Visa, HNC Inc. develop neural network as a weapon to fight fraud

Extracted from "FC NEWSBYTES 1.3", David Geddes <dgeddes@NETCOM.COM> Editor,
	where FC = FutureCulture mailing list <FUTUREC-request@UAFSYSB.UARK.EDU>.    strick
____________________________________________ _ _..........
B Y T E 4:

Visa, HNC Inc. develop neural network as a weapon to fight fraud
SAN FRANCISCO (AUG. 10) PR NEWSWIRE - Visa International and HNC Inc. have
announced a strategic agreement to develop a comprehensive merchant risk
detection system.  The new system will be designed to better control fraud at
the merchant level by determining the risk associated with individual card
transactions.

This agreement continues to support Visa International's active role in
developing effective solutions to the problem of fraud occurring at the point
of sale.  The merchant risk detection system will be available in 1994.

"Visa has combined its core systems capabilities and the premier technology
available -- neural networks -- for fighting credit card fraud," explained
Roger Peirce, Visa International's executive vice president, Delivery Systems.
"HNC, an industry leader in neural network applications and credit card
control services, is a logical partner for Visa," he added.

Michael A. Thiemann, HNC's executive vice president, called the agreement
"another example of our commitment to solving tough business problems through
the application of cutting-edge technologies."

Neural network technology enables a system to predict the probability of fraud
by learning from a large number of past transactions, both legitimate and
fraudulent.  By using neural networks to its full extent, Visa will be able to
provide superior risk analysis for its members.

In combating credit and debit card fraud, Visa already has developed several
programs utilizing information gained from neural network research.  Worldwide
implementation of the International Points-of- Compromise (IPOC) program has
proved highly effective for identifying merchant locations that may be selling
or giving account information to counterfeiters.  Another successful program,
called the Central Deposit Monitoring (CDM) program, matches merchant activity
with sales draft laundering characteristics and identifies unusual merchant
deposits. In addition, close cooperation with law enforcement agencies and
legislatures enhances the value of the programs which, in turn, allow Visa
members to pass on the protection to its cardholders and merchants.

The planned Visa-HNC merchant risk detection system is designed to further
reduce fraud losses by assigning a risk score to each authorization
transaction processed through the VisaNet systems.  "With this new system,
members will be better able to assess risk at the point- of-transaction and,
therefore, make more informed authorization decisions," confirmed Peirce.
According to The Nilson Report, merchant fraud worldwide cost the financial
industry an estimated US $689 million in 1992.

HNC will integrate the risk score into Falcon(TM), their existing, real-time
credit card fraud-detection system that runs at card issuer sites to identify
and prevent a wide range of fraud at the cardholder level. It determines the
probability of fraud on each credit card authorization by comparing it to the
cardholder's purchase patterns and the latest trends in credit card fraud.
Introduced in September 1992, Falcon has already achieved success in reducing
fraud losses of major credit card issuers.

HNC Inc., the world's leader in the application of neural networks, develops,
sells, integrates and supports advanced decision solutions based on neural
network and statistical technology. HNC provides practical products and
services to the financial, credit card, debit card, merchant services,
insurance, mortgage underwriting, retail and direct marketing industries.

Visa is the leading consumer payment system in the world with more than 10.4
million acceptance locations, the largest global ATM network and 309 million
cards issued worldwide.

-0-                          8/10/93 /CONTACT:  Gail Murayama of Visa
International, 415-570-3645; or Ken Jones of HNC Inc., 619-546-8877

____________________________________________ _ _..........

Date: Wed, 25 Aug 93 08:37:07 -0700
   From: strick -- henry strickland <strick@versant.com>

   Extracted from "FC NEWSBYTES 1.3", David Geddes <dgeddes@NETCOM.COM> Editor,
	   where FC = FutureCulture mailing list <FUTUREC-request@UAFSYSB.UARK.EDU>.    strick
   ____________________________________________ _ _..........
   B Y T E 4:

   Visa, HNC Inc. develop neural network as a weapon to fight fraud
   SAN FRANCISCO (AUG. 10) PR NEWSWIRE - Visa International and HNC Inc. have
   announced a strategic agreement to develop a comprehensive merchant risk
   detection system.  The new system will be designed to better control fraud at
   the merchant level by determining the risk associated with individual card
   transactions.

