Re: [OPSEC] [tcpm] draft-gont-tcp-security

[Fernando Gont <fernando at gont.com.ar>]

Joe Touch wrote:

>>> 	The diligent blacksmith knows that hardening a tool also
>>> 	makes it more brittle...
>> This is a nice quote, but... I'd like examples. e.g., start discussing
>> about which specific hardening proposal makes TCP more brittle.
> 
> 1) any security mechanism that increases complexity - of actions, state,
> or message exchanges - any of which increases the potential for
> implementation error

Agreed.



> 2) any security mechanism that has false positives, i.e., that discards
> messages deemed a security threat when they were sent for legitimate reasons

Why would this make e.g., TCP more brittle?

In any case, the actual response to such packets may vary (e.g., in the
case of ICMP hard errors, discard vs. process as soft errors). I believe
that no matter what the recommended response is, it is important to
discuss these issues, and try to get consensus on what's the right thing
to do in each case.


> #1 includes basically everything, from TCP MD5 (and TCP-AO) to tcpsecure
> and ICMP filtering

ICMP filtering actually decreases complexity.



> I.e., AFAICT, *everything* that makes TCP more secure also makes it
> brittle, by definition (ditto for metal hardening, FWIW). The key issue
> is "when/where is the benefit worth the cost".

As I said before, I'd like to have concrete examples from the tcp
security i-d that are deemed to make TCP more brittle.

Thanks!

Kind regards,
-- 
Fernando Gont
e-mail: fernando at gont.com.ar || fgont at acm.org
PGP Fingerprint: 7809 84F5 322E 45C7 F1C9 3945 96EE A9EF D076 FFF1