Fraud Detection During Money Transaction and Prevention: Surbhi, Dr. Sanjeev Kumar

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

2019 2nd International Conference on Issues and Challenges in Intelligent Computing Techniques (ICICT)

Fraud Detection During Money Transaction and


Prevention
Surbhi1, Dr. Sanjeev Kumar2
1,2
KIET Groups of Institutions, Ghaziabad, India

Abstract—With the increased use of digital money, online August, 2019 states that a doctor was fooled to pay 2.62 lakh
financial frauds are observed to be one of the most common cyber rupees while shopping online. He purchased a laundry bag of
crimes. Credit/Debit card frauds are among the easy cyber crimes worth rupees 399. He got a call that he has got the offer in which
targeted by fraudster. In this paper all possible online frauds of a laptop is given to him as a gift. Later on he got message to pay
present era and likely to be occur in future have been discussed. GST Rs 5,580 along with some information about him.
Ignorance, greed, availability of technical tools, malafide
intentions, online and offline security breaches are some reasons 3. Insurance fraud: False insurance claims for automobile
behind these financial cyber crimes. Most of the fraud detections damages, health care expenses and homeowners or rentals
techniques have been studied with their merits and constraints. insurance etc .Example: A report of 9 july 2019 states that 10
Proposal of two techniques to detect and understand dynamics of men arrested for insurance fraud worth crores. This group
evolution of fraudsters with improvement of security systems and includes a doctor, a lawyer and which targets the ill patients.
another technique of analysis of time series data of random They first get them insurance cover. Doctor do long treatment
variables describing characteristics of online frauds have been and make money. In case a person died during treatment then
suggested to use in future scope. certified his death as unnatural so that they can claim for
insurance money.
Keywords—online financial fraud, detection and prevention,
time series data. 4. Work from home scams: Criminal’s offers work from home
oppurnity to the victims and promise the people to payoff big
I. INTRODUCTION amount. Fraudsters offer simple task to the victims and trap them
Fraud is illegal action performed for a specific motive like to in their scheme by offering different courses so that victims can
gain money, goods, services and sometime fame etc. fraud is earn more money by joining such courses and ask victims for
present in our society from ancient time. But the ways of doing their bank details.
frauds is also enhanced with the time. Here are some definitions 5. Credit card or debit card frauds: unauthorized use of
of fraud by different scholars: Fraud is intentional deception to someone’s credit card or debit card or credit or debit card
secure unfair or unlawful gain, or to deprive a victim of a legal information to buy something. Example: In recent news of 9
right. Fraud may violate civil laws, criminal laws. There may be august, 2019 the JCP Atul katiyar sitting at police HQ, Delhi
chances of no loss of money, property or legal rights but still looses Rs 28,000 to credit card fraud. He got a fake bank
matter of another civil or criminal wrong. Fraud is a deliberately message that he had got some rewards point on his credit card.
deceitful activity in order to gain an advantage or generate a To redeem them he has to fill some information. He did so and
profit. In corporate fraud is cheating in business or financial loses his money.
markets. A person or thing intended to deceive others, typically
by unjustifiable claiming or being credited with We mainly focus on the credit card frauds.“A 2016
accomplishments or qualities. Fraud is a crime getting money or Iovation/Aite Group study on EMV’s projected impact on
financial benefits by lying or by using tricks.Fraud is something financial fraud reports that credit card fraud losses may climb to
or someone that deceives people in a way that is illegal or as much as $10 billion in the united states alone by
dishonest. 2020”.According to 2016 data released by ACI worldwide &
financial industry consultant Aite Group, nearly 1 in 3
Types of frauds consumers globally have been victimized by credit card fraud in
There are various types of frauds that can be seen or read in past 5 years.
newspaper. Following are some of them: There are various shapes and forms of credit card frauds.
1. Cheque frauds : when someone give cheque to pay money to Some frauds are done to get money or some are for purchasing
other person by knowing that he has not enough money in his goods for free. Credit card fraud is not just single action but have
account to pay it or when someone steals other individual’s different forms. Below is description of different forms of credit
cheque. Example: A recent report of 30 July 2019 states that two card frauds:1.Application fraud: This type of fraud occurs when
startup fraudsters cheque mate Lucknow businessmen. 2 persons other people apply for credit or new credit card in your name.
brought laptops for wholesale and did payment by cheque of Criminals first steal your information like supporting documents
posterior date which later on declared bounced by the bank. by false calls. 2.Electronic or manual credit card imprints: This
happens when someone skims the information that is placed on
2. Internet sales: Selling fake items online and take payment magnetic strip and then decodes this information to do fraud
with intension of not giving delivery. Example: A report of 29 transaction.3. Card not present fraud (CNP): This type of

