7 May 2014
This blog is for my non-techie, non payment friends.. helping to make sense of all these acronyms.. experts may want to pass on this one.
The GSMA/NFC community is quite stirred up at the moment. This is quite understandable… after all they spent 8 years perfecting their vision of NFC only to have it thrown under the bus by Apple and Google. I’m not knowledgeable enough to go into the depths of the protocol, or EMVco 4.3 Book 3. I’m giving the quasi technical business explanation of what is going on. There is room for disagreement here, as there is substantial interpretation, as well as understanding of what is REALLY happening vs the specifications. Remember this is not my day job… so your comments/corrections are welcome. By far the most useful reference/summary page I have found online is located here http://www.nfc.cc/2012/04/02/android-app-reads-paypass-and-paywave-creditcards/
It’s easiest for me to explain all of this in the context of an example. Credit cards are the easiest example as they are in the market today, with a few different implementations of contactless and touch the areas above.
EMVco has a contactless specification which I challenge any non-techie to read. For this short blog, the key point I wanted to make is that the Credit card number (PAN) is given to the POS unencrypted, in the clear. That’s right… don’t believe me? See:
- Analysis from viaForensics.
- Forbes Article on cloning of RFID creditcards
- And a detailed Tutorial from this Blog
- Great Overview from CryptoMathic
Your next question is probably “Where is the security?” the answer is that that along with the card information, the device sends a cryptogram that is uniquely signed. In other words there is a digital payload that rides along with this credit card primary account number (PAN). This digital payload uniquely identifies the device that EMULATED THE CARD. Think about is as someone validating your SIGNATURE on the document with your social security number on it… Your number is there.. but they make sure it is you by validating the signature.
So why is the SIMAlliance extolling the virtues of a Trusted Execution Environment (TEE) and SIM/UICC? After all we seem to live without this capability quite well in the PC world. Mobile operators want the ability to SIGN and AUTHORIZE more than access to mobile towers. That SIM card in your GSM phone signs and authorizes access to the mobile network, much as MNOs envisioned doing for payments. That is how the GSMA’s version of NFC evolved.. “hey we do this for network access.. lets do it for payments”. To be clear there is nothing technically wrong with the GSMA NFC approach.. it is beautiful… but there are substantial business model issues (see Payments part of the OS).
Apple and Google are both moving aggressively to act as Commerce Orchestrators as handsets become commodities and data moves to cloud, enabling the mobile phone to be the key services platform at the confluence of the virtual and physical world is critical. It is not about payment. Authentication is core to this orchestration role.. authentication is not something that can be given away to MNOs or to Banks.
It makes most sense to jump to TOKENS now. You can imagine that Banks don’t exactly like having their card numbers sent in the clear. In fairness they were involved in the specification, but the EMVCo contactless model is essentially a card number plus authentication. There is more than one way to achieve this, and improve on it by hiding the PAN… this is what tokens are (a few examples described in Money 2020: Tokens and Networks, Apple’s Plans and Google/TXVIA).
Tokens are not new (see Tokens… 10 Approaches). However Tokens are now an official EMVCo specification as of March 2014, with the major issue of Token Assurance outstanding. In this token model, the issuer chooses at Token Service Provider (or does it themselves) and creates a number to replace the PAN. This takes your PAN out of the open… and makes it useless. To be used the Token must be presented by the right party, with the right assurance information. All of this aligns VERY WELL to how banks and networks work today, which is why it is so popular (see blog on HCE). In the GSMA NFC model, the a cryptogram goes along with a PAN in the clear with the PAN stored in the phone in a secure element. In the token/HCE model a Token representing the card is stored in a less secure space, and presented with device and network information for translation by the TSP to the actual PAN. There are substantial Business Implications of Payment Tokens (blog) which I won’t go through again here, but clearly it cuts the mobile operator out of the “signing” role and they become dumb pipes.
My Gemalto friends will howl at how unsecure this is, or how it won’t work if the device has no network access. They are wrong. It is working today, and is secure enough. There is no connectivity requirement, that software token in the phone can change every 10 seconds, 10 minutes or 10 days. The TSP and Issuer can decide whether or not to accept an “old” token based upon the transaction. In other words the intelligence sits IN THE NETWORK.. NOT IN THE PHONE. This is why V/MA/AMEX love it so much. It cements their position (See Perfect Authentication… A Nightmare for Banks?)
Host Card Emulation
This is an Android construct (see Software Secure Element – HCE Breaks the MNO NFC Lock) that allows any application to access the NFC Radio. Without Tokens, HCE would be useless for payments, as payment information can’t be securely maintained without an SE. Think of HCE as dependent on tokens, now a card emulation application can be certified to run outside the secure element. I don’t like to put Apple in the HCE boat, as they have a proprietary secure architecture using tokens. This is a uniquely apple construct where the networks seem to have certified Apple’s card emulation application(s) as well. It is important to note that they use none of the GSMA’s architecture (to my knowledge) and have embedded the TEE in the apple processor (see Apple Insiders note on Secure Enclave and Authentication in Value Nets).
Is it needed? Certainly it is needed for at least 2 functions: Mobile network access (SIM/UICC) and Biometrics. Fingers and Eyes are very hard to reissue.. so the actual information must be highly protected. Apple is handling biometrics in the A7 Secure Enclave (oddly enough has the same “SE” acronym) and Google is a tad bit behind but handling in ARM’s trustzone. Trust zone is largely a hardware construct, and much is made of Gemalto’s marketing announcement here. My view is that there are many more than on software solution for ARM.. and ARM is much more tied to Google and OEMs than Gemalto.
The “big news” here is that both Google and Apple are EMBEDDING SEs in their hardware architecture. Embedded SEs are a threat to Mobile Operators and their preferred Single Wire Protocol architecture. As you can imagine, an embedded SE has all the capabilities of the SE within that micro-SIM card.. and sets up the prospect for a Virtualized SIM (no more of those GSM cards popping into your phone). If the SIM can be virtualized you can switch your network provider anytime you want.. or have them bid for your phone call ( see Carriers as dumb pipes? , Who do you Trust?, Also see Apples patents on Virtualized SIM). To be clear, I believe MNOs can take a leadership position in Emerging markets and payments, but for POS Payments in OECD 20 markets it makes most sense for them to focus on the $5B KYC/Authentication/Fraud opportunity (NOT payments).
OK… now you can shoot me… Open to feedback.