Credit-Carding 101

[RedX]

Part one in a series of postings about the workings of the credit card
industry.

DEFINITIONS
-----------

First some terms, along with the meanings they have in the industry:

Cardholder - an individual to whom a credit card is issued. Typically,
this individual is also responsible for payment of all charges made
to that card. Corporate cards are an exception to this rule.

Card Issuer - an institution that issues credit cards to cardholders.
This institution is also responsible for billing the cardholder for
charges. Often abbreviated to "Issuer".

Card Accepter - an individual, organization, or corporation that
accepts credit cards as payment for merchandise or services. Often
abbreviated "Accepter" or "merchant".

Acquirer - an organization that collects (acquires) credit
authorization requests from Card Accepters and provides guarantees
of payment. Normally, this will be by agreement with the Issuer of
the card in question.

Many issuers are also acquirers. Some issuers allow other acquirers to
provide authorizations for them, under pre-agreed conditions. Other
issuers provide all their own authorizations.


TYPES OF CARDS
----- -- -----

The industry typically divides up cards by the business of the issuer.
So there are bank cards (VISA, Master Card, Discover), Petroleum Cards
(SUN Oil, Exxon, etc.), and Travel and Entertainment (T&E) cards
(American Express, Diners' Club, Carte Blanche). Other cards are
typically lumped together as "Private Label" cards. That would include
department store cards, telephone cards, and the like. Most private
label cards are only accepted by the issuer. People are starting to
divide the telephone cards into a separate class, but it hasn't re-
ceived widespread acceptance. (This is just a matter of terminology,
and doesn't affect anything important.)

Cards are also divided by how they are billed. Thus there are credit
cards (VISA, MC, Discover, most department store cards), charge cards
(American Express, AT&T, many petroleum cards) and debit cards. Credit
cards invoke a loan of money by the issuer to the cardholder under
pre-arranged terms and conditions. Charge cards are simply a payment
convenience, and their total balance is due when billed. When a debit
card is used, the amount is taken directly from the cardholder's ac-
count with the issuer. Terminology is loose - often people use "credit
card" to encompass credit cards and charge cards.

A recent phenomenon is third-party debit cards. These cards are issued
by an organization with which the cardholder has no account relation-
ship. Instead, the cardholder provides the card issuer with the infor-
mation necessary to debit the cardholder's checking account directly
through an Automated Clearing House (ACH), the same way a check would
be cleared. This is sort of like direct deposit of paychecks, in re-
verse. ACHs love third-party debit cards. Banks hate them.

Another recent addition is affinity cards. These cards are valid
credit cards from their issuer, but carry the logo of a third party,
and the third party benefits from their use. There is an incredible
variety of affinity cards, ranging from airlines to colleges to profes-
sional sports teams.


HOW THEY MAKE MONEY
--- ---- ---- -----

Issuers of credit cards make money from cardholder fees and from inter-
est paid on outstanding balances. Not all issuers charge fees. Even
those that do, make most of their money on the interest. They really
LIKE people who pay the minimum each month.

Issuers of charge cards make money from cardholder fees. Some charge
cards actually run at a loss for the company, particularly those that
are free. The primary purpose of such cards is to stimulate business.

Issuers of debit cards may make money on transaction fees. Not all
debit card transactions have fees. Most debit cards exist to stimulate
business for the bank and to offload tellers and back-room departments.
To date, third-party debit cards exist solely to stimulate business.
Providers of such cards make no direct money from their use.

Acquirers make money from transaction charges and discount fees. Unlike
the charges and fees mentioned above, these fees are paid by the ac-
cepter, not (directly) by the cardholder. (Technically, it is not le-
gal for the merchants to pass these charges directly to the consumer.
Some petroleum stations have gotten away with giving a discount for
cash, and it has survived court challenges so far.) Transaction charges
are typically in pennies per transaction, and are sensitive to the type
of communication used for the authorization. Discount fees are a per-
centage of the purchase price and are sensitive to volume and compli-
ance to rules. One way to encourage merchants to follow certain
procedures or to upgrade to new equipment is to offer a lower discount
fee.

Until fairly recently, the only motivation for accepters was to expand
their business by accepting cards. Reduction of fraud was enough rea-
son for many merchants to pay authorization fees, but in many cases, it
isn't worth the cost. (That is, it is cheaper to pay the fraud than to
prevent it.) Recently, electronic settlement has provided merchants
with an added benefit by reducing float on charged purchases. Merchants
can now get their accounts credited much faster than before, which
helps cash flow.

Companies that issue charge cards are real keen on float reduction. The
sooner they can bill you, the sooner they get their money. Credit card
companies are also interested in float reduction, since the sooner they
bill, the sooner they can start charging interest. Debit cards
typically involve little or no float.

Affinity cards usually pay a percentage of purchases to the affinity
organization. Although it may seem obvious to take this money from the
discount fee, this doesn't work since the issuer is not always the
acquirer. The money for this usually comes from the interest paid on
outstanding balances. Essentially, the bank is giving a share of its
profits to an organization in turn for the organization promoting use
of its credit card. The affinity organization is free to use its cut
any way it wishes. An airline will typically put it into the frequent
flyer program (and credit miles to your account). A college may put
the money into the general fund or into a scholarship fund. Lord only
knows what a sports team does with the money!


THE PLAYERS AND THEIR ROLES
--- ------- --- ----- -----

American Express (AMEX) is a charge card issuer and acquirer. (Their
other businesses are not important to this discussion.) All AMEX pur-
chases are authorized by AMEX. They make most of their money from the
discount fees, which is why they have the highest discount fee in the
industry. That's one reason why AMEX isn't accepted in as many places
as VISA and MC, and a reason why many merchants will prefer another
card to an AMEX card. The control AMEX has over authorization allows
them to provide what they consider to be better cardholder
("cardmember" to them) services.

VISA is a non-profit corporation (SURPRISE!) that is best described as
a purchasing and marketing coalition of its member banks. VISA issues
no credit cards itself - all VISA cards are issued by member banks.
VISA does not set terms and conditions for its member banks - the banks
can do pretty much as they please in signing cardholders. All VISA
charges are ultimately approved by the card issuer, regardless of where
the purchase was made. Many smaller banks share their account
databases with larger banks, third parties, or VISA itself, so that the
bank doesn't have to provide authorization facilities itself.

Master Card (MC) is very much like VISA. There are some differences
that are important to those in the industry, but from the consumers
standpoint they operate pretty much the same.

Discover cards are issued by a bank owned by Sears. All Discover pur-
chases are authorized by Sears.

Most petroleum cards, if they are even authorized, are authorized by
the petroleum company itself. There are exceptions. Fraud on petro-
leum cards is so low that the main reason for authorization is to
achieve the float reduction of electronic settlement.


THE BUSINESS RELATIONSHIPS
--- -------- -------------

Card acceptors generally sign up with a local acquirer for authoriza-
tion and settlement of all credit cards. This acquirer may or may not
be a card issuer, but certainly will not have issued all the cards that
the merchant can accept. The accepter does not generally call one
place for VISA and a different place for MC, for example. At one time,
this was necessary, but more and more acquirers are connected to all
networks and are offering a broader range of services.

Acquirers generally are connected to many issuers, and pay transaction
charges and discount fees to those issuers for authorizations. Thus,
the acquirer is actually making money on the difference between fees
paid and fees billed. Most acquirers gather together transactions from
many accepters, allowing them to get volume discounts on fees. Since
the accepters individually have lower volume and are not eligible for
those discounts, there is a markup that the acquirer can get away with.
Acquirers also, of course, provide the convenience of a single contact.

Most large banks are issuers and acquirers. Things get real interest-
ing when it's time to settle up. Some small banks are only issuers.
There are third parties that are only acquirers.

In future episodes, I'll explain how standards help all this chaos work
together, and give details about how the authorization process happens.

Joe Ziegler
att!lznv!ziegler
This is part two in a planned six-part series about the credit card in-
dustry. It would be best if you read part one before reading this
part. Enjoy.

DEFINITIONS
-----------

Some more new terms that are used in this posting.

ABA - American Bankers Association

ACH - Automated Clearing House - an organization that mechanically and
electronically processes checks.

ANSI - American National Standards Institute

Embossing - creating raised letters and numbers on the face of the
card.

Encoding - recording data on the magnetic stripe on the back of the
card.

Imprinting - using the embossed information to make an impression on a
charge slip.

Interchange - sending authorization requests from one host (the
acquirer) to another (the issuer) for approval.

ISO - International Standards Organization

NACHA - National Automated Clearing House Association

PAN - Personal Account Number. The account number associated with a
credit, debit or charge card. This is usually the same as the
number on the card.

PIN - Personal Identification Number. A number associated with the
card, that is supposedly know only to the cardholder and the card
issuer. This number is used for verification of cardholder
identity.


THE ORGANIZATIONS
--- -------------

ISO sets standards for plastic cards and for data interchange, among
other things. ISO standards generally allow for national expansion.
Typically, a national standards organization, like ANSI, will take an
ISO standard and develop a national standard from it. National stan-
dards are generally subsets of the ISO standard, with extensions as al-
lowed in the original ISO standard. Many credit card standards
originated in the United States, and were generalized and adopted by
ISO later.

The ANSI committees that deal with credit card standards are sponsored
by the ABA. Most members of these committees work for banks and other
financial institutions, or for vendors who supply banks and financial
institutions. Working committees report to governing committees.

All standards go through a formal comment and review procedure before
they are officially adopted.


PHYSICAL STANDARDS
-------- ---------

ANSI X4.13, "American National Standard for Financial Services -
Financial Transaction Cards" defines the size, shape, and other
physical characteristics of credit cards. Most of it is of interest
only to mechanical engineers. It defines the location and size of the
magnetic stripe, signature panel, and embossing area. This standard
also includes the Luhn formula used to generate the check digit for the
PAN, and gives the first cut at identifying card type from the account
number. (This part was expanded later in other standards.) Also, this
standard identifies the character sets that can be used for embossing a
card.

Three character sets are allowed - OCR-A as defined in ANSI X3.17,
OCR-B as defined in ANSI X3.49, and Farrington 7B, which is defined in
the appendix of ANSI X4.13 itself. Almost all the cards I have use
Farrington 7B, but Sears uses OCR-A. (Sears also uses the optional,
smaller card size as, allowed in the standard.) These character sets
are intended to be used with optical character readers (hence the OCR),
and large issuers have some pretty impressive equipment to read those
slips.


ENCODING STANDARDS
-------- ---------

ANSI X4.16, "American National Standard for Financial Services - Finan-
cial Transaction Cards - Magnetic Stripe Encoding" defines the
physical, chemical, and magnetic characteristics of the magnetic stripe
on the card. The standard defines a minimum and maximum size for the
stripe, and the location of the three defined encoding tracks. (Some
cards have a fourth, proprietary track.)

Track 1 is encoded at 210 bits per inch, and uses a 6-bit coding of a
64-element character set of numerics, alphabet (one case only), and
some special characters. Track 1 can hold up to 79 characters, six of
which are reserved control characters. Included in these six charac-
ters is a Longitudinal Redundancy Check (LRC) character, so that a card
reader can detect most read failures. Data encoded on track 1 include
PAN, country code, full name, expiration date, and "discretionary
data". Discretionary data is anything the issuer wants it to be.
Track 1 was originally intended for use by airlines, but many Automatic
Teller Machines (ATMs) are now using it to personalize prompts with
your name and your language of choice. Some credit authorization ap-
plications are starting to use track 1 as well.

Track 2 is encoded at 75 bits per inch, and uses a 4-bit coding of the
ten digits. Three of the remaining characters are reserved as
delimiters, two are reserved for device control, and one is left unde-
fined. In practice, the device control characters are never used, ei-
ther. Track 2 can hold up to 40 characters, including an LRC. Data
encoded on track 2 include PAN, country code (optional), expiration
date, and discretionary data. In practice, the country code is hardly
ever used by United States issuers. Later revisions of this standard
added a qualification code that defines the type of the card (debit,
credit, etc.) and limitations on its use. AMEX includes an issue date
in the discretionary data. Track 2 was originally intended for credit
authorization applications. Nowadays, most ATMs use track 2 as well.
Thus, many ATM cards have a "PIN offset" encoded in the discretionary
data. The PIN offset is usually derived by running the PIN through an
encryption algorithm (maybe DES, maybe proprietary) with a secret key.
This allows ATMs to verify your PIN when the host is offline, generally
allowing restricted account access.

