Special Features:
FRAUD DETECTION FITS GRID TECHNOLOGY
By
Kico Borrás Palmer, GridSystems
Thomas arrives
home after a stressing working day. While he opens the black
metal door and
hangs his jacket, realizes that his wife and children have
not arrived yet, he
walks towards the fridge and takes the cold beer he
deserves.
Then he
goes out to check the mailbox and realizes that the monthly summary of
his
credit card has arrived. When he reads the final amount, his heart starts
beating quicker. Much quicker. He has never been in a Casino, but someone has
used his card in one, more than 300 Miles away and has spent a huge amount
of
money.
This is just an example of the fraud risk we are all exposed to
suffer and
that, unfortunately, is getting more usual day to day. Fraud is a
plague
that costs millions of Euros per year both to particulars and
companies. In
the last times some cases have appeared in the media and have
created a
distrust sense against both big companies and even the
administration.
This is especially serious in the USA, where the
problem is enormous in their
Sanitary System, where according to the USA
Countable Headquarters, there have
received 10 % of transactions that include
some type of fraud (charges for not
provided services, adulterated amounts,
etc).
In the world of virtual business, where the users feel even more
unprotected
due to the lack of efficient safety mechanisms and
standardization, the
uncertainty is still bigger.
The existing fear is
reasonable when having to give the number of a credit
card for an on-line
transaction and surely this one is the major obstacle that
e-commerce is
finding before its definitive and irrevocable implantation in
our lives.
The worse thing is that all those doubts are well-founded: some reports
indicate that the fraud in electronic transactions affects 1 % of them, 10
times more than in the physical world (in which Visa has estimated them in
0.06 %). The losses produced by these actions concern not only the credit
cards companies but also the banks that emit them and the companies that
accept the plastic money.
This way, it seems that it is becoming much
easier to commit a crime in the
new Internet economy than it previously was.
Now the delinquents can commit
their actions from their own place, without the
need to physically appear at
the place of the crime, which in this case is not
other one than the Net. The
money, our bank accounts and even ourselves have a
virtual representation in
the Net that is not physical any more, but just bits
of information.
The safety mechanisms that are designed to fight
against them will also need
to be virtual. Nowadays the criminals dedicate
more time to find new ways to
break the security systems than the companies to
find mechanisms to protect
themselves and it is evident that this must change
in their own benefit.
For all that, it is a must that the companies
dedicate enough resources to
deploy efficient protection mechanisms. Lucky,
the solution already exists and
is called Data Mining. According to Ralph
Kimball, co-author of the book "The
Data Webhouse Toolkit: building the
Web-enabled Data Warehouse" the
identification of fraud is one of the key
applications of these technologies.
The Data Mining combines data
analysis techniques with high efficiency
technologies to obtain a knowledge
that the companies can use to discover
hidden information in the data stored
in their huge computer disks. All the
information generated in the computers
that manage the business (behavior
patterns, web pages that are accessed,
transactions performed), can be used to
draw a sketch of the normal, not
fraudulent behavior.
Pattern samples that can be recognized are: an
infrequent number of refund
requests for accidents by the same person, strange
relations like different
companies with the same direction or a sudden fever
for playing in Casinos.
Even if Data Mining exists to fight fraud, big
companies executives will
claim that to get a quick and efficient detection,
huge computers are needed,
since the quantity of information to be analyzed to
discover suspicious
behaviors is enormous and the computational power
necessary to detect it must
be sufficient.
The challenge is that to
discover fraud in so much information is like looking
for a needle in a
haystack. This is not a good excuse to face such an
important problem and
besides, the companies already have enough CPU power to
prevent these attacks,
though only those that are already using grid
technology are efficient in this
aspect.
Fortunately, Data Mining is one of the fields better adapted
to be distributed
by grid technologies.
At GridSystems we have a
funding from the Ministry of Science and Technology
to complete a study on
Data Mining and our experts can not only distribute the
calculation of the
different types of Data Mining techniques (Genetic
Algorithms, for example)
but they have also studied the Data Mining focused to
other fields as for
example the credit risk.
Besides, our InnerGrid product is easily
configurable with specific Connectors
to cooperate with third party tools to
distribute and accelerate their
calculations.
We have the solution,
and it's only necessary to put it into practice. What
are we waiting for?
Kico Borrás Palmer,
GridSystems Consultant.
|
