RFID and the Middleman …

                   RFID and the Middleman

                                   Ross Anderson

                               Cambridge University
                          http://www.ross-anderson.com



      Abstract. Existing bank-card payment systems, such as EMV, have
      two serious vulnerabilities: the user does not have a trustworthy interface,
      and the protocols are vulnerable in a number of ways to man-in-the-
      middle attacks. Moving to RFID payments may, on the one hand, let
      bank customers use their mobile phones to make payments, which will
      go a fair way towards fixing the interface problem; on the other hand,
      protocol vulnerabilities may become worse. By 2011 the NFC vendors
      hope there will be 500,000,000 NFC-enabled mobile phones in the world.
      If these devices can act as cards or terminals, can be programmed by
      their users, and can communicate with each other, then they will provide
      a platform for deploying all manner of protocol attacks. Designing the
      security protocols to mitigate such attacks may be difficult. First, it will
      include most of the hot topics of IT policy over the last ten years (from
      key escrow through DRM to platform trust and accessory control) as
      subproblems. Second, the incentives may lead the many players to try
      to dump the liability on each other, leading to overall system security
      that is equivalent to the weakest link rather than to sum-of-efforts and
      is thus suboptimal.


1   Introduction
Card payment systems have come under repeated attack from forgers and other
fraudsters. The mechanisms used in magnetic-strip cards, and attacks on the
ATMs that rely on them, are documented in [2], while the back-end backing
systems are described in [3]. Banks in Europe are now moving to smartcards
following the EMV standard, branded in the UK and Ireland as `Chip and PIN'.
In the most commonly-deployed variant of this standard, bank cards contain a
smartcard chip that will verify a customer PIN against a locally-stored value,
and also one or more cryptogram generation keys. These are used to compute
a message authentication code (MAC) on transaction data; the MAC is ver-
ified by the card-issuing bank. As many countries, including the USA, have
not adopted EMV, the cards also have a magnetic strip for fall-back operation.
The introduction of EMV has altered the fraud landscape, with a reduction in
cardholer-present forged-card losses, coupled with increases in cardholder-not-
present fraud and mag-stripe-fallback fraud.
    Now that European cardholders are used to entering cards and PINs for
all bank card transactions, rather than just ATM transactions, it has become
much easier for attackers using false terminals to harvest card and PIN data.
Previously, card forgery attacks typically involved skimmers attached to the front
of ATMs; now we see a spate of offences involving wiretapping of the links from
EMV terminals to branch server equipment. With frauds reported in France
and Italy, both card and PIN data were collected using surreptitiously-installed
wiretap devices, which send the data to waiting criminals by wireless [11]. In
a series of recent UK cases, card data have also been harvested using an in-
store wiretap, but the PINs collected by observation. Many offences have been
reported in garages, staffed by Sri Lankan Tamils, leading to ATM withdrawals
in India, Malaysia or Thailand; it's reported that the Tamil Tigers use these
forgeries for fundraising ­ presumably intimidating their UK expatriates into
collaborating [8, 12].
    In addition to such fairly straightforward frauds, it turns out that the EMV
protocols and their implementations are vulnerable to various middleperson at-
tacks [1]. In addition to capturing card details and PINs for use in magnetic-strip
terminals, villains can do various kinds of man-in-the-middle attack, and the
transactions provided by cryptographic hardware security modules to support
EMV have flaws that enable bank insiders to extract PINs. The growing num-
ber of bank customers who complain that stolen chip-and-PIN cards were used
without the PIN possibly having been compromised suggests that at least some
banks have corrupt staff who sell PINs to criminals; this was a known modus
operandi with magnetic-strip cards in the 1990s [2]. Meanwhile, Murdoch and
Drimer have shown that real-time man-in-the-middle attacks are feasible by im-
plementing them [9]; and it's also been noted that a bank customer could use a
middleperson device to fix the protocol by providing a trustworthy user interface
­ she could observe the actual data traffic between the card and the terminal,
rather than the possibly false data displayed by the terminal [5].


2   Implications for RFID

The bank card industry in the USA and Japan is introducing RFID credit cards,
described by Heydt-Benjamin et al. in [7]. Although the specifications are confi-
dential, protocol dumps suggest that the mechanisms are very similar to EMV,
although with the crypto (the MAC) missing. The authors of that paper showed
that cloning and relay attacks work in principle by scanning a credit card in its
sealed delivery envelope and making a purchase with it.
    Meanwhile, the mobile-phone industry is introducing Near-Field Commu-
nications (NFC), whereby a phone acquires the capability to act as either an
RFID device or an RFID terminal under program control. The NFC protocols
themselves provide an abstraction layer between the four different RF interfaces
already deployed, thus providing a clean interface for the software developer.
The NFC Forum envisages about half a billion NFC-enabled handsets in use by
2011 [10].
    This technology holds out the prospect of solving the problem of a trustwor-
thy user interface. The plan is that instead of being a relatively dumb device,
your credit card will be an application on your mobile phone. You bring your

                                        2
mobile into close proximity with the merchant terminal, an application displays
the sale amount, you authorise this, and the transaction goes through. It may
also bring further security benefits, such as a single point of revocation [6].
    However, without protocols giving robust protection against man-in-the-
middle attacks, the trust that might be placed in this interface will be largely
illusory. Worse, NFC is likely to make middleperson attacks much easier. At
present, such an attack requires the construction of custom hardware; in fu-
ture, an attack could be carried out by software installed on commodity mobile
phones. One phone could act as a rogue merchant terminal to the cardholder,
and communicate with another that acts as a card to a merchant elsewhere.
    Transaction forwarding does have `honest' uses. (Once when I stayed at a
hotel in Malawi I found that the bill appeared on my credit card via a South
                ^
African merchant in Rands ­ an obvious attempt to collect a slightly harder
currency, which cost me no more.) However, most forwarding is likely to be
objectionable. Someone thinking she's buying an hour on a parking meter in
Baltimore for $2 might find out, when the credit card bill arrives, that the bank
thinks she paid $2000 for casino chips in Macau.
    Limiting transactions without cryptography to low values ­ as some suggest
­ won't solve all the problems. For example, a crook might collect large numbers
of small payments from passers-by; this might be done by malware installed on
the phone of an unwitting victim, which would steal a few dollars from everyone
he passes. Theft might not be the only objective; RFID will be used for sports
ticketing, so known hooligans who are banned from buying tickets will have every
incentive to spoof the system.
    If cryptography is introduced, it's not obvious how. The banks will want
end-to-end protection, but merchants will want access to transaction data; and
the police will want access to records for intelligence/evidence, raising all the
old issues about escrow. It's unclear that much more can be done than is al-
ready being done with EMV ­ message authentication codes that might just be
upgraded to digital signatures eventually.
    It's also not obvious who will design any new protocol suites to support NFC
payments. The NFC Forum limits itself to interoperability, and the protection
issue are not just about bank payments but transport tickets, sports tickets,
supermarket coupons and much else. How many protocol suites will there be?
What sort of limits are desirable on relaying / cloning bits? Who is going to
specify the protocols? (VISA? Microsoft?) Who will regulate them? (The FTC
and the EU's DG Comp? The Federal Reserve and the European Central Bank?)


3   Conclusions

The introduction of RFID payments based on programmable devices such as
NFC mobile phones may fix one of the problems underlying bank card fraud ­
the lack of a trustworthy user interface. However, this may well be at the cost
of seriously exacerbating the other main problem ­ the vulnerability of current
payment protocols to various man-in-the-middle attacks.

                                       3
    What's more, the responsibility for security is very widely dispersed, and
defenders may hope they can rely on each others' mechanisms. Security eco-
nomics teaches that we get a much more appropriate level of protection where
this results from the sum of defenders' efforts than in the cases where it results
from the weakest link ­ from the least awful of the disparate efforts of a number
of possibly uncoordinated defenders [13]. Unfortunately, the security of RFID
transactions looks set to become a matter of the weakest link. So a large-scale
infrastructure may be deployed that's not only systemically vulnerable to man-
in-the-middle attacks, but that actually provides the platform required for these
attacks to be carried out easily. I'd suggest that application providers think hard
about these issues now; it will be much more expensive later.


References
 1. B Adida, M Bond, J Clulow, A Lin, S Murdoch, RJ Anderson and R Rivest,"Phish
    and Chips", at Security Protocols Workshop, Mar 2006; paper at http://www.
    ross-anderson.com
 2. RJ Anderson, "Why Cryptosystems Fail", in Communications of the ACM v 37
    no 11 (Nov 94) pp 32­40
 3. RJ Anderson, `Security Engineering ­ A Guide to Building Dependable Distributed
    Systems' Wiley (March 2001), ISBN 0-471-38922-6
 4. RJ Anderson, "Why Information Security is Hard ­ An Economic Perspective", in
    Proceedings of the Seventeenth Computer Security Applications Conference, IEEE
    Computer Society Press (2001), pp 358­365; available at http://www.cl.cam.ac.
    uk/ftp/users/rja14/econ.pdf
 5. RJ Anderson, M Bond, "The Man-in-the-Middle Defence", at Security Protocols
    Workshop, Mar 2006; paper at http://www.ross-anderson.com
 6. M Baard, "Will new RFID technology help or hinder security?", 27 Apr
    2005, at http://searchsecurity.techtarget.com/originalContent/0,289142,
    sid14 gci1083417,00.html
 7. TS Heydt-Benjamin, DV Bailey, K Fu, A Juels, T OHare, "Vulnerabilities in First-
    Generation RFID-enabled Credit Cards", at Eleventh International Conference on
    Financial Cryptography and Data Security Scarborough, Tobago, February 2007
 8. Walter Jayawardhana, "Tamil Tigers suspected of scamming millions in Britain",
    at http://lankapage.wordpress.com/2007/01/17/
 9. SJ Murdoch, "Chip & PIN relay attacks", Feb 6th 2007, at http://www.
    lightbluetouchpaper.org/
10. `Near Field Comunication and the NFC Forum: The Keys to Truly Interoperable
    Communications', 2006, at www.nfc-forum.org
11. "Clonavano carte con il bluetooth Scoperta nuova truffa telematica", in la Repub-
    blica Sep 4th 2006, at http://www.repubblica.it/2006/09/sezioni/cronaca/
    truffa-blue/truffa-blue/truffa-blue.html
12. K Shoesmith, "Garage Scam funded Terror Group", Hull Daily Mail, Jan 16th
    2007, p 1, archived online at http://www.srilanka-botschaft.de/NEWSupdates
    neu/Press Releases/Press Pol Government Statement 070119bE.htm
13. H Varian, "System Reliability and Free Riding", available at http:
    //www.sims.berkeley.edu/resources/affiliates/workshops/econsecurity/
    econws/49.pdf



                                         4