Track 3 uses the same density and coding scheme as track 1. The con-
tents of track 3 are defined in ANSI X9.1, "American National Standard
- Magnetic Stripe Data Content for Track 3". There is a slight contra-
diction in this standard, in that it allows up to 107 characters to be
encoded on track 3, while X4.16 only gives enough physical room for 105
characters. Actually, there is over a quarter of an inch on each end
of the card unused, so there really is room for the data. In practice,
nobody ever uses that many characters, anyway. The original intent was
for track 3 to be a read/write track (tracks 1 and 2 are intended to be
read-only) for use by ATMs. It contains information needed to maintain
account balances on the card itself. As far as I know, nobody is actu-
ally using track 3 for this purpose anymore, because it is very easy to
defraud.


COMMUNICATION STANDARDS
------------- ---------

Formats for interchange of messages between hosts (acquirer to issuer)
is defined by ANSI X9.2, which I helped define. Financial message au-
thentication is described by ANSI X9.9. PIN management and security is
described by ANSI X9.8. There is a committee working on formats of
messages from accepter to acquirer. ISO has re-convened the interna-
tional committee on host message interchange (TC68/SC5/WG1), and ANSI
may need to re-convene the X9.2 committee after the ISO committee fin-
ishes. These standards are still evolving, and are less specific than
the older standards mentioned above. This makes them somewhat less
useful, but is a natural result of the dramatic progress in the indus-
try.

ISO maintains a registry of card numbers and the issuers to which they
are assigned. Given a card that follows standards (Not all of them
do.) and the register, you can tell who issued the card based on the
first six digits (in most cases). This identifies not just VISA,
MasterCard, etc., but also which member bank actually issued the card.


DE FACTO INDUSTRY STANDARDS
-- ----- -------- ---------

Most ATMs use IBM synchronous protocols, and many networks are migrat-
ing toward SNA. There are exceptions, of course. Message formats used
for ATMs vary with the manufacturer, but a message set originally de-
fined by Diebold is fairly widely accepted.

Many large department stores and supermarkets (those that take cards)
run their credit authorization through their cash register controllers,
which communicate using synchronous IBM protocols.

Standalone Point-of-Sale (POS) devices, such as you would find at most
smaller stores (i.e. not at department stores), restaurants and hotels
use a dial-up asynchronous protocol devised by VISA. There are two
generations of this protocol, with the second generation just beginning
to get widespread acceptance.

Many petroleum applications use multipoint private lines and a polled
asynchronous protocol known as TINET. This protocol was developed by
Texas Instruments for a terminal of the same name, the Texas Instru-
ments Network E(something) Terminal. The private lines reduce response
time, but cost a lot more money than dial-up.

NACHA establishes standards for message interchange between ACHs, and
between ACHs and banks, for clearing checks. This is important to this
discussion due to the emergence of third-party debit cards, as dis-
cussed in part 1 of this series. The issuers of third-party debit
cards are connecting to ACHs, using the standard messages, and clearing
POS purchases as though they were checks. This puts the third parties
at an advantage over the banks, because they can achieve the same re-
sults as a bank debit card without the federal and state legal restric-
tions imposed on banks.

In the next installment, I'll describe how an authorization happens, as
well as how the settlement process gets the bill to you and your money
to the merchant. After that I'll describe various methods of fraud,
and how issuers, acquirers, and accepters protect themselves. Stay
tuned.

Joe Ziegler
att!lznv!ziegler
Here's part 3 in my six-part series on the credit card industry. This
part discusses how authorization and settlement work. This is a long
one. It will help if you have read parts 1 and 2, since I had to leave
out a lot of overlap to keep this from getting ridiculous. Enjoy.


THE ACCEPTER
--- --------

An important fact to note is that a card accepter does not have to get
approval for any purchases using credit or charge cards. Of course, a
merchant is usually interested in actually getting money, and so must
participate in some form of settlement process (see below). Usually,
the most acceptable (to a merchant) forms of settlement are tied (by
the acquirer) to authorization processes. However, a merchant could
simply accept all cards without any validation, any eat any fraud that
results.

A merchant typically makes a business arrangement with a local bank or
some other acquirer for authorization and settlement services. The
acquirer assigns a merchant identifier to that merchant, which will
uniquely identify the location of the transaction. (This facilitates
compliance with federal regulations requiring that credit card bills
identify where each purchase was made.) The acquirer also establishes
procedures for the merchant to follow. The procedures will vary by
type of the merchant business, geographic location, volume of transac-
tions, and types of cards accepted.

If the merchant follows the procedures given by the acquirer and a
transaction is approved, the merchant is guaranteed payment whether the
card in question is good or bad. The purpose of authorization is to
shift financial liability from the acceptor to the acquirer.

There are two basic tools used - bulletins and online checks. Bulletins
may be hardcopy, or may be downloaded into a local controller of some
form. Online checks could be done via a voice call, a standalone ter-
minal, or software and/or hardware integrated into the cash register.

A low-volume, high-ticket application (a jewelry store) would probably
do all its authorizations with voice calls, or may have a stand-alone
terminal. A high-volume, low-ticket application (a fast-food chain)
will probably do most of its authorizations locally against a bulletin
downloaded into the cash register controller. Applications in between
typically merge the two - things below a certain amount (the "floor
limit") are locally authorized after a lookup in the bulletin, while
things over the floor limit are authorized online.

Usually a lot of effort is taken to use the least expensive tools that
are required by the expected risk of fraud. Typically, communication
costs for authorizations make up the biggest single item in the overall
cost of providing credit cards.

Large accepters are always a special case. Airlines are usually di-
rectly connected, host-to-host, to issuers and/or acquirers, and autho-
rize everything online. Likewise for many petroleum companies and
large department stores. Some large chains use different approaches at
different locations, either as a result of franchising oddities or due
to volume differences between locations. A lot of experimentation is
still going on as well - this is not a mature market.

For voice authorizations, the merchant ID, PAN, expiration date, and
purchase amount are required for an approval. Some applications also
require the name on the card, but this is not strictly necessary. For
data authorizations, the merchant ID, PAN, PIN (if collected), expira-
tion date, and purchase amount are required. Typically, the "discre-
tionary data" from track 2 is sent as well, but this is not strictly
necessary. In applications that do not transmit the PIN with the au-
thorization, it is the responsibility of the merchant to verify iden-
tity. Usually, this should be done by checking the signature on the
card against the signature on the form. Merchants don't often follow
this procedure, and they take a risk in not doing so.

In most applications, the amount of the purchase is known at the time
of the authorization request. For hotels, car rentals, and some petro-
leum applications, an estimated amount is used for the authorization.
After the transaction is complete (e.g. after the gas is pumped or at
check-out time), another transaction may be sent to advise of the ac-
tual amount of the transaction. More on this later.


THE ACQUIRER
--- --------

The acquirer gathers authorization requests from accepters and returns
approvals. If the acquirer is an issuer as well, "on us" transactions
will typically be turned around locally. As before, the acquirer does
not have to forward any requests on to the actual issuer. However,
acquirers are not willing to take the financial risks associated with
generating local approvals. Thus most transactions are sent on to the
issuers (interchanged). The purpose of interchange is to shift finan-
cial liability from the acquirer to the issuer.

Typically, an acquirer connects to many issuers, and negotiates differ-
ent business arrangements with each one of them. But the acquirer gen-
erally provides a uniform interface to the accepter. Thus, the
interchange rules are sometimes less stringent than those imposed on
the accepter. Also, most issuers will trust acquirers to with respon-
sibilities they would never trust to accepters. The acquirer can
therefore perform some front-end screening on the transactions, and
turn some of them around locally without going back to the issuer.

The first screening by the acquirer would be a "sanity" test, for valid
merchant ID, valid Luhn check on PAN, expiration date not past, amount
field within reason for type of merchant, etc. After that, a floor
limit check will be done. Issuers generally give acquirers higher
floor limits than acquirers give accepters, and floor limits may vary
by type of merchant. Next, a "negative file" check would be done
against a file of known bad cards. (This is essentially the same as
the bulletin.) Then a "velocity file" check may be done. A velocity
file keeps track of card usage, and limits are often imposed on both
number of uses and total amount charged within a given time period.
Sometimes multiple time periods are used, and it can get fairly compli-
cated.

Transactions that pass all the checks, and are within the authority
vested in the acquirer by the issuer, are approved by the acquirer.
(Note that, under the business arrangement, financial liability still
resides with the issuer.) An "advice" transaction is sometimes sent to
the issuer (perhaps at a later time), to tell the issuer that the
transaction took place.

Transactions that "fail" one or more checks are denied by the acquirer
(if the cause was due to form, such as bad PAN) or sent to the issuer
for further checking. (Note that "failure" here can mean that it's be-
yond the acquirer's authority, not necessarily that the card is bad.)
Some systems nowadays will periodically take transactions that would
otherwise be approved locally, and send them to the issuer anyway. This
serves as a check on the screening software and as a countermeasure
against fraudulent users who know the limits.

Transactions that go to the issuer are routed according to the first
six digits of the PAN, according to the ISO registry mentioned in an
earlier section. Actually, it's a bit more complicated than that,
since there can be multiple layers of acquirers, and some issuers or
acquirers will "stand in" for other issuers when there are hardware or
communication failures, but the general principal is the same at each
point.


THE ISSUER
--- ------

An issuer receiving an interchanged transaction will often perform many
of the same tests on it that the acquirer performs. Some of the tests
may be eliminated if the acquirer is trusted to do them correctly. This
is the only point where a velocity file can actually detect all usage
of a card. This is also the only point where a "positive file" lookup
against the actual account can be done, since only the issuer has the
account relationship with the cardholder. If a PIN is used in the
transaction, only the issuer can provide true PIN verification -
acquirers may be able to do only "PIN offset" checking, as described in
a previous section. This is one reason why PINs have not become
popular on credit and charge cards.

An account typically has a credit limit associated with it. An ap-
proved authorization request usually places a "hold" against the credit
limit. If the sum of outstanding holds plus the actual outstanding
balance on the account, plus the amount of the current transaction, is
greater than the credit limit, the transaction is (usually) denied.
Often in such a case the issuer will send back a "call me" response to
the merchant. The merchant will then call the issuer's number, and the
operator may even want to talk to the cardholder. The credit limit
could be extended on the spot, or artificially high holds (from hotels
or car rental companies) could be overlooked so that the transaction
can be approved.

The difference between the credit limit and the sum of holds and out-
standing balance is often referred to as the "open to buy" amount. Once
a hold is placed on an account, it is kept there until the actual the
transaction in question is settled (see below), in which case the
amount goes from a hold to a billed amount, with no impact on the open
to buy amount, theoretically. For authorizations of an estimated
amount, the actual settled amount will be less than or equal to the ap-
proved amount. (If not, the settlement can be denied, and the merchant
must initiate a new transaction to get the money.) Theoretically, in
such a case, the full hold is removed and the actual amount is added to
the outstanding balance, resulting in a possible increase in the open
to buy amount.

In practice, older systems were not capable of matching settlements to
authorizations, and holds were simply expired based on the time it
would take most transactions to clear. Newer systems are starting to
get more sophisticated, and can do a reasonable job of matching autho-
rizations for actual amounts with the settlements. Some of them still
don't match estimated amounts well, with varying effects. In some
cases, the difference between actual and estimated will remain as a
hold for some period of time. In other cases, both the authorization
and the settlement will go against the account, reducing the open to
buy by up to twice the actual amount, until the hold expires. These
problems are getting better as the software gets more sophisticated.

Some issuers are also starting to use much more sophisticated usage
checks as well. They will not only detect number of uses and amount
over time, but also types of merchandise bought, or other patterns to
buying behavior. Most of this stuff is new, and is used for fraud pre-
vention. I expect this to be the biggest effort in authorization soft-
ware for the next few years.

American Express does things completely differently. There are no
credit limits on AMEX cards. Instead, AMEX relies entirely on usage
patterns, payment history, and financial data about cardmembers to de-
termine whether or not to automatically approve a transaction. AMEX
also has a policy that a cardmember will never be denied by a machine.
Thus, if the computer determines that a transaction is too risky, the
merchant will receive a "call me" message. The operator will then get
details of the transaction from the merchant, and may talk to the
cardmember as well, if cardmember identity is in question or a large
amount is requested. To verify cardmember identity, the cardmember
will be asked about personal information from the original application,
or about recent usage history. The questions are not the same each
time. If an unusually large amount is requested, the cardmember may be
asked for additional financial data, particularly anything relating to
a change in financial status (like a new job or a promotion). People
who are paranoid about Big Brother and computer databases should not
use AMEX cards.


SETTLEMENT
----------

So far, no money has changed hands, only financial liability. The pur-
pose of settlement is to shift the financial liability back to the
cardholder, and to shift the cardholder's money to the merchant.
Theoretically, all authorization information can be simply discarded
once an approval is received by a merchant. Of course, contested
charges, chargebacks, merchant credits, and proper processing of holds
require that the information stay around. Still, it is important to
realize that an authorization transaction has no direct financial con-
sequences. It only establishes who is responsible for the financial
consequences to follow.

