Re: [MEXT] feedback on draft-ietf-firewall-[vendor|admin]-05

[Suresh Krishnan <[e-mail]>]

Hi Marc,
  Thanks for your comments. Please see responses inline.

On 11/28/2011 03:15 AM, Marc Lampo wrote:
> Hello,
> 
> Looking at the list archive, I see no comments on these drafts, published
> October 17th.
> Nor do I find any traces of discussions on IETF-82.
> 
> However, from a security point of view, I do have a remark.
> In both drafts, the "Security Considerations" state :
> (quoting extracts)
> ... "recommends a liberal setting of firewall rules" ...
> ... "some malicious traffic may be permitted" ...
> ... "allow the initiation of Denial of Service attacks against Mobile IPv6
> capable nodes" ...

This selective quoting misses out the context. This text only refers to
MIPv6 *signaling* traffic that is *encrypted*. I am not sure what more
can be done here. If you get an encrypted packet at a firewall, you can
either drop it or allow it to pass.

> 
> Is this acceptable ?
> Being concerned with security, I'd say that these statements as "Security
> Considerations"
> might lead to the conclusion that these mobility extensions cannot be
> used, in practice ?!

> 
> We hear that mobile networks might be the first to be IPv6 only (if there
> aren't already such networks around)
>  --> IPv6 only in mobile nets : the target for Mobility Extensions
> We see that more and more banks offer clients for smartphones for Internet
> banking
>  --> the "correspondent node" might very well be in some bank's DMZ

These blanket IPsec rules are applicable only for packets to/from the HA
address and not any arbitrary address. They are not relevant to the CNs.

> 
> ??? if they, bank security admins, will accept to setup security devices
>     in such a way that "firewall rules are liberal"
>     and "potentially allow that DOS attacks are started" ???

I am not getting the reference to bank security admins here. The
encrypted packets are between the HA and the MN. So the firewalls that
need to handle them are in the home network (housing the HA) and the
network to which the MN is attached.

> 
> It is my impression that, to put mobility information so "deep" in the
> stack
>  - as "extension header" with IPPROTO_NONE in the next header field -
> implies that security infrastructure will need quite some CPU resources
> to work its way through them.
> Wouldn't it have been wiser to assign a (UDP ?) portnumber to the Mobility
> "application" ?

This is how it is done for DSMIPv6 (port 4191) as described in Section 4.1.

Thanks
Suresh