Fraud Detection in Internet Banking Transactions

Volume 6, Issue 3, March – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
Fraud Detection in Internet Banking Transactions

using Sliding Window Strategy
Abhinandan1,Abhijeet Chauhan2, Divyendu Shekhar3, Narendra Kumar4
1
Computer Science & Engineering, National Institute of Engineering, Mysore
2
3
Information Science & Engineering, National Institute of Engineering
4
Abstract:- Internet Banking(IB) frauds are relatively transactions done by an accountholder and to simultaneously
eaisier for hackers and attackers with malafide overcome the problem of concept drift. Table 1, shows
intentions and so the number of Internet banking frauds basic attributes that are captured when any financial or non
these days are overwhelming. E-commerce platforms financial transaction is done.
and many other online shopping portals have included
Internet Banking as a payment mode, increasing the risk Table 1: Raw attributes of Internet Banking activities
for IB frauds. The primary intent of this research work FEATURE NAME BRIEF DESCRIPTION
is to improvise and develop a unique and new fraud
identification technique for Internet Banking UTR number Unique Identification
Transactions by analysing the past transaction and Number of a transaction
banking details of the customer and deduce the patterns
in the nature of transactions done so as to be able to Account Number Identifier of a user account
detect an anomalous transaction in future. Where IB Total Amount The total Amount
account holders are grouped into different categories transferred
based on their transaction and internet banking Time Time of the transaction
activites. Then we make use of the sliding window IP IP address of the device
protocol or strategy, to assimilate the transactions and from which transaction is
activities done by the customers from different internet done
banking channels so that the patterns and similarities in Marker or Label To specify whether the
the nature and type of the transactions belonging to transaction is legitimate or
different categories or groups can be inferred and fraudulent
extracted respectively.
2.1. Algorithm
Keywords:- Internet Banking Transactions, Sliding Window
Strategy, Concdept Drift;  Firstly, we use a grouping mechanism to categorise the
account holders into different categories or baskets based
I. INTRODUCTION on their transaction volume, i.e., high, medium, and low
using range partitioning technique.
Internet Banking(IB) transactions being the
 Using Sliding-Window protocol and strategy, we classify
cornerstone of cashless economy, there is a spike in the
the Internet Banking activities into respective categories,
volume of such transactions. The statistics available with
i.e., extract features from the sliding window to find the
National Crime Records Bureau says that in excess of three
user’s behavioural patterns. Features like Geo-location
thousand incidences of internet banking frauds were
(IP range), max amount, min amount of transaction, in
recorded in 2018. Internet banking data is highly
the window and even the time taken in the process.
disproportionate because there will often be more number of
legitimate transactions when compared with fraudulent
Algorithm 1: Algorithm to classify and group transaction
transactions. However, despite all adversaries, there are
details and to deducecustomerbehaviors using sliding
umpteen strategies and techniques to overcome this
window technique.
problem.
II. PROPOSED METHOD

Input: account details of the customer, a sequence of
transactions s and window size z.
Internet Banking (IB) activities including financial and
Output: Classified transaction details and behaviors of
non financial transactions are mostly unfamiliar when
account holder legitimate or fraudulent.
studied and compared to past records of the customer. The
main aim of our research work is to device a method or
algorithm to predict the legitimacy of the new IB
IJISRT21MAR446 www.ijisrt.com 1206

Volume 6, Issue 3, March – 2021 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
runtime updations and changes in parameters and features
lead the system or our algorithm to adapt to new
accountholder's behavioural patterns punctualy and timely.
d: length of S
Legitimate= []; Going forward, whenever a new transaction is being
Fraudulent= []; done from the account of a particular accountholder, we can
For g in range 0 to d-z+1: see if the behavior of the transaction matches the behavior of
S: []; the accountholder. If the behavior is new or doubtful, we
may use multifactor authentication to validate the legitimacy
/* sliding window attributes*/
of the transaction before its execution and hense prevent
For j in range g+z-1: potential frauds.
/*Add the transaction to window */
S=S+sj (id); REFERENCES
End
/* behavior extraction related to amount of [1]. http://www.rbi.org.in/Circular/CreditCard
[2]. https://www.kaggle.com/mlg-ulb/creditcardfraud
the transaction*/ [3]. https://www.kaggle.com/uciml/default-of-credit-card-
bi1=MAX_AMT(Sg); clients-dataset
bi2=MIN_AMT(Sg); [4]. https://www.npci.org.in
bi3=AVG_AMT(Sg); [5]. http://www.niti.gov.in
bi4=AMT(Sg); [6]. https://finmin.nic.in
For j in range i+z-1:
/* Time taken for the transaction*/
xi= Time(sj)-Time(sj-1)
End
Pi= (bg1, bg2,bg3,bg4,bg5,);
Q= LABEL(Sg);
/* grouping a transaction into fraudulent
or legitimate */
if Qg=0 then
Legitimate =
(Legitimate)Union(Pg);
Else
Fraudulent
=(Fraudulent)Union(Pg);
End
 Anytime a new activity or transaction is fed into the

sliding window, the previous ones are removed and step-
2 is re-executed for each category and group of
transactions.
 After generating the score, we turn on an assessment
system, wherein the current activities and updated scores
are fed back to the system for subsequent study and
comparisions.
III. CONCLUSION
In this research work we deviced a new approach and

strategy for fraud identification and prevention in Internet
banking transactions, where customers’ data is grouped on
the basis of their financial and non financial activities in
internet banking channels and we extract behavioural
patterns to dynamically build and incrementally develop a
user profile for each and every accountholder or user. These
IJISRT21MAR446 www.ijisrt.com 1207

Fraud Detection in Internet Banking Transactions

Uploaded by

Document Information

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

Fraud Detection in Internet Banking Transactions

Uploaded by

Copyright:

Available Formats

Volume 6, Issue 3, March – 2021 International Journal of Innovative Science and Research Technology

Fraud Detection in Internet Banking Transactions

II. PROPOSED METHOD

IJISRT21MAR446 www.ijisrt.com 1206

 Anytime a new activity or transaction is fed into the

In this research work we deviced a new approach and

IJISRT21MAR446 www.ijisrt.com 1207

You might also like