Traditionally, a merchant would take the charge slips to the bank that
was that merchant's acquirer, and "deposit" them into the merchant ac-
count. The acquirer would take the slips, sort them by issuer, and
send them to the issuing banks, receiving credits by wire once they ar-
rived and were processed. The issuer would receive the slips, micro-
film them (to save the transaction information, as required by federal
and state laws) charge them against the cardholder's accounts, send
credits by wire to the acquirer, and send out the bill to the
cardholder. Problem is, this took time. Merchants generally had to
wait a couple of weeks for the money to be available in their accounts,
and issuers often suffered from float on the billables of about 45
days.

Therefore, nowadays many issuers and acquirers are moving to on-line
settlement of transactions. This is often called "draft capture" in
the industry. There are two ways this is done - one based on the host
and one based on the terminal at the merchant's premises. In the
host-based case, the terminal generally only keeps counts and totals,
while the acquirer host keeps all the transaction details. Peri-
odically, the acquirer host and the terminal communicate, and verify
that they both agree on the data. In the terminal-based case, the ter-
minal remembers all the important transaction information, and peri-
odically calls the acquirer host and replays it all for several
transactions. In either case, once the settlement is complete the mer-
chant account is credited. The acquirer then sends the settlement in-
formation electronically to the issuers, and is credited by wire
immediately (or nearly so). The issuer can bill directly to the
cardholder account, and float can be reduced to an average of 15 days.

The problem is, what to do with the paper? Current regulations in many
states require that it be saved, but there is no need for it to be sent
to the issuer. Also, for contested charges, a paper trail is much more
likely to stand up in court, and much better to use for fraud investi-
gations. Currently, the paper usually ends up back at the issuer, as
before, but it doesn't need to be processed, just microfilmed and
stored.

Much of the market still uses paper settlement methods. Online settle-
ment will replace virtually all of this within the next 5 to 10 years,
because of its many benefits.

This was pretty long, but there is a lot of information, and I skimmed
over a lot of details. Future installments should be shorter. Coming
up next is a discussion of fraud and security, and then a special dis-
cussion of debit cards. Hang on, we're halfway through this!


Joe Ziegler
att!lznv!ziegler
This is part four of a planned six-part series on the credit card in-
dustry. It will be helpful if you have read parts one through three,
as I use a lot of terminology here that was introduced earlier. Enjoy.

WARNING

This installment describes various methods of perpetrating fraud
against credit and charge card issuers, acquirers, and cardholders. Le-
gal penalties for using these methods to commit fraud are severe. The
reason for sharing this information is so that consumers will be aware
of the importance of security and be aware of the procedures used by
financial institutions to protect against fraud. Neither I nor my em-
ployer advocate use of the fraudulent methods described herein.

All the information here is publicly available from other sources. Un-
necessary detail is purposely not included, particularly as it applies
to detection and prevention of fraud.


CARDHOLDER FRAUD
---------- -----

The most common type of fraud against credit cards is cardholders fal-
sifying applications to get higher credit limits than they can afford
to pay, or to get multiple cards that they cannot afford to pay off.
Sometimes this is done with intent to defraud, but most often it is
done out of desperation or sheer financial ineptitude. Those who in-
tend to defraud generally use the multiple-card approach. They give
false names and financial data on several (sometimes as many as hun-
dreds) of applications. Often, the address of a vacant house that the
crook has access to is given, making it difficult to track the crook's
real identity. Once cards start showing up, the crook uses them for
cash advances or charges merchandise that is easy to sell, like con-
sumer electronics. The crook will run all the cards up to the limit
immediately, and will generally move on by the time the bills start ar-
riving. This type of fraud is not applicable to debit cards, since
they require an available account balance equal to or greater than any
purchases or withdrawals.

Protecting against this type of fraud, either intentional or otherwise,
is exactly the purpose of credit bureaus such as TRW. Issuers have be-
come more aware of the need for careful screening of applications, and
are using better techniques for detecting similar applications sent to
multiple issuers. More sophisticated velocity file screening can also
be used to detect possibly fraudulent usage patterns. Since this is a
method of fraud that can be used to gain really large amounts of
money, it is a high priority with issuers' security departments.

A variant of this scheme is much like check kiting. Can you use your
VISA to pay your MasterCard? Well, you might be able to manage it, but
if you're doing it with intent to defraud, you can be prosecuted. Kit-
ing schemes typically don't last long, have a low payoff, and are very
easy to detect.

Another type of cardholder fraud is simply contesting legitimate
charges. Most often, retrieving the documents gives pretty convincing
proof. Frequently, a family member will be found to have used the card
without the cardholder's permission. Such cases are usually pretty
easy to resolve. In the case of an ATM card, cameras are often placed
at ATMs (sometimes hidden) to record users of the machine. The camera
is usually tied to the ATM, so that a single retrieval stamp can be
placed on the film and the ATM log. If a withdrawal is contested, the
bank can then retrieve the picture of the person standing at the ma-
chine, and conclusively tie that picture to the transaction.

A type of cardholder fraud that is endemic only to ATMs is making false
deposits. You could, theoretically, tell the ATM that you are deposit-
ing a large amount of money, and put in an empty envelope. Most banks
will not let you withdraw amounts deposited into an ATM until the de-
posit has been verified, but some will allow part of the deposit to be
withdrawn. Typically, you can't get away with much. If you have any
money actually in your account, the bank has easy, legal recourse to
seize those funds. Most banks have no sense of humor about such
things, and will remove ATM card privileges after the first offense.


THIRD-PARTY FRAUD
----------- -----

The simplest way for a third party to commit fraud is for them to get
their hands on a legitimate card. There is a large black market for
credit cards obtained from hold-ups, break-ins and muggings. Perhaps
one of the cruelest methods of getting a card is a "Good Samaritan"
scam. In such a scam, credit cards are stolen by pick-pockets,
purse-snatchers, etc. That same day, someone looks up your number in
the phone book and calls you up. "I just found your wallet. All the
money is gone, but the credit cards and your driver's license are still
here. It just happens that I'll be in your neighborhood next Wednesday
and I'll drop it off then." Since the cards are found, you don't re-
port them stolen, and the crooks get until next Wednesday before you're
even suspicious. If such a thing happens to you, ask if you can come
and pick the cards up immediately. A true good samaritan won't mind,
but a crook will stall you. If you can't get your hands on the cards
immediately, report them as stolen. Most issuers will be able to get
you a new card by next Wednesday, anyway.

Often stolen cards will be used for a time exactly as is. The best
tool for preventing this is verification of the signature, but this is
ineffective because most merchants don't consistently check signatures
and some people don't even sign their cards. (I guess these people
figure that all purse snatchers are accomplished forgers as well.)
Many cards will eventually be modified as the various security schemes
start catching up.

It is a very easy matter, for example, to re-encode a different number
on the magnetic stripe. Since the card still looks fine, a merchant
will accept it and run it through the POS terminal, completely ignorant
of the fact that the number read off the back is not the same as that
on the front. Although the number on the front would fail a negative
file check, the number on the back is one that hasn't been reported
yet. A card can be re-encoded almost any number of times, as long as
you can keep coming up with new valid PANs. To protect against this,
some merchants purposely avoid using the magnetic stripe. Others have
terminals that display the number read from the stripe, so the cashier
can compare it to the number on the card. Some issuers are experiment-
ing with special encoding schemes, to make re-encoding difficult, but
most of these schemes would require replacing the entire embedded base
of POS terminals. An interesting approach I've seen (it's probably
patented) uses a laser to burn off the parts of the magnetic stripe
where zeroes are encoded, leaving only the ones. This severely limits
the changes you can make to the card number. Some issuers use the
"discretionary data" field to encode data unique to the card, that a
crook would not be able to guess, to combat this type of fraud.

Since an ATM doesn't have a human looking at the card, it is especially
susceptible to re-encoding fraud. A crook could get a number from a
discarded receipt and encode it on a white card blank, which is easy to
obtain legally. Many people use PINs that are easy to guess, and the
crook has an easy job of it. Most ATMs will not give you your card
back if you don't enter a correct PIN, and will only give you a few
tries to get it right, to prevent this type of fraud. Velocity file
checks are also important in detecting this. You should always take
your ATM receipts with you, pick a non-obvious PIN, and make sure that
nobody sees you enter it.

One place that a crook can get valid PANs to encode on credit cards is
from dumpsters outside of stores and restaurants. The credit slip
typically is a multipart form, with one copy for you, one for the mer-
chant, and one for the issuer (ultimately). If carbon paper is used,
and the carbons are discarded intact, it's pretty easy to read the num-
bers off of them. Carbonless paper and forms that either rip the car-
bons in half or attach them to the cardholder copy automatically are
used to prevent this.

There are a lot of scams for getting people to tell their credit card
numbers over the phone. Never give your card number to anyone unless
you are buying something from them, and make sure that it is a le-
gitimate business you are buying from. "Incredible deal!! Diamond
jewelry at half price!! Call now with your VISA number, and we'll rush
you your necklace!!" When you don't get the necklace for four weeks,
you might start to wonder. When you get your credit card bill, you'll
stop wondering.

There are other, more sophisticated ways to modify a credit card. If
you're skillful, you can change the embossing on the card and even the
signature on the back. For most purposes, these techniques are more
trouble than they're worth, since it's not difficult to come up with a
new stolen card, or fake ID to match the existing card.


MERCHANT FRAUD
-------- -----

There are many urban rumors of merchants imprinting a card multiple
times while the cardholder isn't looking, and then running through a
bunch of charges after the cardholder leaves. I don't know of any case
where this is an official policy of a merchant, but this is certainly
one technique a dishonest cashier could use. The cashier can then take
home a bunch of merchandise charged to your account. Although some
people are afraid of this happening in a restaurant, where a waiter
takes your card away for a while, it's actually less likely there,
since there isn't anything the waiter can charge against your card and
take home.

A merchant could also make copies of charge slips, to sell the PANs to
other crooks. (See above for use of PANs.) Most credit card investi-
gation departments are sensitive to this possibility, and catch on real
fast if it's happening just by looking at usage history of cards with
fraudulent charges.

A merchant is also in a position to create many false charges against
bogus numbers, to attempt to defraud the acquirer or issuer. These
schemes are usually not too effective, since acquirers generally re-
spond very quickly to an unusual number of fraudulent transactions by
tightening restrictions on the merchant.


ACQUIRER AND ISSUER FRAUD
-------- --- ------ -----

The place to make really big bucks in fraud is at the acquirer or is-
suer, since this is where you can get access to large amounts of money.
Fortunately, it's also fairly easy to control things here with audit
procedures and dual control. People working in the back offices, pro-
cessing credit slips, bills, etc. have a big opportunity to "lose"
things, introduce false things, artificially delay things, and tempo-
rarily divert things. Most of the control is standard banking stuff,
and has been proven effective for decades, so this isn't a big problem.
A bigger potential problem to the consumer is the possibility of an em-
ployee at the issuer or acquirer selling PANs to crooks. This would be
very hard to track down, and could compromise a large part of the card
base. I know of no cases where this has happened.

Programmers, in particular, are very dangerous because they know where
the data is, how to get it, and what to do with it. In most shops, de-
velopment is done on completely separate facilities from the production
system. Certification and installation are done by non-developers, and
developers are not allowed any access to the production facilities.
Operations and maintenance staff are monitored very carefully as well,
since they typically have access to the entire system as part of their
jobs.

Another type of fraud that is possible here is diversion of materials,
such as printed, but not embossed or encoded, card blanks. Such mate-
rials are typically controlled using processes similar to those used at
U.S. mints. Since most of the cards issued in the United States are
actually manufactured by only a handful of companies, it's not too hard
to keep things under control.

There are many types of fraud that can be perpetrated by tapping data
communication lines, and using protocol analyzers or computers to in-
tercept or introduce data. These types of fraud are not widespread,
mainly because of the need for physical access and because sophisti-
cated computer techniques are required. There are message authentica-
tion, encryption, and key management techniques that are available to
combat this type of fraud, but currently these techniques are far more
costly than the minimal fraud they could prevent. About the only such
security technique that is in widespread use is encryption of PINs.

The next episode will be devoted to debit cards, and the final episode
will talk about the networks that make all this magic happen.


