Re: Privacy and its costs (was: Re: Mandatory encryption)

On 25 July 2012 18:13, Greg Wilkins <> wrote:
> Also consider a Server using some HTTP/2.0 push feature to push out
> stock market prices as they change and users can have a custom
> portfolio of stocks that they can watch.  It can be very valuable
> information to know what stocks a top trader has in their portfolio,
> so if you sniff packets on their network, it does not matter that the
> contents are encrypted, because over a period you can correlate the
> time that they receive encrypted packets with known fluctuations of
> stock prices and thus work out the contents of their portfolio.

I don't want to disagree with your general argument, but what you are
describing is a traffic analysis attack on the application that works
even when TLS is employed.  That's a good argument against the idea
that TLS solves all security problems, but not one that supports the
case for cleartext HTTP.

We've seen a number of compelling cases for intermediation based on
the value that intermediaries can provide.  But we've also seen that
with the right solution, the decision to enable intermediation can be
made independently of the choice to require or disallow TLS.  But, so
far, I've not seen a workable solution for this problem - only
expressions that it should be possible (i.e.,
draft-rpeon-httpbis-exproxy).  Until such a solution manifests with
reasonable backing, I'd consider these to form a fairly strong
argument in favour of allowing use of HTTP/2 without TLS.

That doesn't mean that we can't also motivate the use of cleartext
HTTP/2 by pointing out disadvantages of secured HTTP/2 or advantages
of cleartext HTTP/2.

The fact that you can't hide who you are talking to is perhaps one
such disadvantage.  Intermediation actually helps those cases by
creating K-anonymity sets... and in the end, that's half of what TOR


Received on Thursday, 26 July 2012 16:22:08 UTC