Re: channel bindings and address types
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: channel bindings and address types



On Sat, Jun 27, 2009 at 06:39:21PM +0200, Simon Josefsson wrote:
> Shawn M Emery <Shawn.Emery at Sun.COM> writes:
> 
> > RFC 5554 states that network address as channel bindings is not very
> > secure, which the chair would imply as not being very useful.
> 
> Where does it say that?  This part
> 
>    Language bindings that use OCTET STRING (or equivalent) for channel
>    bindings will not support the use of network addresses as channel
>    bindings.  This should not cause any security problems, as the use of
>    network addresses as channel bindings is not generally secure.
> 
> appears to me imply that network address alone is not secure as a
> channel binding, but that isn't the same.

How is the quote ("...not generally secure") not in agreement with what
Shawn said ("not very secure")?

> If there is something inherent insecure about using network addresses in
> channel bindings, I'd like to understand more.
> 
> As far as I can tell, there is no security problem in using network
> addresses in channel bindings.  There is the obvious and major
> deployment problem wrt NAT-like situations though.

Indeed, no security _harm_ results from using channel binding to network
addresses.

However, channel binding to network addresses adds little or no
security.  It can only add security when the network is physically
secure or IPsec (or equivalent, for non-IP networks) is in use, and then
only if a) IPsec protection is end-to-end, b) the IPsec policy ties
peers and IP addresses together tightly, and c) the IPsec policy does
not change adversely during the life of the connection.

Channel binding to network addresses could be seen as a proxy for
channel binding to IPsec, but only in very specific circumstances,
circumstances that applications cannot check for without having IPsec
APIs, and if they do then they might as well have access to IPsec
channel APIs that export channel bindings.  There are no APIs for
determining that the network path between two hosts is physically secure
either.

Given that channel binding to network addresses adds little or no
security, may lead to a sense of false security, is difficult to use in
any way that does add security, and will cause problems when NAT is in
the picture, channel binding to network addresses is best discouraged.

Nico
-- 

Note: Messages sent to this list are the opinions of the senders and do not imply endorsement by the IETF.