Joe Ziegler
att!lznv!ziegler
Part 5 - Debit Cards

EVOLUTION OF DEBIT CARDS
--------- -- ----- -----

The debit card originated as a method for bank customers to have access
to their funds through Automatic Teller Machines (ATMs). This was seen
as a way for banks to automate their branches and save money, as well
as a benefit for customers. A secondary intent was for the card to be
used as a method of identification when dealing with a human teller.
Although that idea never really caught on, it has seen renewed interest
from time to time.

One problem with using cards to access bank accounts is that federal
regulations required a signature be used for each withdrawal transac-
tion. After much debate, the concept of a Personal Identification Num-
ber (PIN) was invented, and federal regulations were modified to allow
PINs for use in place of signatures with bank withdrawals. ATMs also
faced many other regulatory difficulties. In many states, for example,
there are limitations on the number of branches a bank can have. In a
conflict that only a lawyer could conceive of, a ruling was required
about whether an ATM constitutes a bank branch or not. Since such rul-
ings were made on a state by state basis, it varies across the country.
This results in some very odd arrangements in some states, because of
requirements placed on bank branches.

In early attempts, the card actually carried account information and
balances. The cardholder would bring the card into a branch, and bank
personnel would "load" money onto the card, based on the customer's ac-
tual account balance. The cardholder could then use the card at a
stand-alone machine that would update the information on the card as
money was withdrawn. The information was stored on track 3 of the mag-
netic stripe, as mentioned in an earlier installment. This approach
had many problems. It was far too susceptible to fraud, it could not
reasonably handle multiple accounts, and it could not be used as a ve-
hicle for other services. Since it was pretty much limited to with-
drawals, it didn't even automate much of the bank branch functions.

The online ATM offered a solution to the problems of the early ATM
cards. Since the ATM was connected to the bank's host, it was no
longer necessary to maintain account balances on the card itself, which
removed a major source of fraud. Also, access to multiple accounts be-
came possible, as did additional services, such as bill payment.

Once banks started buying and installing ATMs, they quickly realized
that it is very expensive to maintain a large number of machines. Yet
customers began demanding more machines, so they could have easier ac-
cess to their funds. Since many banks in an area would have ATMs, the
obvious solution was to somehow cross-connect bank hosts so that cus-
tomers could use ATMs at other banks, for convenience. The lawyers
struck again. Does a shared ATM count as a branch for both banks? Does
a transaction at a shared ATM mean that one bank is doing financial
transactions for another, which is not allowed? If two banks share
ATMs, but refuse to allow a third bank, is that monopolizing or re-
straint of trade? Strange restrictions on shared ATM transactions re-
sulted.

Soon interchange standards began to evolve, and ATM networks became a
competitive tool. Regional and national networks started to emerge.
And the lawyers struck again. If a network allows transactions in one
state for a bank in another state, isn't that interstate banking, which
was at the time forbidden? Should an ATM network that dominates a re-
gion become a regulated monopoly? Should an ATM network that gets re-
ally big be considered a public utility?

Today, the regional and national networks continue to grow and offer
more services and more interconnections. All of the regulatory issues
have not been resolved, and this is creating a lot of tension for eas-
ing banking restrictions.

An ATM card is just an ATM card, regardless of how many ATMs it works
in. Most banks long ago saw an opportunity for the ATM card to be used
as a debit card, presumably to replace checks. A tremendous number of
checks are used each year, and it costs banks a lot of money to process
them. Debit card transactions could cost less to process, given an ap-
propriate infrastructure. Some of the costs could potentially be
passed on to the merchants or the consumers, who are notoriously reluc-
tant to directly pay the cost of checks. So far there have been many
trials of using ATM cards as debit cards at the point of sale, but they
have, in general, met with consumer apathy. In some areas, where banks
have aggressively promoted debit, things have gone better. Still, gen-
eral acceptance of debit seems a ways off.

One interesting twist to the debit card story, as mentioned earlier, is
the emergence of third party debit cards. Issuers of these cards have
no real account relationship with the cardholders. Instead, they ob-
tain permission from the cardholders to debit their checking accounts
directly through the Automated Clearing Houses (ACHs), the same way
checks are cleared. (Think of it as direct deposit, in reverse.) Oil
companies first started experimenting with this a couple of years ago,
and it has met with surprising success. Banks dislike this concept,
because it competes directly with their debit cards, but isn't subject
to the same state and federal regulations. ACHs like this, because it
bolsters their business, which otherwise stands to lose a lot by
acceptance of debit cards. Merchants generally like this, especially
the large retailers, because it allows them to get their payment sys-
tems out from under the control of the banks.


THE ATM
--- ---

An ATM is an interesting combination of computer, communication, bank-
ing, and security technology all in one box. A typical machine has a
microprocessor, usually along the lines of an 8086, a communications
module (which may have it's own microprocessor), a security module
(also with a microprocessor), and special-purpose controllers for the
hardware. The user interface is typically a CRT, a telephone-style
keypad, and some soft function keys. Typically there is a lot of
memory, but no disk. The screens and program are usually downloaded
from the host at initialization, and are stored in battery-backed RAM
indefinitely. The machine typically interacts with the host for every
transaction, but it can operate offline if necessary, as dictated by
the downloaded program. The downloaded program is often in an
industry-standard "states and screens" format that was created by
Diebold, a manufacturer of various banking equipment, including ATMs.

Most machines can use a few IBM protocols (bisync, SNA, and an outmoded
but still used "loop" protocol), Burroughs poll/select, and perhaps
some others, depending on which communications module is in place.
This allows the manufacturer to make a standard machine, and plug in
different communications hardware to suit the customer. The IBM bisync
and SNA protocols are most common, with most networks moving toward
SNA.

The security modules do all encryption for the ATM. They are separate
devices that are physically sealed and cannot be opened or tapped with-
out destroying the data within them. In a truly secure application, no
sensitive data entering or leaving the security module is in cleartext.
Arranging this and maintaining it is more complicated than I can go
into here.

Most ATMs contain two bill dispensers, a "divert" bin for bills, a
"capture" bin for cards, a card reader, receipt printer, journal
printer, and envelope receptacle. Some ATMs have more than two bill
dispensers, and can even dispense coins.

When an ATM is dispensing money, it counts the appropriate bills out of
the bill dispensers, and uses a couple of mechanical and optical checks
to make sure it counted correctly. If the checks fail, it shunts the
bills into the divert bin and tries again. Typically, this is because
two bills were stuck together. I've seen ATMs have sensor faults, and
divert the total contents of both bill dispensers the first time a user
asks for a withdrawal. "Gee, all I did was ask for $50, and this ma-
chine made all kinds of funny whirring noises and shut down." Most
banks will put twenty-dollar bills in one of the dispensers and five
dollar bills in the other. Some use tens and fives, or tens and twen-
ties. Depending on the denominations of the bills, the size of the
dispensers, and the policy of the bank, an ATM can hold tens of thou-
sands of dollars.

The journal printer keeps a running log of every use of the machine,
and exactly what the machine is doing, for audit purposes. you can of-
ten hear it printing as soon as you put your card in or after your
transaction is complete.

When you put an envelope into an ATM, the transaction information is
usually printed directly on the envelope, so that verifying the deposit
is easier. Bank policies typically require that any deposit envelope
be opened and verified by two people. In this, you're actually safer
depositing cash at an ATM than giving it to a human teller.

A card will be diverted to the capture bin if it is on the "hot card"
list, if the user doesn't enter a correct PIN, or if the user walks
away and forgets to take the card.

On some machines, the divert bin, capture bin, envelope receptacle, and
bill dispenser bins are all separately locked containers, so that re-
stocking can be done by courier services who simply swap bins and re-
turn the whole thing to a central site.

The entire ATM is typically housed in a hardened steel case with alarm
circuitry built in. These suckers have been known to survive dynamite
explosions. The housing typically has a combination lock on the door,
and no single person knows the entire combination. The machine can
thus be opened for restocking, maintenance, or repair, only if at least
two people are present.

DEBIT CARD PROCESSING
----- ---- ----------

Debit card processing is fairly similar to credit and charge card pro-
cessing, with a few exceptions. First, in the case of ATMs, the ac-
cepter and acquirer are usually the same. For debit card use at the
point of sale, the usual acquirer-accepter relationship holds. In gen-
eral, acquirers may do front-end screening on debit cards, but all ap-
provals are generated by the issuer - the floor limit is zero. This
makes it possible to eliminate a separate settlement process for debit
card transactions, but places additional security and reliability con-
straints on the "authorization". Often a separate settlement is done
anyway.

One problem that has caused difficulties for POS use of debit cards is
the use of PINs. Many merchants and cardholders would rather use sig-
nature for identity verification. But most debit systems grew out of
ATM systems, and require PINs. This is an ironic reversal of the early
ATM card days, when people were trying to avoid requiring signature.
Other than the PIN, the information required for a debit transaction is
the same as that required for a credit transaction.

One last installment on the networks that tie this all together, and
the Credit Card 101 course will be complete. There will be no final
exam - you will be graded entirely on classroom participation. Most of
you are failing miserably...


Joe Ziegler
att!lznv!ziegler
Part 6 - Networks


ACCESS NETWORKS
------ --------

For most credit card applications, the cost of the access network is
the single biggest factor in overall costs, often accounting for over
half of the total. For that reason, there are many different solu-
tions, depending on the provider, the application, and geographical
constraints.

The simplest form of access network uses 800 service, in one of its
many forms. Terminals at merchant locations across the country dial an
800 number that is terminated on a large hunt group of modems, con-
nected directly to the acquirer's front-end processor (FEP). The FEP
is typically a fault-tolerant machine, since an outage here will take
out the entire service. A large acquirer will typically have two or
more centers for terminating the 800 service. This allows better
economy, due to the nature of 800 service tariffs, and allows for di-
saster recovery in case of a failure of one data center. An advantage
of 800 service is that it is quite easy to cover the entire country
with it. It also provides the most effective utilization of your FEP
resources. (A little queuing theory will show you why.) However, 800
service is quite expensive. It always requires 10 (or 11) digits di-
aled, and in areas with pulse dialing it can take almost three seconds
just to dial 1-800. The delay between dialing and connection is longer
for 800 calls than many other calls, because of the way the calls get
routed. All of this adds to the perceived response time at the mer-
chant location, even though the acquirer has no control over it.

Large acquirers prefer to offer some form of local access service. In
this service, terminals at the merchants dial a local telephone number
to gain access to the acquirer. Typically, the local number actually
connects to a packet network, which then connects to the acquirer. If
the packet network is a public network, the terminal must go through a
login sequence to get connected across the packet network. Typically,
local calls are much less expensive than 800 service calls, and local
calls typically connect faster than 800 calls. The cost of those calls
are absorbed by the merchants directly. In those few remaining areas
where local calls are still free from a business line, this works out
well for the merchant. Otherwise, the merchant can end up spending a
lot of money on phone calls. Usually, the acquirer has to offer lower
prices to accepters who use local calls, to help offset this. Even so,
these networks are generally much less expensive for the acquirers.
Such networks are difficult to maintain, due to the distributed nature
of the access network. Since most packet networks are much more likely
to experience failures than the phone network is, the merchant's POS
terminal is usually programmed to dial an 800 number for fallback if
the local number doesn't work. Also, it is generally not cost-effec-
tive to cover every free calling area in the entire country with access
equipment, so some 800 service is required anyway. There is also an
administrative headache associated with keeping track of the different
phone numbers that each merchant across the country needs to dial.
When you have tens of thousands of terminals to support, this can be
formidable.

Acquirers are beginning to experiment with Feature Group B (FGB) ac-
cess. FGB access was the method of access used to get to alternative
long-distance carriers before "equal access" was available. The
tariffs are still on the books, and they are favorable for this appli-
cation. FGB access provides a single number, nationwide, for all mer-
chants to dial in order to gain access to the acquirer. The call has
simpler (hence, presumably, faster) routing than 800 service, and the
call is charged to the acquirer, not the accepter. FGB access does
have to terminate on equipment that is physically located in the Local
Access Toll Area (LATA) where the call originated, so there is the
problem of having distributed equipment, as above. This also implies
that it is not cost-effective to deploy FGB access everywhere, as well.
There are also some technical oddities of FGB, due to its original in-
tent, that have made it difficult to implement so far.

The other big switched access capability that is likely to have an im-
pact in the future is ISDN. So far, this has been inhibited by limited
availability and lack of adequate equipment on the merchant end, but it
could be very beneficial when these problems are solved.

Private-line networks are pretty straightforward applications of
point-to-point and multipoint private lines. Since private lines are
quite expensive, engineering of the networks is challenging. Usually,
sophisticated software is used to determine the optimum placement of
concentrators in order to minimize costs. Since tariffs, real estate
prices, and business needs change frequently, maintaining a stable,
cost-effective network is hard work. A typical asynchronous private
line network will have multiplexers at remote sites, with backbone
links to companion multiplexers at a central site. Synchronous private
line networks may use multiplexers, or remote controllers, or remote
FEPs, depending on the application and the availability of real estate.

