Re: [Taler] denomination manipulation

[Jeff Burdges]

On Sat, 2015-11-28 at 12:43 +0100, Christian Grothoff wrote:
> 1) RESTful API with 'GET" over HTTP implies that HTTP cache control
>    is always a possibility; for /keys, this was made an explicit
>    requirement a month ago:
>    https://gnunet.org/bugs/view.php?id=4036

Tor Browser does not cache between sessions.  And FireFox's 50mb HTTP
cache can be cleared in minutes of really heavy browsing anyways. 

Any caching for security purposes would need to be done separately.


On Sat, 2015-11-28 at 12:37 +0100, Christian Grothoff wrote:
> No, because there is no *correlation*. You get a request to /keys at
> 5pm. What do you correlate that with? Well, you also got a request to
> say 'index.html at 4:59:59. Great. So those two likely came from the
> same user.

That's a correlation.  Our state level attacker who hacked both the
mint and bank can use the fact that their target customer identified
themselves with their bank right before their bank sent them to the
mint.  That's plenty.


I'm fairly happy with caching /keys as a protection, but HTTP caching
won't cut it.

Jeff

signature.asc
Description: This is a digitally signed message part