Re: [Tsvwg] Port Randomization issues summary

[Joe Touch <[e-mail]>]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1



Fernando Gont wrote:
> Joe Touch wrote:
> 
>>>> 1) randomizing connection IDs helps protect against blind attacks
>>>>
>>>> 	- but ultimately the only way to avoid old segments
>>>> 	from interfering with new connections is to keep state
>>> Yes. But... what does this have to do with port randomization?
>> It underscores the need to keep state in the endpoints, rather than just
>> trust that randomization will provide safety.
> 
> Are you talking about the vtag randomization thing you were discussiing
> with Anantha?

Yes. I think you already have this in your draft, though. It differs
from what SCTP describes, but as you saw there's already a suggestion to
address that separately.

>>>> 	- randomness requires keeping more state at the endpoints
>>>> 	than sequential use of the ID space
>>> Why?
>> If IDs are used in sequence, and endpoint can keep track of "do not use"
>> ones using two IDs (representing the range to be avoided). If IDs are
>> used randomly, then the endpoint needs a copy of each value to avoid.
> 
> Again, I guess you msut be referring to some vtag-thing here, right?

Well, yes and no. This also refers to your solution; if you select
source ports in sequence, it may be simpler/faster/less state to avoid
reuse during the required period than to do randomization. The point is
that randomization may increase protection against blind attacks, but it
comes at a cost of additional state at the endpoints. I don't know if
you say that explicitly, but it seems worth noting.

Joe
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.9 (MingW32)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iEYEARECAAYFAkofCdUACgkQE5f5cImnZrsHXQCfQiEOHotWfh7iMoxXz466ZYrx
qqwAoMx8WMIsg3kVGTtfl8EvFEtuvZkG
=SDi/
-----END PGP SIGNATURE-----