For those who are not familiar with the details of neural networks,  I
thought I would point out that this represents a departure from the 
current notion of a credit rating in two ways:

1) There is no clear way to fix your "neural credit rating" if there
   is a problem.

The neural network program which predicts the probability of fraud will
give its guess as to the probability of fraud.  If you are a cardholder
and it predicts that a transaction is likely to be fraudulent,  then
your purchase won't be accepted.  But,  unlike conventional credit reporting
firms which use a credit report,  the neural network cannot explain
anything about how it came to its decision.  With existing credit reporting
schemes,  you at least have the option of acquiring your credit report and
taking the necessary steps to repair your credit rating if there is a
problem.  With the use of neural networks,  this is no longer possible.

Given the current state of neural network research,  a percentage of the 
rejections will be false.  This means that a number of card users will be 
denied service for no other reason than the fact that neural networks make 
mistakes.

2) You are no longer judged on your own actions,  but on the similarity
   of your purchasing patterns with those who have committed fraudulent acts.

Instead of being judged on your trustworthiness based on your past actions,
you will be judged based on whether people whose purchasing profiles are
similar to yours are trustworthy.  An example of this being problematic
is say you purchase a particular CD and the neural network decides that,
partly based on this and partly on other information,  that you won't
pay your bill because most of the people in the database who bought that
CD didn't pay their bills.

Andy

In article <9308252001.AA14017@custard.think.com>,
Andy Wilson <ajw@Think.COM> wrote:
: [mostly bogus stuff]

That is irrelevant to cypherpunks, as I understand the list.

There is no technology, including that of privacy, that cannot be
used for ill. We don't know how they're going to be using the
neural network. They could, as was suggested, abandon their minds
and and rely on the neural net. I don't think they will because
doing so would be a really bad business decision. Furthermore, on
the evidence, the neural network output will only be used as one
datum in a process involving many inputs and a human making the
final decision. Finally, in the examples I'm familiar with (from
reading AI Expert), when a neural net is used as a decision
element, precisely because of its error rate, the decision isn't
"go/no go" but "go/refer the problem to a human".

From: bill@twwells.com (T. William Wells)
   Date: Wed, 25 Aug 1993 21:04:05 GMT

   In article <9308252001.AA14017@custard.think.com>,
   Andy Wilson <ajw@Think.COM> wrote:
   : [mostly bogus stuff]

   That is irrelevant to cypherpunks, as I understand the list.

The prospect of the impossibility of anonymity and the uses
to which personal information is made in a cashless economy is 
not relevant?  I beg to differ.  This is exactly what digital
cash is meant to prevent.

   There is no technology, including that of privacy, that cannot be
   used for ill. We don't know how they're going to be using the
   neural network. They could, as was suggested, abandon their minds
   and and rely on the neural net. I don't think they will because
   doing so would be a really bad business decision. Furthermore, on
   the evidence, the neural network output will only be used as one
   datum in a process involving many inputs and a human making the
   final decision. Finally, in the examples I'm familiar with (from
   reading AI Expert), when a neural net is used as a decision
   element, precisely because of its error rate, the decision isn't
   "go/no go" but "go/refer the problem to a human".

The problem with referring a neural network's decision to a human
is that the neural network gives no information other than the 
probability of fraud.  It does not tell the human why it determined
the transaction was likely to be flawed,  like a system based on 
rules or case-based reasoning would be able to do.  There is not any 
good way to combine the judgement of the neural net with that of a
human for that reason.

