Re: [TLS] Last Call: draft-ietf-tls-rfc4366-bis (Transport Layer

> -----Original Message-----
> From: Simon Josefsson [mailto:simon@josefsson.org] 
> Sent: Tuesday, September 29, 2009 10:20 PM
> To: Joseph Salowey (jsalowey)
> Cc: Michael D'Errico; martin.rex@sap.com; ietf@ietf.org; tls@ietf.org
> Subject: Re: [TLS] Last Call: draft-ietf-tls-rfc4366-bis 
> (Transport Layer
> 
> "Joseph Salowey (jsalowey)" <jsalowey@cisco.com> writes:
> 
> > It seems that this is really up to the application.  Both 
> server names 
> > are authenticated under the same session.  It seems an application
> > server may require them to be the same or allow them to be 
> different.   
> 
> I would agree if the draft wouldn't prevent clients from 
> requesting a different server name at the application layer:
> 
>    negotiated in the application protocol. If the server_name is
>    established in the TLS session handshake, the client SHOULD NOT
>    attempt to request a different server name at the 
> application layer.
> 
> At least that is how I read it.
> 
[Joe] Good point, however I still think it is application protocol that
needs to enforce the matching if it cares.  Perhaps we can add some text
that states

"Since it is possible for a client to present a different server_name in
the application protocol, application server implementators should take
this into account and take appropriate action to avoid introducing
security vulnerabilities if the names do not match.  "

Peter's text also included the possibility of server name change during
a renegotiation handshake, do you think we should include this
consideration here as well?   

> /Simon
> 
> >
> >> -----Original Message-----
> >> From: tls-bounces@ietf.org [mailto:tls-bounces@ietf.org] 
> On Behalf Of 
> >> Michael D'Errico
> >> Sent: Tuesday, September 29, 2009 4:48 PM
> >> To: martin.rex@sap.com
> >> Cc: simon@josefsson.org; ietf@ietf.org; tls@ietf.org
> >> Subject: Re: [TLS] Last Call: draft-ietf-tls-rfc4366-bis 
> (Transport 
> >> Layer
> >> 
> >> >>> I do not see why you consider this a vulnerability in the
> >> _server_!
> >> >>
> >> >> Because a malicious client could theoretically 
> establish a secure 
> >> >> connection using one server domain and then ask for 
> pages from a 
> >> >> different domain.  If the server does not check for 
> this, it could 
> >> >> potentially leak sensitive information.
> >> > 
> >> > You're barking up the wrong tree.  If the client did not
> >> use TLS, the
> >> > server wouldn't even know that.
> >> 
> >> You must be talking about a particular server 
> implementation that has 
> >> this shortcomings.  There is nothing inherent in TLS that 
> prevents a 
> >> server from knowing when it is used.  Your library and/ or use of 
> >> that library is the problem.
> >> 
> >> > It is inappropriate to assume that virtual hosting provides
> >> seperation
> >> > of content and draw a conclusion that, when accesses via
> >> HTTPS, will
> >> > provide a secure seperation of content instead.
> >> 
> >> I'm not assuming anything; I have written a TLS library 
> and an HTTP 
> >> server that provides the separation of content that you deny is 
> >> possible.
> >> 
> >> > If the lack of such a server-side check is a problem for
> >> your server,
> >> > then your server problably has a severe design flaw in 
> its session 
> >> > management.
> >> 
> >> I never said my server suffered from this problem....
> >> 
> >> >> And I'm curious: why do you call matching the commonName weak?
> >> > 
> >> > Because in the vast majority of situatins it is the last 
> step in a 
> >> > long row of flawed assumptions.
> >> 
> >> OK, so you are complaining about the entirety of e-commerce on the 
> >> web.  Do you have any proposed solutions to these problems?
> >> 
> >> Mike
> >> 
> >> 
> >> > Security is only as strong as its weakest link.  The 
> authentication 
> >> > process based on a DNSName involves a number of very weak
> >> authentications.
> >> > 
> >> > DNS domain names are not very genuine, and it is very 
> non-obvious 
> >> > which domain names are used by the business or peer someone
> >> is looking
> >> > for and which are used by others (different businesses with
> >> the same
> >> > name, cybersquatters or attackers).  Most HTTPS-URLs 
> opened by Web 
> >> > Browsers are served through plaintext HTTP pages.
> >> > 
> >> > Then most Browser PKIs come with a hundred or more trusted CAs 
> >> > preconfigured, and browsers trust them equally.  Whether or
> >> how secure
> >> > the authentication is that the CA performs before issuing a 
> >> > certificate is another flawed assumption that weakens the
> >> > rfc-2818 server endpoint authentication.
> >> > 
> >> > A final flaw that is still present in most browsers is 
> the lack of 
> >> > memory.  Not memorizing the certificate that a server
> >> presented on the
> >> > last contact perpetuates the weakness of the original
> >> authentication.
> >> > 
> >> > Personally, I think that deriving a server endpoint
> >> identifier from a
> >> > network address is the most flawed assumption of all.
> >> > 
> >> > That is like asserting that if someone opens on a random
> >> door on which
> >> > you knock, and shows you an ID card with the correct street
> >> address --
> >> > then he must be a GOOD guy.
> >> > 
> >> > 
> >> > -Martin
> >> _______________________________________________
> >> TLS mailing list
> >> TLS@ietf.org
> >> https://www.ietf.org/mailman/listinfo/tls
> >> 
> 