INTERCHANGE NETWORKS
----------- --------

Interchange networks physically consist mostly of point-to-point pri-
vate lines. In many of the large interchange networks, there is a cen-
tral "switch" that takes transactions from acquirers (thereby acting as
an issuer), and routes them to issuers (thereby acting as an acquirer).
Often the switch provider will actually be an acquirer or issuer as
well, but this is not always the case. Usually, the provider of the
switch defines standard message formats, protocols, and interchange
rules. These formats and protocols usually comply with national and
international standards, but sometimes do not. Often the switch will
provide translation between different message formats and protocols.

The switch provider is generally very concerned that settlement com-
plete successfully. Failure to settle with one or more large issuers
can leave the switch provider with an overnight deficit of a couple
million dollars. Even though this is a temporary situation, it has
significant financial impact.

In some current networks, authorization and settlement take place on
completely separate facilities, with separate hosts in some cases.
This is mainly due to the history of the industry in this country. Re-
call that authorizations were originally done by voice calls, and
settlement was done by moving paper around. These two processes were
automated at different times, by separate means. Thus VISA has a BASE
1 network for authorization, and a BASE 2 network for settlement.
Likewise, MasterCard has INET and INES, one for authorization and one
for settlement. These functions are becoming less and less separated
as communication and computer facilities evolve, and will probably be
completely integrated over the next five to ten years.

Interchange networks are probably the most volatile part of the ATM
market right now. There is currently a shakeout going on in much of
the market, with larger, more aggressive regionals buying out
standalone networks and smaller regionals. This causes local banks to
change local and national network affiliation from time to time. So a
card may work in a given ATM one day, but fail in that machine the
next, which confuses many consumers. Most large regional and national
networks have operating regulations requiring labeling of ATMs and
cards, so that if you see the same logo on your card and the ATM, you
can be pretty sure it will work.

Some regionals are interconnected, and others are not. The two biggest
nationals, Cirrus and Plus, have operating regulations that effectively
prohibit a member of one network from connecting to the other. But a
regional on Cirrus could be connected to a regional on Plus. In that
case, whether a machine will take your ATM card depends on the routing
algorithm used. In most cases, the acquirer will have a table of issu-
ers that are directly connected, and will send anything else to the re-
gional switch. The regional switch will have a table of each issuer
it is directly connected to, and tables of which cards are acceptable
to other regionals it interchanges with. Anything else goes to the na-
tional switch. The same process happens in reverse from there. Often
the order of search in the routing tables is determined by fee scales,
not geography, so transactions can be routed in completely non-obvious
ways.

So the easiest way to tell if your card will work in a given ATM is to
stick the card in and try. I don't know of any machine that will eat a
card just because it can't route the transaction - it will generally
give some non-specific message about being unable to complete the
transaction and spit the card back out. Of course, if the transaction
is completed from a machine that you're not sure of, you also aren't
sure what the fee is going to be if your bank passes those fees on to
you. Sometimes the fee will be printed on the receipt, but usually it
isn't. If you do the transaction in a foreign country, you may not
know the exchange rate used. (I once couldn't balance my checkbook for
a month until I got a statement with the transaction I did at Banc du
Canada in Montreal.) But if you need the money and are willing to pay
the fee, you have little to lose by trying out just about any ATM.

This completes the course in Credit Card 101. Hope you all found it
enjoyable and informative.


Joe Ziegler
att!lznv!ziegler


X-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-X

Another file downloaded from: NIRVANAnet(tm)

& the Temple of the Screaming Electron Jeff Hunter 510-935-5845
The Salted Slug Strange 408-454-9368
Burn This Flag Zardoz 408-363-9766
realitycheck Poindexter Fortran 415-567-7043
Lies Unlimited Mick Freen 415-583-4102
Tomorrow's 0rder of Magnitude Finger_Man 415-961-9315
My Dog Bit Jesus Suzanne D'Fault 510-658-8078
New Dork Sublime Demented Pimiento 415-566-0126

Specializing in conversations, obscure information, high explosives,
arcane knowledge, political extremism, diverse sexuality,
insane speculation, and wild rumours. ALL-TEXT BBS SYSTEMS.

Full access for first-time callers. We don't want to know who you are,
where you live, or what your phone number is. We are not Big Brother.

"Raw Data for Raw Nerves"

X-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-X

This is part two in a planned six-part series about the credit card in-
dustry. It would be best if you read part one before reading this
part. Enjoy.

DEFINITIONS
-----------

Some more new terms that are used in this posting.

ABA - American Bankers Association

ACH - Automated Clearing House - an organization that mechanically and
electronically processes checks.

ANSI - American National Standards Institute

Embossing - creating raised letters and numbers on the face of the
card.

Encoding - recording data on the magnetic stripe on the back of the
card.

Imprinting - using the embossed information to make an impression on a
charge slip.

Interchange - sending authorization requests from one host (the
acquirer) to another (the issuer) for approval.

ISO - International Standards Organization

NACHA - National Automated Clearing House Association

PAN - Personal Account Number. The account number associated with a
credit, debit or charge card. This is usually the same as the
number on the card.

PIN - Personal Identification Number. A number associated with the
card, that is supposedly know only to the cardholder and the card
issuer. This number is used for verification of cardholder
identity.


THE ORGANIZATIONS
--- -------------

ISO sets standards for plastic cards and for data interchange, among
other things. ISO standards generally allow for national expansion.
Typically, a national standards organization, like ANSI, will take an
ISO standard and develop a national standard from it. National stan-
dards are generally subsets of the ISO standard, with extensions as al-
lowed in the original ISO standard. Many credit card standards
originated in the United States, and were generalized and adopted by
ISO later.

The ANSI committees that deal with credit card standards are sponsored
by the ABA. Most members of these committees work for banks and other
financial institutions, or for vendors who supply banks and financial
institutions. Working committees report to governing committees.

All standards go through a formal comment and review procedure before
they are officially adopted.


PHYSICAL STANDARDS
-------- ---------

ANSI X4.13, "American National Standard for Financial Services -
Financial Transaction Cards" defines the size, shape, and other
physical characteristics of credit cards. Most of it is of interest
only to mechanical engineers. It defines the location and size of the
magnetic stripe, signature panel, and embossing area. This standard
also includes the Luhn formula used to generate the check digit for the
PAN, and gives the first cut at identifying card type from the account
number. (This part was expanded later in other standards.) Also, this
standard identifies the character sets that can be used for embossing a
card.

Three character sets are allowed - OCR-A as defined in ANSI X3.17,
OCR-B as defined in ANSI X3.49, and Farrington 7B, which is defined in
the appendix of ANSI X4.13 itself. Almost all the cards I have use
Farrington 7B, but Sears uses OCR-A. (Sears also uses the optional,
smaller card size as, allowed in the standard.) These character sets
are intended to be used with optical character readers (hence the OCR),
and large issuers have some pretty impressive equipment to read those
slips.


ENCODING STANDARDS
-------- ---------

ANSI X4.16, "American National Standard for Financial Services - Finan-
cial Transaction Cards - Magnetic Stripe Encoding" defines the
physical, chemical, and magnetic characteristics of the magnetic stripe
on the card. The standard defines a minimum and maximum size for the
stripe, and the location of the three defined encoding tracks. (Some
cards have a fourth, proprietary track.)

Track 1 is encoded at 210 bits per inch, and uses a 6-bit coding of a
64-element character set of numerics, alphabet (one case only), and
some special characters. Track 1 can hold up to 79 characters, six of
which are reserved control characters. Included in these six charac-
ters is a Longitudinal Redundancy Check (LRC) character, so that a card
reader can detect most read failures. Data encoded on track 1 include
PAN, country code, full name, expiration date, and "discretionary
data". Discretionary data is anything the issuer wants it to be.
Track 1 was originally intended for use by airlines, but many Automatic
Teller Machines (ATMs) are now using it to personalize prompts with
your name and your language of choice. Some credit authorization ap-
plications are starting to use track 1 as well.

Track 2 is encoded at 75 bits per inch, and uses a 4-bit coding of the
ten digits. Three of the remaining characters are reserved as
delimiters, two are reserved for device control, and one is left unde-
fined. In practice, the device control characters are never used, ei-
ther. Track 2 can hold up to 40 characters, including an LRC. Data
encoded on track 2 include PAN, country code (optional), expiration
date, and discretionary data. In practice, the country code is hardly
ever used by United States issuers. Later revisions of this standard
added a qualification code that defines the type of the card (debit,
credit, etc.) and limitations on its use. AMEX includes an issue date
in the discretionary data. Track 2 was originally intended for credit
authorization applications. Nowadays, most ATMs use track 2 as well.
Thus, many ATM cards have a "PIN offset" encoded in the discretionary
data. The PIN offset is usually derived by running the PIN through an
encryption algorithm (maybe DES, maybe proprietary) with a secret key.
This allows ATMs to verify your PIN when the host is offline, generally
allowing restricted account access.

Track 3 uses the same density and coding scheme as track 1. The con-
tents of track 3 are defined in ANSI X9.1, "American National Standard
- Magnetic Stripe Data Content for Track 3". There is a slight contra-
diction in this standard, in that it allows up to 107 characters to be
encoded on track 3, while X4.16 only gives enough physical room for 105
characters. Actually, there is over a quarter of an inch on each end
of the card unused, so there really is room for the data. In practice,
nobody ever uses that many characters, anyway. The original intent was
for track 3 to be a read/write track (tracks 1 and 2 are intended to be
read-only) for use by ATMs. It contains information needed to maintain
account balances on the card itself. As far as I know, nobody is actu-
ally using track 3 for this purpose anymore, because it is very easy to
defraud.


COMMUNICATION STANDARDS
------------- ---------

Formats for interchange of messages between hosts (acquirer to issuer)
is defined by ANSI X9.2, which I helped define. Financial message au-
thentication is described by ANSI X9.9. PIN management and security is
described by ANSI X9.8. There is a committee working on formats of
messages from accepter to acquirer. ISO has re-convened the interna-
tional committee on host message interchange (TC68/SC5/WG1), and ANSI
may need to re-convene the X9.2 committee after the ISO committee fin-
ishes. These standards are still evolving, and are less specific than
the older standards mentioned above. This makes them somewhat less
useful, but is a natural result of the dramatic progress in the indus-
try.

ISO maintains a registry of card numbers and the issuers to which they
are assigned. Given a card that follows standards (Not all of them
do.) and the register, you can tell who issued the card based on the
first six digits (in most cases). This identifies not just VISA,
MasterCard, etc., but also which member bank actually issued the card.


DE FACTO INDUSTRY STANDARDS
-- ----- -------- ---------

Most ATMs use IBM synchronous protocols, and many networks are migrat-
ing toward SNA. There are exceptions, of course. Message formats used
for ATMs vary with the manufacturer, but a message set originally de-
fined by Diebold is fairly widely accepted.

Many large department stores and supermarkets (those that take cards)
run their credit authorization through their cash register controllers,
which communicate using synchronous IBM protocols.

Standalone Point-of-Sale (POS) devices, such as you would find at most
smaller stores (i.e. not at department stores), restaurants and hotels
use a dial-up asynchronous protocol devised by VISA. There are two
generations of this protocol, with the second generation just beginning
to get widespread acceptance.

Many petroleum applications use multipoint private lines and a polled
asynchronous protocol known as TINET. This protocol was developed by
Texas Instruments for a terminal of the same name, the Texas Instru-
ments Network E(something) Terminal. The private lines reduce response
time, but cost a lot more money than dial-up.

NACHA establishes standards for message interchange between ACHs, and
between ACHs and banks, for clearing checks. This is important to this
discussion due to the emergence of third-party debit cards, as dis-
cussed in part 1 of this series. The issuers of third-party debit
cards are connecting to ACHs, using the standard messages, and clearing
POS purchases as though they were checks. This puts the third parties
at an advantage over the banks, because they can achieve the same re-
sults as a bank debit card without the federal and state legal restric-
tions imposed on banks.

In the next installment, I'll describe how an authorization happens, as
well as how the settlement process gets the bill to you and your money
to the merchant. After that I'll describe various methods of fraud,
and how issuers, acquirers, and accepters protect themselves. Stay
tuned.

Joe Ziegler
att!lznv!ziegler

Here's part 3 in my six-part series on the credit card industry. This
part discusses how authorization and settlement work. This is a long
one. It will help if you have read parts 1 and 2, since I had to leave
out a lot of overlap to keep this from getting ridiculous. Enjoy.


