Re:Prof. Ross Anderson's response to a takedown request abou...
by huhtenberg on Sunday December 26, 2010 @04:14PM
(#2041184)
Actually, no, that's not how it worked when I last looked at it. At the high-level it was:
- bank asks terminal "Here's a random token, have the chip sign it with its private key"
- terminal asks the customer for the PIN
- terminal feeds PIN into the chip, and this enables signing function
- terminal feeds bank's token into the chip, gets the signature back and forwards it to the bank
What you described looks like something designed by a layman with very basic understanding of the cryptography. I will not be shocked if this was in fact deployed, but I still find it very unlikely.
Re:Prof. Ross Anderson's response to a takedown request abou...
by Someone on Sunday December 26, 2010 @07:22PM
(#2041446)
I reread http://www.cl.cam.ac.uk/~sjm217/papers/oakland10chipbroken.p..., and things are indeed similar to what you describe. If I understand things correctly, the difference is that the "and this enables signing function" part isn't there. It is enabled by default, and gets disabled only when PIN check fails. That is 'necessary' because the terminal can skip this step if it wants to do a transaction without PIN check.
A man in the middle can hide the "check this PIN" request that the terminal sends from the chip and send a "PIN is OK" reply to the terminal. That way, the terminal thinks PIN check succeeded, and the chip thinks it is doing a payment without PIN check.
That may or may not be case (there is no way for customers to check that), but the problem is that this encrypted communication is between some third party's terminal and my bank, not between my chip and my bank. IIRC correctly, the protocol in the end boils down to:
- bank asks terminal 'Can you verify that the customer is who he claims to be?'
- terminal asks user for PIN
- terminal asks chip: is this PIN correct?
- chip replies: yes.
So, I have to trust that the terminal will not e.g. put my PIN on Twitter.
Worse, that terminal-chip communication is not encrypted. Hence, it is vulnerable to a man-in-the-middle attack. That is what 'chip and PIN is broken' demonstrated.