W3C home > Mailing lists > Public > ietf-http-wg@w3.org > July to September 2012

Re: Semantics of HTTPS

From: Stephen Farrell <stephen.farrell@cs.tcd.ie>
Date: Tue, 07 Aug 2012 10:21:25 +0100
Message-ID: <5020DE15.3070008@cs.tcd.ie>
To: Willy Tarreau <w@1wt.eu>
CC: Yoav Nir <ynir@checkpoint.com>, "Adrien W. de Croy" <adrien@qbik.com>, Mark Nottingham <mnot@mnot.net>, "ietf-http-wg@w3.org Group" <ietf-http-wg@w3.org>

So I think Yoav is basically right, his TLS proposal
for a MITM that was rejected by the tls WG is more or
less the same as the other ideas being sketched here
even if those are done as HTTP mechanisms. And all
of them represent a significant change to the semantics
of https.

On 08/07/2012 08:37 AM, Willy Tarreau wrote:
> Hi Yoav,
> 
> On Tue, Aug 07, 2012 at 10:30:04AM +0300, Yoav Nir wrote:
>> Best we can do is something along the lines of "Your traffic to
>> "www.mybank.com" is being decrypted an inspected by "sslproxy.example.com".
>> Is this OK?"

I agree with Yoav that this is no different from the
current impact on user interfaces caused by MITM'ing TLS.

>> Do you think this allows a user to make an informed decision? Usability
>> studies suggest that the user will click on whatever button makes him get to
>> www.mybank.com, without thinking about the implications. This may or may not
>> be the correct decision, but changing browser UI to fit security geeks does
>> not necessarily make sense.
> 
> Precisely having the proxy take care of TLS could improve the situation a lot.

Personally, I have to say I doubt that.

> The proxy would simply refuse to connect to improperly configured sites and
> the client would not be able to do anything about this, except try to pass
> using the CONNECT method if it's whitelisted in the proxy.
> 
> The client would only have the proxy's cert and the browser could be
> configured to always refuse to connect to an invalid proxy instead of
> prompting the user.
> 
> So in the end, we would significantly improve end-user security by enforcing
> security where it matters and where people can make informed decisions, and
> leaving the user with less hesitation caused by warnings he doesn't understand.

There are so many ways in which I disagree with the above
statements/speculations that a blow-by-blow response is
probably not going to get us anywhere;-)

But I do think we've answered the question posed in the
subject line in the affirmative, so we could end this
thread with:

  Yes, any MITM significantly changes the semantics of
  https.

Cheers,
S.


> 
> Regards,
> Willy
> 
> 
> 
Received on Tuesday, 7 August 2012 09:22:03 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 7 August 2012 09:22:11 GMT