Re: [TLS] Security concerns around co-locating TLS and

[Martin Rex <[e-mail]>]

Yoav Nir wrote:
> 
> On Nov 9, 2010, at 6:28 PM, t.petch wrote:
> > 
> > I recently was involved in syslog over TLS and the RFC there
> > explicitly calls it out.  Diagnostic messages are good for
> > diagnosing problems and so should be sent in the clear unless
> > there is a reason to do otherwise (eg relating to security
> > breaches, password misuse etc) and this is a general truth
> > in operations (SNMP etc).  Yes, I must have integrity, but
> > privacy can be a real pain, stopping the determination of problems.
> 
> Nobody is forcing privacy down your throat. Just as IPsec has
> AH and ESP-NULL, TLS has TLS_RSA_WITH_NULL_SHA256 and
> TLS_ECDHE_ECDSA_WITH_NULL_SHA among others.

There is much more DoS-potential in the server endpoint identification.

If you _can_ prevent syslog entries from being written by messing with
OCSP requests of the client that wants to submit syslog entries,
then you have a vulnerability -- as would be any kind of insistence of
involvement of third parties for a TLS-enabled syslog client.


The popularity of TLS is not that it is good, efficient or secure.
The primary motivator is that implementations are available and
it allows everyone to continue using weak passwords to authenticate
end users (and noone really cares about authenticating servers,
never did).


-Martin