978-1-7281-1772-0 ©2019 IEEE


Authorized licensed use limited to: National Inst of Training & Indust Eng - Mumbai. Downloaded on August 12,2021 at 08:06:58 UTC from IEEE Xplore. Restrictions apply.
fraud is done when criminals don’t have your card but have operational analysis. In strategic analysis, exploration of data is
some details like account no, expiry date etc.4.Lost and stolen done on its operating environment to make strategy while in
fraud cards: In this transactions are made online by using your operating analysis of data is exploring about current information
missing or stolen cards details. 5.Card id theft: when someone to be processed. In the end he suggests that by monitoring the
gets your credit card details and this information can be used for system we can avoid fraud to enter.[4]Varsha Nemade(2018)
getting new credit card. Example: A report of 30 July, 2019 focused on the predictive modeling for credit card fraud
states that a Mumbai senior citizen looses 2.5 Lakh rupees by detection using data analytics. He used 3 approaches after data
filling online form. She got a fake call from fraudsters that she analysis i.e. logistic regression, random forest and random forest
has to update her details by filling up online form. If she doesn’t decision tree. He found that random forest work best among all.
do so her card will be blocked. She did as it and loses her But there is a problem of overfitting in memory as data
money.6.Mail receipt card fraud: It is also called never received increases. [5] Devi meenakshi(2019) applied random forest for
issues or intercepts fraud. When you are expecting a new card detecting credit card frauds. She found that random forest also
or replacement one but criminal intercept that card.7.Doctored suffer from imbalance problem of data. [6]Alejandro correa
cards: It is a card in which a strong magnet has erased its metallic Bahnsen(2013) predicted a cost sensitive method based on
stripe. Criminals do this to make changes so that they match with Bayes minimum risk. He first optimize the threshold and then
it valid cards. 8.Fake cards: criminals make fake credit card applied Bayes minimum risk classifier to predict the cost
although it requires a lot of skills. 9. Assumed identity: In this sensitivity and then he compared it with state of art algorithms
type of fraud, criminals use assumed information like temporary which gave result by improving 23% cost. [7]Sanjeev jha(2011)
address with false name. Although now a day’s banks verify this proposed a comparative study for data mining for credit card
by account reference to check genuinity of the card fraud. He compared 2 approaches of data mining i.e. support
holder.10.Account Takeover: This type of fraud is most vector machine and random forest with logistic regression. He
commonly done by online. Criminals first get your information found that random forest work well. [8] Joseph pun (2012)
like supporting documents, card details and then contact to your suggested improving credit card fraud detection using Meta
credit card company to get new card and send it to changed or classification strategy. He did analysis of different paired
modified fake address. approaches i.e. Naïve Bayes, Decision tree, k-nearest neighbor,
SVM for detecting fraud. The best out of them are selected by
II. LITERATURE SURVEY the Meta classifier to detect the fraud. This Meta classifier is
To avoid any fraud, there are two stages first one is fraud then combined with ranking method. Its prediction is then
prevention and second one is fraud detection. In fraud compared with neural network prediction. [9]C.Withrow (2009)
prevention we try to restrict ourselves from being victims of any applied transaction aggregation as a strategy for credit card fraud
fraud by using firewall or any encryption and decryption detection. He predicted that transaction aggregation sometime
method. But in fraud detection if we found that there is a fraud can be useful but not all the time. He found that random forest
then there are some methods which we can use to get overcome work well as compare to SVM, KNN, and logistic regression for
from them. Mohd Aizaini Maarof(2016) discussed issues & classification. [10]Fabrizio carcillo (2017) predict a Scalable
challenges that impact the performance of fraud detection real time fraud finder (SCARFF) which works with tools spark
system. He took 5 areas of frauds i.e. credit card, in detecting credit card fraud detection. [11]Heta naik (2019)
telecommunication, health insurance, automobile insurance and compared the different machine learning algorithms i.e. k-
online auction. He also predict the techniques that are recently nearest neighbor, Naïve bayes, Random tree, Logistic
used in detecting the frauds i.e. SVM, ANN, Decision tree, regression, outlier, Adaboost and J48 for detecting credit card
logistic regression [1].There are various issues that faced by the frauds. He found that Adaboost algorithm perform well as
investigators while detecting credit card fraud. Yann-Ael Le compared to other algorithms.[12]Gowtham sethupathi(2019)
Borgne(2014) focused on 3 issues i.e. unbalancedness, non- applied random forest algorithm to detect the dynamic malware.
stationarity, and assessment. He suggested doing ranking of He classified the benign apps from malicious apps by studying
transactions by their probability of fraud rather than classifying the android apps and its pattern. [13]Krishna kumar tripathi
transaction correctly, as a solution of the problem unbalanced. (2012) did a survey about the various algorithms used in
He urged for non stationary problem to create new model every detecting credit card frauds. He discussed various approaches
time that in turn create a new chunk. For assessment problem, like neural network, Bayesian network, Genetic algorithm, k-
he predicted 3 approaches i.e. static, update & forgetting. In nearest neighbor, SVM, Decision tree, meta learning strategy,
these 3 approaches he found best strategy is forgetting approach artificial immune system. [14]Venkata ratnam ganji (2012)
along with easyEnsembler and daily update. [2]Masoumeh predict the concept of credit card fraud detection by using a data
Zareapoor(2015) compared 4 data mining approaches for credit stream outlier detection algorithm which is based on reverse k-
card fraud detection i.e. Naïve bayes classifier, Support vector nearest neighbors (SODRNN). [15]
machine, K nearest neighbor algorithm and bagging ensemble
classifier based on decision tree. He found that bagging III. TECHNIQUES : OVERVIEW
ensemble classifier based on decision tree work best among all Let’s take a Brief overview of techniques which we have studied
because of its independency of attribute values and its ability to from the above stated.
handle the class imbalance problem. [3] Adrian
Banarescu(2015) suggested that by doing data analysis we can A. Random forest: RF creates a set of models and combines its
detect the fraud and able to prevent from it. He predicts two prediction to predict the label of class for any data at any point.
approaches for data analysis i.e. strategic analysis and It creates each tree on separate bootstrap sample and randomly
selects the attributes from the subset of data as node. Error rate

