Cybercrime bust highlights PIN terminal insecurity
Clear and present danger
Analysis UK police arrests of a gang reckoned to have tampered with Chip and PIN entry devices to harvest PIN numbers and cardholder details have sparked calls to revamp the security of devices.
Banking industry sources maintain that this type of fraud is rare but recent posts on underground forums suggest that the know-how on how to bypass anti-tampering protection is available for as little as $4,000.
Chip and PIN is based on the EMV standard for secure payments developed by Visa and Mastercard, so the issue has relevance far beyond the UK.
Police from the Dedicated Cheque and Plastic Crime Unit (DCPCU) recovered kit for tampering with PIN pads and hundreds of fake cards in a raid on a counterfeit card factory in Birmingham earlier this week. Two people have subsequently been charged in what is described as an ongoing police investigation.
Concerns about the security of chip and PIN first surfaced when Shell temporarily suspended the authorisation method in May 2006 following the discovery of a systematic fraud that led to losses estimated at £1m. The banking industry maintained the scam did not affect the integrity of the payment method more generally.
However Cambridge University security researchers Saar Drimer, Steven Murdoch and Ross Anderson last year revealed that two popular PEDs, the Ingenico i3300 and Dione Xtreme, fail to adequately protect card details and PINs. Data exchanged between the card and the PED during a transaction is not encrypted. By tapping this communication fraudsters can make counterfeit cards that can be used to make withdrawals from ATMs overseas that rely only on mag stripe readers, exactly the type of scam the Birmingham gang are allegedly involved in. The Cambridge research, which highlighted concerns with the evaluation and certification process, was not accepted by the banking industry.
"We have sent a report on how easy it is to attack PEDs back in November 2007 to all concerned parties: GCHQ, APACS, Visa, MasterCard, Ingenico and Verifone," Drimer told El Reg.
"We were effectively ignored until just before the report was made public by the BBC in February 2008. Then, responses included shifting blame and asserting that what we found is not really a problem. It was also made clear to us that little is about to happen as a result of our findings; PEDs will not be recalled or de-certified from use. One frequent response to our findings by APACS is that PED tampering is only possible under lab conditions."
An APACS spokeswoman said that Chip and PIN had made a big dent in fraud levels. "For fraudsters it's not a question of what's possible but what's cost-effective. Card details can be captured in a variety of ways. This is just a new technique, focused on the PIN pad. Details captured are written onto mag stripe cards and used fraudulently abroad so the attack ultimately relies on flaws with older technology.
"There's no evidence to suggest the chip was actually cracked and used in a meaningful way. If it was then chip security would be upgraded," she said.
Sorry to blow our horm, but the problem *is* solved..
Could I humbly suggest you look at www.axsionics.ch, a Swiss startup? I'm working on the docs so if you want decent details mention it (once I have this I will send El Reg a token to play with, give me a couple of weeks).
In short, it's a trusted display (graphical OLED), combining more or less all of your above comments. To address question one upfront: no, the use of biometrics does not mean that a "disconnected" finger is of use (or its friendlier equivalent, the copied fingerprint a la Chaos Computer Club). The reader is quite good at rejecting fakes, and you have to "name" your fingers - only you know which finger "g" is, for instance, if you used the word "frog" to name them.
A message for the token is AES128 symmetric dual cert encrypted, so it has to (a) come from a defined source (the token accepts 128 different origin certs) and (b) has to be encoded for that token or it won't be able to decode it. It picks that encrypted message up via a screen animation, and after taking a valid fingerprint it will show it, together with a password if an answer is required. So, "To: BT, A/C Household, Val. GBP 125,23, PIN ABC45F" is quite possible (or "Please call us on +44 1234 4568") - and that PIN is also meaningless to anyone but you and the sending server because it's a One Time Password, generated on the card.
This means that a Man in The Middle Attack won't work, and -VERY- important, that you do NOT need a secure terminal. There is no reason why you can't use one of the card's channels as a payment method, which ends the need for secure terminals altogether. Instead, you just pop up an iframe in the POS display (or an external one), supply the required finger sweep, read the message and enter the PIN (numeric or alphanumeric) if required. Ditto at home - regardless whether your system is virus infested or not.
To give you an idea where I'm coming from, I was consulting private Swiss banks on next generation eBanking, and the basic premise there too is that we have to assume the client PC *is* infected. Yet, you still need to supply secure eBanking.
Expect to hear more from us soon :-).
Chip and Sig
With a chip and signiture card does the liability rest with the retailer/ bank rather than the customer as with chip and pin?
The PED shouldn't be trusted at all
I think one big weakness here is that the PED is inside the security boundary and therefore has to be trusted.
Regardless of whether the card details are encrypted between the PED and card it is implied that they exist, in plaintext form, in the PED. So, the PED can be attacked to get the card details in this case it's easy - just tap the wires between PED and card. However, If the info going between the PED and card was encrypted they could attack the PED itself - still possible but probably more difficult and expensive.
End-to-end encryption between the card and bank would take the PED outside the security boundary (data would be encrypted from the card right to the bank). OK, the PIN and payment amount would still need to be sent to the card from the PED but that has limited use without the card details. Better still, as someone said above, would be to put the keypad & display on the card so the PED just proxies data between card and bank.
I don't know for sure but I bet the idiots that developed this sh** system traded off security against additional card complexity (and therefore cost). By putting more "intelligence" (& therefore trust) in the PED the card becomes simpler and cheaper.