[nsp-sec] ccTLD Security - What are your top 5 "security" recommendations?

[Nick Hilliard]

On 11/03/2011 17:01, sthaug at nethelp.no wrote:
> I look at this from a slightly different point of view: You *want* the
> name service for your ccTLD to be distributed, both geographically and
> across different ASes. To do this, you obviously need name servers that
> are distributed correspondingly.

Agreed completely: just look at the recent problems with .ly and .eg.

> Unless your ccTLD organization is big, you're highly unlikely to own all
> the physical facilities yourself. So most likely *somebody* outside the
> ccTLD organization has physical access to some of the servers, and can
> therefore break into them.

Yes.  But breaking in takes more effort and involves much more noticeable 
disruption to services.  It will probably require a reboot, which is much 
easier to notice.  If you're running your dns server on a VM (some 
operators do), it's even more difficult because you can't even reboot to 
single user mode, without access to the underlying virtual machine.

> I don't see the big difference between such physical access and having
> third party people with root access running the name servers, given a
> reasonable (legally binding) *contract* between the ccTLD organization
> and the organizations of the third party people with root access. Such
> contracts certainly exist, for instance in the case of the .no ccTLD.

Contracts enforce procedure, which is generally a good thing.  But 
malicious intent will not be hindered by the existence of a contract.

Nick