Authorized licensed use limited to: National Inst of Training & Indust Eng - Mumbai. Downloaded on August 12,2021 at 08:06:58 UTC from IEEE Xplore. Restrictions apply.
depends upon correlation between trees and strength of each E. K Nearest neighbor: It is simple algorithms that stores
tree in the ensemble. In [7] RF performs better when threshold all available cases and classify new cases on the basis of
optimization is applied. It result in increase in recall which similarity measure i.e. distance function. It is based on
means number of fraud detected but precision decreases but F1 instances. Every time when a new instance come it compare
score remain constant. On adjusting estimated probability new one with existing one by using distance metric and
Bayes Minimum is applied on RF it give better result as assign the class to the nearest existing instance. In [3] author
compare to other algorithms. In [2] RF performs better i.e. calculates the nearest point to new incoming transaction and
accuracy obtained is improved by increasing training size. if nearest point is fraudulent then this transaction is fraud.
B. Support vector machine (SVM): SVM is a linear classifier Generally value of k is small. For calculating distance value
that do non linear mapping of data points in input space of between two points there are different ways. Euclidean
problem in high dimensional feature space. It doesn’t work in distance calculation is used for continuous attributes while
straight line that helps it in capturing the much more complex for categorical attributes, simple matching coefficient is
relationship between data points. It performs well with used.
unbalanced data without adding any computational complexity.
In [8] SVM uses kernel representation and margin optimization IV. CONCLUSION
to perform classification. Kernel representation is done through It is prominently well established fact that fraudsters are
the kernel function that is the dot product of two data points in always available in society. Online fraud cases of financial
high dimensional space that help in mapping of data points in transactions are big challenges since with the evolution of
learning the classification. Whereas margin optimization is done security mechanism fraudsters are also evolved and using new
through the classification function that is a hyper plane used to techniques. Lack of awareness, greed and fear are the main
overfit the training data by calculating margin separation parameters behind online cheating of customers so there is
between two classes. There are various kernel function strong demand of smart mechanism to deal with such cases.
depending upon the problem classification i.e. linear function, Proposal of smart software which can track suspicious textual
Gaussian radial basis function, polynomial function. messages and audio conversations along with detection of
C. Naïve Bayes classifier: NVC is a supervised machine malafide software agents, one proposal of development of
algorithm used to predict the class of future instance on the basis intelligent software agent is also very important which may be
of training dataset with known target classes. It is used for self-evolvable and can track source of fraudulent and nullify the
classification problem mainly text classification. It is called effects of these and it may be helpful to nab fraudsters too.
naïve because it makes assumption that the occurrence of each
feature is independent to other feature.
REFERENCES
P(A│B)=(P(A│B)P(A))/P(B)
[1] Aisha Abdallah, Mohd Aizaini Maarof, Anazaida Zainal, “Fruad detection
Where P (A|B) =occurrence of event A given the event B system: A Survey”, ELSEVIER, April 2016.
is true, P(B|A)=probability of occurrence of event B given event [2] Andrea Dal Pozzolo, Olivier Caelen, Yann-Ael Le Borgne,Serge