THE ACCEPTER
--- --------

An important fact to note is that a card accepter does not have to get
approval for any purchases using credit or charge cards. Of course, a
merchant is usually interested in actually getting money, and so must
participate in some form of settlement process (see below). Usually,
the most acceptable (to a merchant) forms of settlement are tied (by
the acquirer) to authorization processes. However, a merchant could
simply accept all cards without any validation, any eat any fraud that
results.

A merchant typically makes a business arrangement with a local bank or
some other acquirer for authorization and settlement services. The
acquirer assigns a merchant identifier to that merchant, which will
uniquely identify the location of the transaction. (This facilitates
compliance with federal regulations requiring that credit card bills
identify where each purchase was made.) The acquirer also establishes
procedures for the merchant to follow. The procedures will vary by
type of the merchant business, geographic location, volume of transac-
tions, and types of cards accepted.

If the merchant follows the procedures given by the acquirer and a
transaction is approved, the merchant is guaranteed payment whether the
card in question is good or bad. The purpose of authorization is to
shift financial liability from the acceptor to the acquirer.

There are two basic tools used - bulletins and online checks. Bulletins
may be hardcopy, or may be downloaded into a local controller of some
form. Online checks could be done via a voice call, a standalone ter-
minal, or software and/or hardware integrated into the cash register.

A low-volume, high-ticket application (a jewelry store) would probably
do all its authorizations with voice calls, or may have a stand-alone
terminal. A high-volume, low-ticket application (a fast-food chain)
will probably do most of its authorizations locally against a bulletin
downloaded into the cash register controller. Applications in between
typically merge the two - things below a certain amount (the "floor
limit") are locally authorized after a lookup in the bulletin, while
things over the floor limit are authorized online.

Usually a lot of effort is taken to use the least expensive tools that
are required by the expected risk of fraud. Typically, communication
costs for authorizations make up the biggest single item in the overall
cost of providing credit cards.

Large accepters are always a special case. Airlines are usually di-
rectly connected, host-to-host, to issuers and/or acquirers, and autho-
rize everything online. Likewise for many petroleum companies and
large department stores. Some large chains use different approaches at
different locations, either as a result of franchising oddities or due
to volume differences between locations. A lot of experimentation is
still going on as well - this is not a mature market.

For voice authorizations, the merchant ID, PAN, expiration date, and
purchase amount are required for an approval. Some applications also
require the name on the card, but this is not strictly necessary. For
data authorizations, the merchant ID, PAN, PIN (if collected), expira-
tion date, and purchase amount are required. Typically, the "discre-
tionary data" from track 2 is sent as well, but this is not strictly
necessary. In applications that do not transmit the PIN with the au-
thorization, it is the responsibility of the merchant to verify iden-
tity. Usually, this should be done by checking the signature on the
card against the signature on the form. Merchants don't often follow
this procedure, and they take a risk in not doing so.

In most applications, the amount of the purchase is known at the time
of the authorization request. For hotels, car rentals, and some petro-
leum applications, an estimated amount is used for the authorization.
After the transaction is complete (e.g. after the gas is pumped or at
check-out time), another transaction may be sent to advise of the ac-
tual amount of the transaction. More on this later.


THE ACQUIRER
--- --------

The acquirer gathers authorization requests from accepters and returns
approvals. If the acquirer is an issuer as well, "on us" transactions
will typically be turned around locally. As before, the acquirer does
not have to forward any requests on to the actual issuer. However,
acquirers are not willing to take the financial risks associated with
generating local approvals. Thus most transactions are sent on to the
issuers (interchanged). The purpose of interchange is to shift finan-
cial liability from the acquirer to the issuer.

Typically, an acquirer connects to many issuers, and negotiates differ-
ent business arrangements with each one of them. But the acquirer gen-
erally provides a uniform interface to the accepter. Thus, the
interchange rules are sometimes less stringent than those imposed on
the accepter. Also, most issuers will trust acquirers to with respon-
sibilities they would never trust to accepters. The acquirer can
therefore perform some front-end screening on the transactions, and
turn some of them around locally without going back to the issuer.

The first screening by the acquirer would be a "sanity" test, for valid
merchant ID, valid Luhn check on PAN, expiration date not past, amount
field within reason for type of merchant, etc. After that, a floor
limit check will be done. Issuers generally give acquirers higher
floor limits than acquirers give accepters, and floor limits may vary
by type of merchant. Next, a "negative file" check would be done
against a file of known bad cards. (This is essentially the same as
the bulletin.) Then a "velocity file" check may be done. A velocity
file keeps track of card usage, and limits are often imposed on both
number of uses and total amount charged within a given time period.
Sometimes multiple time periods are used, and it can get fairly compli-
cated.

Transactions that pass all the checks, and are within the authority
vested in the acquirer by the issuer, are approved by the acquirer.
(Note that, under the business arrangement, financial liability still
resides with the issuer.) An "advice" transaction is sometimes sent to
the issuer (perhaps at a later time), to tell the issuer that the
transaction took place.

Transactions that "fail" one or more checks are denied by the acquirer
(if the cause was due to form, such as bad PAN) or sent to the issuer
for further checking. (Note that "failure" here can mean that it's be-
yond the acquirer's authority, not necessarily that the card is bad.)
Some systems nowadays will periodically take transactions that would
otherwise be approved locally, and send them to the issuer anyway. This
serves as a check on the screening software and as a countermeasure
against fraudulent users who know the limits.

Transactions that go to the issuer are routed according to the first
six digits of the PAN, according to the ISO registry mentioned in an
earlier section. Actually, it's a bit more complicated than that,
since there can be multiple layers of acquirers, and some issuers or
acquirers will "stand in" for other issuers when there are hardware or
communication failures, but the general principal is the same at each
point.


THE ISSUER
--- ------

An issuer receiving an interchanged transaction will often perform many
of the same tests on it that the acquirer performs. Some of the tests
may be eliminated if the acquirer is trusted to do them correctly. This
is the only point where a velocity file can actually detect all usage
of a card. This is also the only point where a "positive file" lookup
against the actual account can be done, since only the issuer has the
account relationship with the cardholder. If a PIN is used in the
transaction, only the issuer can provide true PIN verification -
acquirers may be able to do only "PIN offset" checking, as described in
a previous section. This is one reason why PINs have not become
popular on credit and charge cards.

An account typically has a credit limit associated with it. An ap-
proved authorization request usually places a "hold" against the credit
limit. If the sum of outstanding holds plus the actual outstanding
balance on the account, plus the amount of the current transaction, is
greater than the credit limit, the transaction is (usually) denied.
Often in such a case the issuer will send back a "call me" response to
the merchant. The merchant will then call the issuer's number, and the
operator may even want to talk to the cardholder. The credit limit
could be extended on the spot, or artificially high holds (from hotels
or car rental companies) could be overlooked so that the transaction
can be approved.

The difference between the credit limit and the sum of holds and out-
standing balance is often referred to as the "open to buy" amount. Once
a hold is placed on an account, it is kept there until the actual the
transaction in question is settled (see below), in which case the
amount goes from a hold to a billed amount, with no impact on the open
to buy amount, theoretically. For authorizations of an estimated
amount, the actual settled amount will be less than or equal to the ap-
proved amount. (If not, the settlement can be denied, and the merchant
must initiate a new transaction to get the money.) Theoretically, in
such a case, the full hold is removed and the actual amount is added to
the outstanding balance, resulting in a possible increase in the open
to buy amount.

In practice, older systems were not capable of matching settlements to
authorizations, and holds were simply expired based on the time it
would take most transactions to clear. Newer systems are starting to
get more sophisticated, and can do a reasonable job of matching autho-
rizations for actual amounts with the settlements. Some of them still
don't match estimated amounts well, with varying effects. In some
cases, the difference between actual and estimated will remain as a
hold for some period of time. In other cases, both the authorization
and the settlement will go against the account, reducing the open to
buy by up to twice the actual amount, until the hold expires. These
problems are getting better as the software gets more sophisticated.

Some issuers are also starting to use much more sophisticated usage
checks as well. They will not only detect number of uses and amount
over time, but also types of merchandise bought, or other patterns to
buying behavior. Most of this stuff is new, and is used for fraud pre-
vention. I expect this to be the biggest effort in authorization soft-
ware for the next few years.

American Express does things completely differently. There are no
credit limits on AMEX cards. Instead, AMEX relies entirely on usage
patterns, payment history, and financial data about cardmembers to de-
termine whether or not to automatically approve a transaction. AMEX
also has a policy that a cardmember will never be denied by a machine.
Thus, if the computer determines that a transaction is too risky, the
merchant will receive a "call me" message. The operator will then get
details of the transaction from the merchant, and may talk to the
cardmember as well, if cardmember identity is in question or a large
amount is requested. To verify cardmember identity, the cardmember
will be asked about personal information from the original application,
or about recent usage history. The questions are not the same each
time. If an unusually large amount is requested, the cardmember may be
asked for additional financial data, particularly anything relating to
a change in financial status (like a new job or a promotion). People
who are paranoid about Big Brother and computer databases should not
use AMEX cards.


SETTLEMENT
----------

So far, no money has changed hands, only financial liability. The pur-
pose of settlement is to shift the financial liability back to the
cardholder, and to shift the cardholder's money to the merchant.
Theoretically, all authorization information can be simply discarded
once an approval is received by a merchant. Of course, contested
charges, chargebacks, merchant credits, and proper processing of holds
require that the information stay around. Still, it is important to
realize that an authorization transaction has no direct financial con-
sequences. It only establishes who is responsible for the financial
consequences to follow.

Traditionally, a merchant would take the charge slips to the bank that
was that merchant's acquirer, and "deposit" them into the merchant ac-
count. The acquirer would take the slips, sort them by issuer, and
send them to the issuing banks, receiving credits by wire once they ar-
rived and were processed. The issuer would receive the slips, micro-
film them (to save the transaction information, as required by federal
and state laws) charge them against the cardholder's accounts, send
credits by wire to the acquirer, and send out the bill to the
cardholder. Problem is, this took time. Merchants generally had to
wait a couple of weeks for the money to be available in their accounts,
and issuers often suffered from float on the billables of about 45
days.

Therefore, nowadays many issuers and acquirers are moving to on-line
settlement of transactions. This is often called "draft capture" in
the industry. There are two ways this is done - one based on the host
and one based on the terminal at the merchant's premises. In the
host-based case, the terminal generally only keeps counts and totals,
while the acquirer host keeps all the transaction details. Peri-
odically, the acquirer host and the terminal communicate, and verify
that they both agree on the data. In the terminal-based case, the ter-
minal remembers all the important transaction information, and peri-
odically calls the acquirer host and replays it all for several
transactions. In either case, once the settlement is complete the mer-
chant account is credited. The acquirer then sends the settlement in-
formation electronically to the issuers, and is credited by wire
immediately (or nearly so). The issuer can bill directly to the
cardholder account, and float can be reduced to an average of 15 days.

The problem is, what to do with the paper? Current regulations in many
states require that it be saved, but there is no need for it to be sent
to the issuer. Also, for contested charges, a paper trail is much more
likely to stand up in court, and much better to use for fraud investi-
gations. Currently, the paper usually ends up back at the issuer, as
before, but it doesn't need to be processed, just microfilmed and
stored.

Much of the market still uses paper settlement methods. Online settle-
ment will replace virtually all of this within the next 5 to 10 years,
because of its many benefits.

This was pretty long, but there is a lot of information, and I skimmed
over a lot of details. Future installments should be shorter. Coming
up next is a discussion of fraud and security, and then a special dis-
cussion of debit cards. Hang on, we're halfway through this!


Joe Ziegler
att!lznv!ziegler

This is part four of a planned six-part series on the credit card in-
dustry. It will be helpful if you have read parts one through three,
as I use a lot of terminology here that was introduced earlier. Enjoy.

WARNING

This installment describes various methods of perpetrating fraud
against credit and charge card issuers, acquirers, and cardholders. Le-
gal penalties for using these methods to commit fraud are severe. The
reason for sharing this information is so that consumers will be aware
of the importance of security and be aware of the procedures used by
financial institutions to protect against fraud. Neither I nor my em-
ployer advocate use of the fraudulent methods described herein.

All the information here is publicly available from other sources. Un-
necessary detail is purposely not included, particularly as it applies
to detection and prevention of fraud.


