On 10/26/11 9:39 PM, "Yaron Sheffer" <[email protected]> wrote:

>There is a common use case where we don't worry about malicious spokes,
>i.e. where they are all trusted.
>
>We do worry about misconfigured spokes, but that would most likely
>result in loss of connectivity, which I expect to be fixed in due time.
>Or it can be prevented by (static) configuration on the hub.

Malicious/misconfigured/compromised, it doesn't matter. One way or another
we have a spoke that is configured to protect the 69.171.0.0/16 subnet,
that includes all facebook.com servers.

If the hub learns the protected network of the spokes from the spokes,
then this information will propagate to the entire network. The
configuration is easy, because each protected network needs to be
configured in one place only, but we have the security problem. Maybe the
use of sidr resource certificates could mitigate this, but I don't think
they're in wide use yet.

On the other hand, we could have all the protected networks of the spokes
statically configured on the hub, as you suggest. Then it doesn't matter
if the spoke thinks it owns facebook, because none of the others will
agree. This has several implications:
 - Configuration is harder and error prone - spoke and hub must match
 - It introduces an asymmetry. IPsec is all about peers with separate but
matching configuration. In this scenario there is an all-knowing hub that
introduces the spokes to one another.

I'm not necessarily saying that the second scenario is bad, but it's not
as scalable as the first. I'd rather solve both, but solving just the
first may be a good first step.


_______________________________________________
IPsec mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/ipsec

Reply via email to