used to model much more complicated stochastic processes as compared to a traditional Markov model. 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. A Hidden Markov Model [8] is initially trained with the normal behavior of a cardholder. It works on the user spending profiles which can be divided into three types such as 1. Lower profile; 2. Middle profile; and 3. Higher profile For every credit card, the spending profile is different, so it can figure out an inconsistency of user profile and try to find fraudulent transaction. It tries to find any anomaly in the transaction based on the spending profile of the cardholder, shipping address, and billing address, etc. Every user is represented by specific patterns of set which containing information about last 10 transaction using credit card [1][8]. The set of information contains spending profile of card holder, money spent in every transaction, the last purchase time, category of purchase etc. The potential threat for fraud detection will be a deviation from set of patterns. In the process of HMM 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. 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 misused. 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. E. Genetic Algorithm (GA) : Genetic algorithms, inspired from natural evolution were first introduced by Holland (1975). Genetic algorithm (GA) is a search technique used in computing to find exact or approximate solutions to optimization and search problems. GA is used in data mining mainly for variable selection [7] and is mostly coupled with other DM algorithms [1]. And their combination with other techniques has a very good performance. ISSN: Page 189 They have been used in a number of applications in engineering and social science. Recently, they applied for optimization of the parameters of support vector machine for predicting bankruptcy [7], and hybrid with neural net work for detecting credit card fraud with high accuracy [1], and have been used along with Artificial Immune System for reducing a number of false alarm in credit card fraud detection. GA has been used in credit card fraud detection for minimizing the wrongly classified number of transactions. And is easy accessible for computer programming language implementation, thus, make it strong in credit card fraud detection. But this method has high performance and is quite. IV. GA IMPLEMENTATION Genetic algorithms are inspired from Darwin's evolutiontheory and were introduced by Holland [1]. Its basic idea isthat after sufficiently long amount of time has passed weaker members of a species will be destroyed and the fittermembers will survive. The new generations will be produced by the cross-over of two parent members. However, sometimes some random mutations can occur on individuals which in tum increase the diversity in the population. The GA as a solution procedure is depicted in Figure 1. It starts with a number of initial solutions which act as the parents of the current generation. New solutions are generated from these solutions by the cross-over and mutation operators. The less fit members of this generation are eliminated and the fitter members are selected as the parents for the next generation. This procedure is repeated until a pre-specifie number of generations have passed, and the best solution found until then is selected [6], [7]. GA is a parametric solution procedure and it needs to be fine-tuned well for the problem undertaken to get a betterperformance. The list of such parameters and the settings we have taken are discussed below. a) Number of parent solutions: Number of starting solutions which also equal to the number of parents selected for each generation is an important parameter which can influence the convergence speed of the procedure. The population size is determined according to the size of the problem, i.e. bigger population for larger problem [7]. We have taken this to be 50 where three of them are determined as to be the solutions which will generate the maximum number of alerts (MAX), the one that will generate the minimum number of alerts (MIN) and the one currently used in the production (PRD). The remaining 47 parameters are obtained by producing random numbers for all 43 parameters. The produced set of parameters is validated against the scoring logic and the infeasible ones are corrected. b) Number of children: For the ease of implementation we decided to cross-over every possible pair of parents and obtained 1225 children in each generation. c) Cross-over operator: We took the weighted average of the parameter values of the two parent solutions and obtained the child solution. For each generation a random number between zero and one is determined and this number is used as the weight of the first parent in the all cross-over operations in that generation. The weight of the second parent is equal to one minus the determined random number. This type of cross-over operator is not a typical one in GA literature but in our application domain, it is quite suitable. d) Mutation operator: One of the 43 parameters is picked up randomly and its value is changed randomly within its allowable range. e) Cross-over and mutation probabilities: All children are generated by the cross-over operator. Then, one of the Childrensolutions is randomly picked up and mutation Operatoris applied to it. f) Fitness function: As described above, the fitness values of an individual solution is determined as the total amount of savings incurred from fraud losses. g) Selection: The best three members of the generation are automatically selected. To keep having diversity in all generations, the three named solutions, MAX, MIN and PRD are also automatically transferred to the next generation. The remaining 44 solutions are determined by the roulette selection procedure. h) Termination criterion: We decided to run the generations until no improvements are observed for some sufficiently long amount of time V. CONCLUSION AND FUTURE WORK In this paper, we present a comparative study of fraud detection methods based on credit card. The main objective ISSN: Page 190 of this paper is to review methodology of different detection methods based on credit card. We have considered the most important parameter in different methods such as, accuracy, speed and cost. Comparison table was prepared in order to compare various credit card fraud detection mechanisms. All the techniques of credit card fraud detection described in the table1 have its own strengths and weaknesses. We found these result is mentioned in following table from the references that we have mentioned in end As the results show, the fraud detection systems based on Fuzzy Darwinian, has a very high accuracy with 100% true positive but with very low processing speed. In another view, HMM has a fast processing speed with low accuracy. At the same time, the processing speed in decision tree is very fast enough to enable detection of credit card fraud.[5] All these techniques of credit card fraud detection discussed in this survey paper, have its own weaknesses as well as strengths. Thus, this survey enables us to build a hybrid approach for developing some effective algorithms which can perform well for the classification problem with variable misclassification costs and with higher accuracy. [6] Tej Paul Bhatla, Vikram Prabhu & Amit Dua Understanding Credit Card Frauds, [7] Blickle, T., & Thiele, L. (1995). A Comparison of Selection Schemes used in Genetic Algorithms (Vol. 2). Zurich: Swiss Federal Institute. [8] Abhinav Srivastava, Amlan Kundu, Shamik Sural, Arun K. Majumdar. Credit Card Fraud Detection using Hidden Markov Model. IEEE Transactions on dependable and secure computing,volume 5; (2008) (37-48). [9] A. Vellidoa, P.J.G. Lisboaa, J. Vaughan Neural networks in business: a survey of applications. Elsevier, Expert Systems with Applications, (1999). [10] P.K. Chan, W. Fan, A.L. Prodromidis, S.J. Stolfo Distributed Data Mining in Credit Card Fraud Detection. Data Mining; (1999). [11] R.C. Chen, T.S. Chen, C.C. Lin ( A new binary support vector system for increasing detection rate of credit card fraud. International Journal of Pattern Recognition2006). [12] Peter J. Bentley, Jungwon K
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks