Datasets:
Devilishcode
/
hacking

Dataset card Data Studio Files
xet
 Community
text
stringlengths
0
1.99k
So, if you don't believe hackers will invest a lot of time and attention to
fraudulently purchase a bag of potato chips, you may deem it unnecessary to
cryptographically authenticate the smart card. We have seen other instances
where industry has a blind spot about the capabilities and interests of
hackers (especially university-affiliated hackers).
"User experience" considerations may also play a role. Like most smart
cards, the MCard featured an 8-bit microprocessor running at 3.57 MHz,
communicating half-duplex through one contact at about 1 msec per byte.
Slow processing and communication produces authentication round-trip times
longer than a second. If too many seconds elapse between card insertion and
the vending machine signaling readiness, or if even a second goes by
between selection of a product and the delivery of that product, the
customer is going to be frustrated.
If round trips were fast, the protocol might look something like
1 Mutually authenticate
2 Check purse value
3 Customer makes selection
4 Update the purse
5 Deliver the selection
6 Eject card
The critical section lies between steps 3 and 5, updating the purse between
the time the customer presses a button to make a selection and the start of
product delivery. This is when the customer is most anxious about their
purchase: if that little spiral holding the KitKat Bar doesn't begin to
spin immediately, the user experience breaks down and havoc can ensue.
But even this protocol fails in the face of a determined hacker. Taking the
inverse card reader to the next level, CITI developed a
monkey-in-the-middle apparatus that filters messages from the vending
machine, directing them either to a legitimate card or to a rogue
application. This can defeat any protocol that doesn't cryptographically
sign messages (at the very least).
--[ 5 - Responsible disclosure
We had good contacts on the card manufacturer's research team, and a
productive phone call took place right away. Naturally, we shared our
discovery with university officials as well; actually, that was the first
phone call. The response was lukewarm thanks. But one official with whom we
met surprised us by pulling a Visitor MCard out of a desk drawer and
informed us that every time they inserted this card into a vending machine,
five quarters dropped into the coin return!
We borrowed the card and examined it in our lab. The card has a subtle
flaw: the key file for authenticating the vending machine is unreadable, so
when the vending machine wants to reserve funds in the log, authentication
fails and the card reports an error.
What happens next requires us to speculate. The smart card module is
retrofitted to an OG vending machine stocked with items that cost less than
$1.25. The module verifies that there is $1.25 in the purse, tells the
vending machine to credit the customer with $1.25, and reserves $1.25 in
the log. We don't know in which order the module takes these steps, but the
order is important!
When the authentication step fails, the smart card module appears to throw
up its hands, eject the card, and go offline. The vending machine is now
holding the bag for $1.25, and the smart card module is offline. The
vending machine can't credit the refund to the card, so it does what an OG
vending machine knows how to do: cash refund. Five quarters drop into the
coin return.
We were excited by this discovery! Together with the earlier vulnerability,
we could strip a vending machine of all of its product and all of its cash
(if we dared). We were especially amused by the sound of a vending machine
eventually dry-heaving quarters it didn't have: <cough cough cough cough
cough>. I wish we had recorded that.
Nonetheless, our sense at CITI was that these vulnerabilities were too
shallow to merit a refereed publication, and to an extent was a distraction
from our sponsored research, so we talked about it at a few conference rump
sessions and that was the end of it.
But we also saw another possibility: funding ongoing smart card research.
No, not like that! Having proved our skills to the smart card vendor, we
were encouraged to submit a modest proposal (to tunnel IP and ICMP through
ISO 7816), but it turned out that vendor higher-ups had bigger ideas and
approached CITI with a much broader scope and an order of magnitude larger
price tag in mind. CITI initiated the Program in Smartcard Technology in
the fall of 1998 and celebrated: for the first time under my direction,
CITI was on solid financial ground.
I know what you're thinking, but I don't think Schlumberger bought CITI's
silence. We sincerely believed that our discoveries were below the LPU
("least publishable unit") threshold. And there was genuine excitement on
both sides about the potential for collaboration. But I'm sure they were
grateful that they weren't forced to reflash their worldwide installed base
of vending machine smart card modules, and perhaps the grant was one way
for them to show their gratitude.
Whatever, we spent it all ... on student stipends, tuition, staff salaries,
conference organizing, publishing, traveling, etc.
One of our earliest objectives was a smart card-based Kerberos sign-on for
our UNIX and Windows computers and we got pretty far with it: we found a