CARDHOLDER FRAUD
---------- -----

The most common type of fraud against credit cards is cardholders fal-
sifying applications to get higher credit limits than they can afford
to pay, or to get multiple cards that they cannot afford to pay off.
Sometimes this is done with intent to defraud, but most often it is
done out of desperation or sheer financial ineptitude. Those who in-
tend to defraud generally use the multiple-card approach. They give
false names and financial data on several (sometimes as many as hun-
dreds) of applications. Often, the address of a vacant house that the
crook has access to is given, making it difficult to track the crook's
real identity. Once cards start showing up, the crook uses them for
cash advances or charges merchandise that is easy to sell, like con-
sumer electronics. The crook will run all the cards up to the limit
immediately, and will generally move on by the time the bills start ar-
riving. This type of fraud is not applicable to debit cards, since
they require an available account balance equal to or greater than any
purchases or withdrawals.

Protecting against this type of fraud, either intentional or otherwise,
is exactly the purpose of credit bureaus such as TRW. Issuers have be-
come more aware of the need for careful screening of applications, and
are using better techniques for detecting similar applications sent to
multiple issuers. More sophisticated velocity file screening can also
be used to detect possibly fraudulent usage patterns. Since this is a
method of fraud that can be used to gain really large amounts of
money, it is a high priority with issuers' security departments.

A variant of this scheme is much like check kiting. Can you use your
VISA to pay your MasterCard? Well, you might be able to manage it, but
if you're doing it with intent to defraud, you can be prosecuted. Kit-
ing schemes typically don't last long, have a low payoff, and are very
easy to detect.

Another type of cardholder fraud is simply contesting legitimate
charges. Most often, retrieving the documents gives pretty convincing
proof. Frequently, a family member will be found to have used the card
without the cardholder's permission. Such cases are usually pretty
easy to resolve. In the case of an ATM card, cameras are often placed
at ATMs (sometimes hidden) to record users of the machine. The camera
is usually tied to the ATM, so that a single retrieval stamp can be
placed on the film and the ATM log. If a withdrawal is contested, the
bank can then retrieve the picture of the person standing at the ma-
chine, and conclusively tie that picture to the transaction.

A type of cardholder fraud that is endemic only to ATMs is making false
deposits. You could, theoretically, tell the ATM that you are deposit-
ing a large amount of money, and put in an empty envelope. Most banks
will not let you withdraw amounts deposited into an ATM until the de-
posit has been verified, but some will allow part of the deposit to be
withdrawn. Typically, you can't get away with much. If you have any
money actually in your account, the bank has easy, legal recourse to
seize those funds. Most banks have no sense of humor about such
things, and will remove ATM card privileges after the first offense.


THIRD-PARTY FRAUD
----------- -----

The simplest way for a third party to commit fraud is for them to get
their hands on a legitimate card. There is a large black market for
credit cards obtained from hold-ups, break-ins and muggings. Perhaps
one of the cruelest methods of getting a card is a "Good Samaritan"
scam. In such a scam, credit cards are stolen by pick-pockets,
purse-snatchers, etc. That same day, someone looks up your number in
the phone book and calls you up. "I just found your wallet. All the
money is gone, but the credit cards and your driver's license are still
here. It just happens that I'll be in your neighborhood next Wednesday
and I'll drop it off then." Since the cards are found, you don't re-
port them stolen, and the crooks get until next Wednesday before you're
even suspicious. If such a thing happens to you, ask if you can come
and pick the cards up immediately. A true good samaritan won't mind,
but a crook will stall you. If you can't get your hands on the cards
immediately, report them as stolen. Most issuers will be able to get
you a new card by next Wednesday, anyway.

Often stolen cards will be used for a time exactly as is. The best
tool for preventing this is verification of the signature, but this is
ineffective because most merchants don't consistently check signatures
and some people don't even sign their cards. (I guess these people
figure that all purse snatchers are accomplished forgers as well.)
Many cards will eventually be modified as the various security schemes
start catching up.

It is a very easy matter, for example, to re-encode a different number
on the magnetic stripe. Since the card still looks fine, a merchant
will accept it and run it through the POS terminal, completely ignorant
of the fact that the number read off the back is not the same as that
on the front. Although the number on the front would fail a negative
file check, the number on the back is one that hasn't been reported
yet. A card can be re-encoded almost any number of times, as long as
you can keep coming up with new valid PANs. To protect against this,
some merchants purposely avoid using the magnetic stripe. Others have
terminals that display the number read from the stripe, so the cashier
can compare it to the number on the card. Some issuers are experiment-
ing with special encoding schemes, to make re-encoding difficult, but
most of these schemes would require replacing the entire embedded base
of POS terminals. An interesting approach I've seen (it's probably
patented) uses a laser to burn off the parts of the magnetic stripe
where zeroes are encoded, leaving only the ones. This severely limits
the changes you can make to the card number. Some issuers use the
"discretionary data" field to encode data unique to the card, that a
crook would not be able to guess, to combat this type of fraud.

Since an ATM doesn't have a human looking at the card, it is especially
susceptible to re-encoding fraud. A crook could get a number from a
discarded receipt and encode it on a white card blank, which is easy to
obtain legally. Many people use PINs that are easy to guess, and the
crook has an easy job of it. Most ATMs will not give you your card
back if you don't enter a correct PIN, and will only give you a few
tries to get it right, to prevent this type of fraud. Velocity file
checks are also important in detecting this. You should always take
your ATM receipts with you, pick a non-obvious PIN, and make sure that
nobody sees you enter it.

One place that a crook can get valid PANs to encode on credit cards is
from dumpsters outside of stores and restaurants. The credit slip
typically is a multipart form, with one copy for you, one for the mer-
chant, and one for the issuer (ultimately). If carbon paper is used,
and the carbons are discarded intact, it's pretty easy to read the num-
bers off of them. Carbonless paper and forms that either rip the car-
bons in half or attach them to the cardholder copy automatically are
used to prevent this.

There are a lot of scams for getting people to tell their credit card
numbers over the phone. Never give your card number to anyone unless
you are buying something from them, and make sure that it is a le-
gitimate business you are buying from. "Incredible deal!! Diamond
jewelry at half price!! Call now with your VISA number, and we'll rush
you your necklace!!" When you don't get the necklace for four weeks,
you might start to wonder. When you get your credit card bill, you'll
stop wondering.

There are other, more sophisticated ways to modify a credit card. If
you're skillful, you can change the embossing on the card and even the
signature on the back. For most purposes, these techniques are more
trouble than they're worth, since it's not difficult to come up with a
new stolen card, or fake ID to match the existing card.


MERCHANT FRAUD
-------- -----

There are many urban rumors of merchants imprinting a card multiple
times while the cardholder isn't looking, and then running through a
bunch of charges after the cardholder leaves. I don't know of any case
where this is an official policy of a merchant, but this is certainly
one technique a dishonest cashier could use. The cashier can then take
home a bunch of merchandise charged to your account. Although some
people are afraid of this happening in a restaurant, where a waiter
takes your card away for a while, it's actually less likely there,
since there isn't anything the waiter can charge against your card and
take home.

A merchant could also make copies of charge slips, to sell the PANs to
other crooks. (See above for use of PANs.) Most credit card investi-
gation departments are sensitive to this possibility, and catch on real
fast if it's happening just by looking at usage history of cards with
fraudulent charges.

A merchant is also in a position to create many false charges against
bogus numbers, to attempt to defraud the acquirer or issuer. These
schemes are usually not too effective, since acquirers generally re-
spond very quickly to an unusual number of fraudulent transactions by
tightening restrictions on the merchant.


ACQUIRER AND ISSUER FRAUD
-------- --- ------ -----

The place to make really big bucks in fraud is at the acquirer or is-
suer, since this is where you can get access to large amounts of money.
Fortunately, it's also fairly easy to control things here with audit
procedures and dual control. People working in the back offices, pro-
cessing credit slips, bills, etc. have a big opportunity to "lose"
things, introduce false things, artificially delay things, and tempo-
rarily divert things. Most of the control is standard banking stuff,
and has been proven effective for decades, so this isn't a big problem.
A bigger potential problem to the consumer is the possibility of an em-
ployee at the issuer or acquirer selling PANs to crooks. This would be
very hard to track down, and could compromise a large part of the card
base. I know of no cases where this has happened.

Programmers, in particular, are very dangerous because they know where
the data is, how to get it, and what to do with it. In most shops, de-
velopment is done on completely separate facilities from the production
system. Certification and installation are done by non-developers, and
developers are not allowed any access to the production facilities.
Operations and maintenance staff are monitored very carefully as well,
since they typically have access to the entire system as part of their
jobs.

Another type of fraud that is possible here is diversion of materials,
such as printed, but not embossed or encoded, card blanks. Such mate-
rials are typically controlled using processes similar to those used at
U.S. mints. Since most of the cards issued in the United States are
actually manufactured by only a handful of companies, it's not too hard
to keep things under control.

There are many types of fraud that can be perpetrated by tapping data
communication lines, and using protocol analyzers or computers to in-
tercept or introduce data. These types of fraud are not widespread,
mainly because of the need for physical access and because sophisti-
cated computer techniques are required. There are message authentica-
tion, encryption, and key management techniques that are available to
combat this type of fraud, but currently these techniques are far more
costly than the minimal fraud they could prevent. About the only such
security technique that is in widespread use is encryption of PINs.

The next episode will be devoted to debit cards, and the final episode
will talk about the networks that make all this magic happen.


Joe Ziegler
att!lznv!ziegler

Part 5 - Debit Cards

EVOLUTION OF DEBIT CARDS
--------- -- ----- -----

The debit card originated as a method for bank customers to have access
to their funds through Automatic Teller Machines (ATMs). This was seen
as a way for banks to automate their branches and save money, as well
as a benefit for customers. A secondary intent was for the card to be
used as a method of identification when dealing with a human teller.
Although that idea never really caught on, it has seen renewed interest
from time to time.

One problem with using cards to access bank accounts is that federal
regulations required a signature be used for each withdrawal transac-
tion. After much debate, the concept of a Personal Identification Num-
ber (PIN) was invented, and federal regulations were modified to allow
PINs for use in place of signatures with bank withdrawals. ATMs also
faced many other regulatory difficulties. In many states, for example,
there are limitations on the number of branches a bank can have. In a
conflict that only a lawyer could conceive of, a ruling was required
about whether an ATM constitutes a bank branch or not. Since such rul-
ings were made on a state by state basis, it varies across the country.
This results in some very odd arrangements in some states, because of
requirements placed on bank branches.

In early attempts, the card actually carried account information and
balances. The cardholder would bring the card into a branch, and bank
personnel would "load" money onto the card, based on the customer's ac-
tual account balance. The cardholder could then use the card at a
stand-alone machine that would update the information on the card as
money was withdrawn. The information was stored on track 3 of the mag-
netic stripe, as mentioned in an earlier installment. This approach
had many problems. It was far too susceptible to fraud, it could not
reasonably handle multiple accounts, and it could not be used as a ve-
hicle for other services. Since it was pretty much limited to with-
drawals, it didn't even automate much of the bank branch functions.

The online ATM offered a solution to the problems of the early ATM
cards. Since the ATM was connected to the bank's host, it was no
longer necessary to maintain account balances on the card itself, which
removed a major source of fraud. Also, access to multiple accounts be-
came possible, as did additional services, such as bill payment.

Once banks started buying and installing ATMs, they quickly realized
that it is very expensive to maintain a large number of machines. Yet
customers began demanding more machines, so they could have easier ac-
cess to their funds. Since many banks in an area would have ATMs, the
obvious solution was to somehow cross-connect bank hosts so that cus-
tomers could use ATMs at other banks, for convenience. The lawyers
struck again. Does a shared ATM count as a branch for both banks? Does
a transaction at a shared ATM mean that one bank is doing financial
transactions for another, which is not allowed? If two banks share
ATMs, but refuse to allow a third bank, is that monopolizing or re-
straint of trade? Strange restrictions on shared ATM transactions re-
sulted.

Soon interchange standards began to evolve, and ATM networks became a
competitive tool. Regional and national networks started to emerge.
And the lawyers struck again. If a network allows transactions in one
state for a bank in another state, isn't that interstate banking, which
was at the time forbidden? Should an ATM network that dominates a re-
gion become a regulated monopoly? Should an ATM network that gets re-
ally big be considered a public utility?

Today, the regional and national networks continue to grow and offer
more services and more interconnections. All of the regulatory issues
have not been resolved, and this is creating a lot of tension for eas-
ing banking restrictions.

An ATM card is just an ATM card, regardless of how many ATMs it works
in. Most banks long ago saw an opportunity for the ATM card to be used
as a debit card, presumably to replace checks. A tremendous number of
checks are used each year, and it costs banks a lot of money to process
them. Debit card transactions could cost less to process, given an ap-
propriate infrastructure. Some of the costs could potentially be
passed on to the merchants or the consumers, who are notoriously reluc-
tant to directly pay the cost of checks. So far there have been many
trials of using ATM cards as debit cards at the point of sale, but they
have, in general, met with consumer apathy. In some areas, where banks
have aggressively promoted debit, things have gone better. Still, gen-
eral acceptance of debit seems a ways off.

