2. Analyse and find most
effective way to fight
against fraud
Fraud and his types
Which combination of
methods and actions will
be the most effective?
Purpose
Subject
Research
question
AGENDA
3. LITERATURE REVIEW
“According to the 2008 edition of CyberSource’s Annual Online
Fraud Report, 38% of merchants reported a fraudulent order rate of
1% or more – that’s accepted orders; That means 38% of merchants
are in danger of losing their accounts as the 1% mark is where
merchant account providers start getting toey” - (Ecommerce)
“Data and device security will continue to be a major focus,
particularly for consumers. The more devices we allow our
technologies to connect with the more susceptible we become to
hacks and fraud – the October 2016 DDoS attack on Twitter, Spotify,
PayPal and other major services in the U.S. was a prime example” -
(Computerworld)
5. REGRESSION MODEL
Y = α + β1X1+ β2X2+ β3X3+ β4X4+ β5X5+ β6X6+ ℮
Where,
Y = Fraud Score;
β = coefficient (fraud rate/number of purchases);
α - is the significance level;
X1 - IP range (checking the geographical location of the customer, compliance
between IP cardholder/ online store buyer and IP buyer;
X2 - Email;
X3 - Proxy;
X4 - Fraud level in particular country;
X5 - Distance between IP address of buyer and IP of delivering;
X6 - BIN Bank;
℮ - is a random error;
6. TAXONOMIC ANALYSE
Fraud
Web Network
Web
Advertising
Credit cards
Telecommunic
ation
SubscriptionSuperimposedCredit card
Credit
application
Online OfflineClick inflation Hit shaving
For example: Who is the source? - What is fraud category? - What is the type of fraud?
7. QUALITATIVE METHOD
Adaptive method
Guidance method
Analytical method
use the previous data to
predict normal behavior of
user
analyse real world
examples of fraud
investigation of violations
8. FORMALISATION OF FRAUD SCHEMES (DETECTION OF EVENTS NOT TYPICAL
FOR THIS TECHNOLOGY, DO NOT FIT INTO A CERTAIN PATTERN OR
TEMPLATE SPECIFIC TO THE BUSINESS PROCESS OR CUSTOMER);
TIME ANALYSIS OF CLICK;
DEVICE-INFORMATION ANALYSIS;
CONVERSION TIME;
PROXY TIME;
IP;
USER-AGENT;
MAIN QUALITATIVE PARAMETERS