With respect,  I have found AI Expert to consist more of marketing
hype than correct and useful information on artificial intelligence 
technology.

Andy

In article <9308252250.AA17771@custard.think.com>,
Andy Wilson <ajw@Think.COM> wrote:
:    Andy Wilson <ajw@Think.COM> wrote:
:    : [mostly bogus stuff]
:
:    That is irrelevant to cypherpunks, as I understand the list.
:
: The prospect of the impossibility of anonymity and the uses
: to which personal information is made in a cashless economy is
: not relevant?

But that wasn't what you were writing about. You were writing
about bad business decisions, not violations of privacy.

For that matter, your notions on neural networks seem
contradictory. On the one hand, you complain about a violation of
privacy and on the other you complain that a neural network won't
tell you how it reached its conclusions!

:                I beg to differ.  This is exactly what digital
: cash is meant to prevent.

Digital cash and the use of neural networks to authenticate
transactions are essentially orthogonal issues.

: The problem with referring a neural network's decision to a human
: is that the neural network gives no information other than the
: probability of fraud.

1) This statement is false. It is true of some neural networks
   but not all. We have no way of knowing whether their neural
   network is among those.

2) A problem with *any* decision system is that people may place
   an unsupportable weight on some particular piece of evidence.
   Your "problem" is not that (some) neural networks give answers
   that can't be interpreted but that some people will use their
   answers in an inappropriate way.

Blaming neural networks for bad *human* decision making is just
plain silly.

:                                                     There is not any
: good way to combine the judgement of the neural net with that of a
: human for that reason.

Nonsense. As the existence of rule based systems that incorporate
neural networks shows.

: With respect,  I have found AI Expert to consist more of marketing
: hype than correct and useful information on artificial intelligence
: technology.

Oh, goodie, an ad hominem argument.

But, as it happens, it is because AI Expert is so commercially
oriented that it is an appropriate reference. It speaks to how,
and why, AI gets deployed in business and that makes it just the
right place to go.

From: bill@twwells.com (T. William Wells)
   Date: Thu, 26 Aug 1993 01:21:03 GMT

   [...]

   But that wasn't what you were writing about. You were writing
   about bad business decisions, not violations of privacy.

No, you were writing about bad business decisions.  I was providing
a few details on how credit/charge-card information is used in this
process and a few potential problems resulting from it.

   For that matter, your notions on neural networks seem
   contradictory. On the one hand, you complain about a violation of
   privacy and on the other you complain that a neural network won't
   tell you how it reached its conclusions!

You are deliberately confusing two different points: 1) the fact that
neural networks do not provide useful explanations of how they arrived
at a particular decision,  and 2) some potential problems that arise
from this fact that concern privacy issues.

   :                I beg to differ.  This is exactly what digital
   : cash is meant to prevent.

   Digital cash and the use of neural networks to authenticate
   transactions are essentially orthogonal issues.

I will reiterate that the whole point of digital cash is to provide
anonymity,  which will prevent these kinds of uses made of personal
information which are not done with the explicit approval of the
person involved.

   : The problem with referring a neural network's decision to a human
   : is that the neural network gives no information other than the
   : probability of fraud.

   1) This statement is false. It is true of some neural networks
      but not all. We have no way of knowing whether their neural
      network is among those.

It is true of all commercial applications of neural networks to my
knowledge,  and certainly true of the neural networks developed
by Hecht-Nielsen.

   :                                                     There is not any
   : good way to combine the judgement of the neural net with that of a
   : human for that reason.

   Nonsense. As the existence of rule based systems that incorporate
   neural networks shows.

That shows no such thing.  The only way to combine the judgement of
a neural network with that of a rule-based system,  or anything else,
is to see if both arrive at the same conclusion.  You cannot see the
reasoning process of the neural network to help the human understand
why it made the judgement that it did,  the marketing hype of neural
network vendors notwithstanding.

This is my last post on this thread.

Andy