A is true. It works on Bayesian method of probability to Waterschoot, Gianluca Bomtempi, ”learned lessons in credit catd fraud
calculate the probability of correct class which is particular detection from a practitioner perspective”,ELSEVIER,2014.
attributes in the dataset. It is as follows: Prior probability of X: [3] Masoumeh Zareapoor, Pourya Shamsolmoali, “Application of credit card
fraud detection: Based on Bagging Ensemble Classifier”, ELSEVIER,
Number of X instances/total number of instances. Likelihood of 2015.
W given X: Number of X in vicity of W/total number of X. Final [4] Adrian Banarescu, ”Detecting and preventing fraud with data analytics”,
classification is produced by combining of prior probability and ELSEVIER, 2015.
likelihood to posterior probability. [5] Suraj patil,Varsha Namade, Piyush kumar soni, “Predictive modeling for
credit card fraud detection using data Analystics”, ELSEVIER,2018.
Posterior=(likelihood)×prior/(evidence prior probability)
[6] Devi Meenakshi.B, Janani.B, Gayathri,s, Mrs.Indra.N, “credit card fraud
D. Decision tree: it is a supervised machine learning algorithm. detection using random forest”, IRJET,Page no-6662 to 6666,March,2019
It is widely used tool for classification and regression. It is kind .
of flow chart in which each leaf node contains a class label, each [7] Alejandro correa Bahnsen, Aleksandar Stojanovic, Djamila Aouada and
Bjorn Ottersten, “cost sensitive credit card fraud detection using Bayes
internal node have a test(IF-Else) on attributes whereas branch Minimum Risk”, IEEE, Page no-333 to 338,2013.
holds the outcome of the test. There is no need of prior
[8] Siddhartha Bhattacharyya, Sanjeev Jha, Kurain Tharkunnel, J.Christopher
knowledge to work in any domain and can handle continuous Westland, “Data mining for credit card fraud: A comparative study”,
and categorical variables. In [5] decision tree uses ID3 for ELSEVIER, August, 2010.
constructing tree by considering the entropy of dataset i.e. [9] Joseph Pun, Yuri Lawryshyn, “Improving credit card fraud detection using
amount of uncertainty in n. Split point of tree is also depend on meta classification strategy”, International journel of computer
entropy. It can be calculated as follows: applications, volume 56-no.10, October, 2012.
[10] C.Whitron, D.J.Hand, P.Juszezak,D.Weston,N.M.Adams, “Transaction
H(p1,p2,…pi)=∑_(i=1)^s(pilog(i/pi) ) aggregation as a strategy for credit card fraud detection”, Springer,
August, 2008.
Where p1, p2,….ps are probability of attribute of dataset [11] Fabrizio Carcillo, Andrea Dal Pozzolo, Yann-Ael Le Borgne, Olivier
Now gain function is calculated by subtracting entropy of entire Caelen, Yannis Mazzer, Glnuca Bontempi, “SCARFF: a scalable
dataset to entropy of splitting attribute. The highest value of framework for streaming credit card fraud detection with spark”,
entropy will be selected as root of tree. September, 2012 .

Authorized licensed use limited to: National Inst of Training & Indust Eng - Mumbai. Downloaded on August 12,2021 at 08:06:58 UTC from IEEE Xplore. Restrictions apply.
[12] Heta Naik, Prashasti Kanikar, ”credit card fraud detection based on
machine learning algorithms”, International journel of computer
applications, volume 182-no.44,March 2019.
[13] Gowtham Sethupathi, Swapnil Siddharth, Vikash Kumar, Pratyush Kumar,
Ashwani Yadav, “Malroid: Dynamic Malware detection using Random
Forest Algorithm”, IJITEE, volume-8 Issuse-6, April, 2019.
[14] Krishna Kumar Tripathi, Mahesh A.Pavaskar, “survey on credit card
fraud detection methods”, IJETAE,Volume 2,issue 11,November 2012.
[15] Venkata Ratnam Ganji, Siva Naga Prasad Mannem, ”Credit card fraud
detection using anti-k nearest algorithm”, IJCSE,Vol.4 No.6, June 2012.

Authorized licensed use limited to: National Inst of Training & Indust Eng - Mumbai. Downloaded on August 12,2021 at 08:06:58 UTC from IEEE Xplore. Restrictions apply.

You might also like