One interesting twist to the debit card story, as mentioned earlier, is
the emergence of third party debit cards. Issuers of these cards have
no real account relationship with the cardholders. Instead, they ob-
tain permission from the cardholders to debit their checking accounts
directly through the Automated Clearing Houses (ACHs), the same way
checks are cleared. (Think of it as direct deposit, in reverse.) Oil
companies first started experimenting with this a couple of years ago,
and it has met with surprising success. Banks dislike this concept,
because it competes directly with their debit cards, but isn't subject
to the same state and federal regulations. ACHs like this, because it
bolsters their business, which otherwise stands to lose a lot by
acceptance of debit cards. Merchants generally like this, especially
the large retailers, because it allows them to get their payment sys-
tems out from under the control of the banks.


THE ATM
--- ---

An ATM is an interesting combination of computer, communication, bank-
ing, and security technology all in one box. A typical machine has a
microprocessor, usually along the lines of an 8086, a communications
module (which may have it's own microprocessor), a security module
(also with a microprocessor), and special-purpose controllers for the
hardware. The user interface is typically a CRT, a telephone-style
keypad, and some soft function keys. Typically there is a lot of
memory, but no disk. The screens and program are usually downloaded
from the host at initialization, and are stored in battery-backed RAM
indefinitely. The machine typically interacts with the host for every
transaction, but it can operate offline if necessary, as dictated by
the downloaded program. The downloaded program is often in an
industry-standard "states and screens" format that was created by
Diebold, a manufacturer of various banking equipment, including ATMs.

Most machines can use a few IBM protocols (bisync, SNA, and an outmoded
but still used "loop" protocol), Burroughs poll/select, and perhaps
some others, depending on which communications module is in place.
This allows the manufacturer to make a standard machine, and plug in
different communications hardware to suit the customer. The IBM bisync
and SNA protocols are most common, with most networks moving toward
SNA.

The security modules do all encryption for the ATM. They are separate
devices that are physically sealed and cannot be opened or tapped with-
out destroying the data within them. In a truly secure application, no
sensitive data entering or leaving the security module is in cleartext.
Arranging this and maintaining it is more complicated than I can go
into here.

Most ATMs contain two bill dispensers, a "divert" bin for bills, a
"capture" bin for cards, a card reader, receipt printer, journal
printer, and envelope receptacle. Some ATMs have more than two bill
dispensers, and can even dispense coins.

When an ATM is dispensing money, it counts the appropriate bills out of
the bill dispensers, and uses a couple of mechanical and optical checks
to make sure it counted correctly. If the checks fail, it shunts the
bills into the divert bin and tries again. Typically, this is because
two bills were stuck together. I've seen ATMs have sensor faults, and
divert the total contents of both bill dispensers the first time a user
asks for a withdrawal. "Gee, all I did was ask for $50, and this ma-
chine made all kinds of funny whirring noises and shut down." Most
banks will put twenty-dollar bills in one of the dispensers and five
dollar bills in the other. Some use tens and fives, or tens and twen-
ties. Depending on the denominations of the bills, the size of the
dispensers, and the policy of the bank, an ATM can hold tens of thou-
sands of dollars.

The journal printer keeps a running log of every use of the machine,
and exactly what the machine is doing, for audit purposes. you can of-
ten hear it printing as soon as you put your card in or after your
transaction is complete.

When you put an envelope into an ATM, the transaction information is
usually printed directly on the envelope, so that verifying the deposit
is easier. Bank policies typically require that any deposit envelope
be opened and verified by two people. In this, you're actually safer
depositing cash at an ATM than giving it to a human teller.

A card will be diverted to the capture bin if it is on the "hot card"
list, if the user doesn't enter a correct PIN, or if the user walks
away and forgets to take the card.

On some machines, the divert bin, capture bin, envelope receptacle, and
bill dispenser bins are all separately locked containers, so that re-
stocking can be done by courier services who simply swap bins and re-
turn the whole thing to a central site.

The entire ATM is typically housed in a hardened steel case with alarm
circuitry built in. These suckers have been known to survive dynamite
explosions. The housing typically has a combination lock on the door,
and no single person knows the entire combination. The machine can
thus be opened for restocking, maintenance, or repair, only if at least
two people are present.

DEBIT CARD PROCESSING
----- ---- ----------

Debit card processing is fairly similar to credit and charge card pro-
cessing, with a few exceptions. First, in the case of ATMs, the ac-
cepter and acquirer are usually the same. For debit card use at the
point of sale, the usual acquirer-accepter relationship holds. In gen-
eral, acquirers may do front-end screening on debit cards, but all ap-
provals are generated by the issuer - the floor limit is zero. This
makes it possible to eliminate a separate settlement process for debit
card transactions, but places additional security and reliability con-
straints on the "authorization". Often a separate settlement is done
anyway.

One problem that has caused difficulties for POS use of debit cards is
the use of PINs. Many merchants and cardholders would rather use sig-
nature for identity verification. But most debit systems grew out of
ATM systems, and require PINs. This is an ironic reversal of the early
ATM card days, when people were trying to avoid requiring signature.
Other than the PIN, the information required for a debit transaction is
the same as that required for a credit transaction.

One last installment on the networks that tie this all together, and
the Credit Card 101 course will be complete. There will be no final
exam - you will be graded entirely on classroom participation. Most of
you are failing miserably...


Joe Ziegler
att!lznv!ziegler

Part 6 - Networks


ACCESS NETWORKS
------ --------

For most credit card applications, the cost of the access network is
the single biggest factor in overall costs, often accounting for over
half of the total. For that reason, there are many different solu-
tions, depending on the provider, the application, and geographical
constraints.

The simplest form of access network uses 800 service, in one of its
many forms. Terminals at merchant locations across the country dial an
800 number that is terminated on a large hunt group of modems, con-
nected directly to the acquirer's front-end processor (FEP). The FEP
is typically a fault-tolerant machine, since an outage here will take
out the entire service. A large acquirer will typically have two or
more centers for terminating the 800 service. This allows better
economy, due to the nature of 800 service tariffs, and allows for di-
saster recovery in case of a failure of one data center. An advantage
of 800 service is that it is quite easy to cover the entire country
with it. It also provides the most effective utilization of your FEP
resources. (A little queuing theory will show you why.) However, 800
service is quite expensive. It always requires 10 (or 11) digits di-
aled, and in areas with pulse dialing it can take almost three seconds
just to dial 1-800. The delay between dialing and connection is longer
for 800 calls than many other calls, because of the way the calls get
routed. All of this adds to the perceived response time at the mer-
chant location, even though the acquirer has no control over it.

Large acquirers prefer to offer some form of local access service. In
this service, terminals at the merchants dial a local telephone number
to gain access to the acquirer. Typically, the local number actually
connects to a packet network, which then connects to the acquirer. If
the packet network is a public network, the terminal must go through a
login sequence to get connected across the packet network. Typically,
local calls are much less expensive than 800 service calls, and local
calls typically connect faster than 800 calls. The cost of those calls
are absorbed by the merchants directly. In those few remaining areas
where local calls are still free from a business line, this works out
well for the merchant. Otherwise, the merchant can end up spending a
lot of money on phone calls. Usually, the acquirer has to offer lower
prices to accepters who use local calls, to help offset this. Even so,
these networks are generally much less expensive for the acquirers.
Such networks are difficult to maintain, due to the distributed nature
of the access network. Since most packet networks are much more likely
to experience failures than the phone network is, the merchant's POS
terminal is usually programmed to dial an 800 number for fallback if
the local number doesn't work. Also, it is generally not cost-effec-
tive to cover every free calling area in the entire country with access
equipment, so some 800 service is required anyway. There is also an
administrative headache associated with keeping track of the different
phone numbers that each merchant across the country needs to dial.
When you have tens of thousands of terminals to support, this can be
formidable.

Acquirers are beginning to experiment with Feature Group B (FGB) ac-
cess. FGB access was the method of access used to get to alternative
long-distance carriers before "equal access" was available. The
tariffs are still on the books, and they are favorable for this appli-
cation. FGB access provides a single number, nationwide, for all mer-
chants to dial in order to gain access to the acquirer. The call has
simpler (hence, presumably, faster) routing than 800 service, and the
call is charged to the acquirer, not the accepter. FGB access does
have to terminate on equipment that is physically located in the Local
Access Toll Area (LATA) where the call originated, so there is the
problem of having distributed equipment, as above. This also implies
that it is not cost-effective to deploy FGB access everywhere, as well.
There are also some technical oddities of FGB, due to its original in-
tent, that have made it difficult to implement so far.

The other big switched access capability that is likely to have an im-
pact in the future is ISDN. So far, this has been inhibited by limited
availability and lack of adequate equipment on the merchant end, but it
could be very beneficial when these problems are solved.

Private-line networks are pretty straightforward applications of
point-to-point and multipoint private lines. Since private lines are
quite expensive, engineering of the networks is challenging. Usually,
sophisticated software is used to determine the optimum placement of
concentrators in order to minimize costs. Since tariffs, real estate
prices, and business needs change frequently, maintaining a stable,
cost-effective network is hard work. A typical asynchronous private
line network will have multiplexers at remote sites, with backbone
links to companion multiplexers at a central site. Synchronous private
line networks may use multiplexers, or remote controllers, or remote
FEPs, depending on the application and the availability of real estate.

INTERCHANGE NETWORKS
----------- --------

Interchange networks physically consist mostly of point-to-point pri-
vate lines. In many of the large interchange networks, there is a cen-
tral "switch" that takes transactions from acquirers (thereby acting as
an issuer), and routes them to issuers (thereby acting as an acquirer).
Often the switch provider will actually be an acquirer or issuer as
well, but this is not always the case. Usually, the provider of the
switch defines standard message formats, protocols, and interchange
rules. These formats and protocols usually comply with national and
international standards, but sometimes do not. Often the switch will
provide translation between different message formats and protocols.

The switch provider is generally very concerned that settlement com-
plete successfully. Failure to settle with one or more large issuers
can leave the switch provider with an overnight deficit of a couple
million dollars. Even though this is a temporary situation, it has
significant financial impact.

In some current networks, authorization and settlement take place on
completely separate facilities, with separate hosts in some cases.
This is mainly due to the history of the industry in this country. Re-
call that authorizations were originally done by voice calls, and
settlement was done by moving paper around. These two processes were
automated at different times, by separate means. Thus VISA has a BASE
1 network for authorization, and a BASE 2 network for settlement.
Likewise, MasterCard has INET and INES, one for authorization and one
for settlement. These functions are becoming less and less separated
as communication and computer facilities evolve, and will probably be
completely integrated over the next five to ten years.

Interchange networks are probably the most volatile part of the ATM
market right now. There is currently a shakeout going on in much of
the market, with larger, more aggressive regionals buying out
standalone networks and smaller regionals. This causes local banks to
change local and national network affiliation from time to time. So a
card may work in a given ATM one day, but fail in that machine the
next, which confuses many consumers. Most large regional and national
networks have operating regulations requiring labeling of ATMs and
cards, so that if you see the same logo on your card and the ATM, you
can be pretty sure it will work.

Some regionals are interconnected, and others are not. The two biggest
nationals, Cirrus and Plus, have operating regulations that effectively
prohibit a member of one network from connecting to the other. But a
regional on Cirrus could be connected to a regional on Plus. In that
case, whether a machine will take your ATM card depends on the routing
algorithm used. In most cases, the acquirer will have a table of issu-
ers that are directly connected, and will send anything else to the re-
gional switch. The regional switch will have a table of each issuer
it is directly connected to, and tables of which cards are acceptable
to other regionals it interchanges with. Anything else goes to the na-
tional switch. The same process happens in reverse from there. Often
the order of search in the routing tables is determined by fee scales,
not geography, so transactions can be routed in completely non-obvious
ways.

So the easiest way to tell if your card will work in a given ATM is to
stick the card in and try. I don't know of any machine that will eat a
card just because it can't route the transaction - it will generally
give some non-specific message about being unable to complete the
transaction and spit the card back out. Of course, if the transaction
is completed from a machine that you're not sure of, you also aren't
sure what the fee is going to be if your bank passes those fees on to
you. Sometimes the fee will be printed on the receipt, but usually it
isn't. If you do the transaction in a foreign country, you may not
know the exchange rate used. (I once couldn't balance my checkbook for
a month until I got a statement with the transaction I did at Banc du
Canada in Montreal.) But if you need the money and are willing to pay
the fee, you have little to lose by trying out just about any ATM.

This completes the course in Credit Card 101. Hope you all found it
enjoyable and informative.