Hi,
Sorry if this is a bit OT. Recently several different vendors (in completely different fields) where they white label for us asked us to remove A records that we have going to them and replace them with CNAME records. Is there anything going around in the security aranea that has caused this?
Hi,
They will set a dynamic IP address on the server, or use a CDN service.
It’s not a security thing. We do this with the the resellers who white label our VOIP. CNAMEs allow us to be flexible with our own hosts and infrastructure without having all of our resellers change DNS records.
Ray Orsini
Chief Executive Officer
OIT, LLC
305.967.6756 x1009 |
305.571.6272
How are we doing? We’d love to hear your feedback. https://go.oit.co/review
Interesting. We got a few requests at the same time which is what made we wonder. I wanted to make sure that there wasn’t something I was missing.
Are you using A records in a domain you own and pointing at their IPs? I’m not aware of any security vulnerability exploiting A vs CNAME. If they are hosting on a domain they own vs one you own, the use of a CNAME would allow them to change the A record IP without less impact to you, it would also allow them to remove the A record and effectively stop traffic targeting the host via a resolved IP.
While the change from A to CNAME itself is probably not based on security considerations, a CNAME pointing to a CDN or similar can result in future security issues, i.e. you want to closely monitor your externally pointing CNAMEs when you get rid of external services: https://www.hackerone.com/blog/Guide-Subdomain-Takeovers
Hi,
In my experience this has been the most abused security vector in the cases I saw.
I would add: Externally monitoring the signature of the non static objects (html, javascript) returned by the CDN.
While you can easily recover from image defacing, having your customers getting their private information (creds, identity, CC) stolen is another ball game.
Hi,
Sorry if this is a bit OT. Recently several different vendors (in completely
different fields) where they white label for us asked us to remove A records
that we have going to them and replace them with CNAME records. Is there
anything *going around* in the security aranea that has caused this?
Security-wise, you should be good. But make sure you're not attempting to deliver e-mail to such a domain; CNAMEs cannot be used in MX records.
Thanks,
Sabri
Or NS records, since you mentioned it. 
Doug
The closest thing to a *security* issue I can think of is IP agility in the
face of DDoS attacks -- most booter-style attacks are dumb as rocks, and
null-routing the target IP and moving all the customers on that IP to
another one is the easiest solution.
However, there are many *other* great reasons to get customers to CNAME onto
their SaaS vendors, including:
* No need to coordinate routine renumbering events;
* IPv6 support;
* CAA record (SSL cert issuance) support; and
* no doubt a bunch of other reasons I've forgotten for the moment.
Basically, if you sign up for a SaaS that uses your own domain and they
*don't* give you a CNAME target to point at, I'd be very cautious, because
they're either *very* new to the game, or they're probably also
operationally deficient in a lot of other areas, too.
- Matt
except - don't forget that the root of a domain (that domain without "www." or any other label) - cannot have a CNAME as the "A" record - fwiw...
Sorry if this is a bit OT. Recently several different vendors (in
completely different fields) where they white label for us asked us to
remove A records that we have going to them and replace them with CNAME
records. Is there anything *going around* in the security aranea that has
caused this?The closest thing to a *security* issue I can think of is IP agility in the
face of DDoS attacks -- most booter-style attacks are dumb as rocks, and
null-routing the target IP and moving all the customers on that IP to
another one is the easiest solution.However, there are many *other* great reasons to get customers to CNAME onto
their SaaS vendors, including:* No need to coordinate routine renumbering events;
* IPv6 support;
* CAA record (SSL cert issuance) support; and
* no doubt a bunch of other reasons I've forgotten for the moment.Basically, if you sign up for a SaaS that uses your own domain and they
*don't* give you a CNAME target to point at, I'd be very cautious, because
they're either *very* new to the game, or they're probably also
operationally deficient in a lot of other areas, too.- Matt
except - don't forget that the root of a domain (that domain without "www.”
or any other label) - cannot have a CNAME as the "A" record - fwiw…
Which is why there are HTTPS and SVCB records coming and SRV exists.
You don’t need CNAME, you need indirection. Indirection does require
a small amount of client support.
Sorry if this is a bit OT. Recently several different vendors (in
completely different fields) where they white label for us asked us to
remove A records that we have going to them and replace them with CNAME
records. Is there anything *going around* in the security aranea that has
caused this?The closest thing to a *security* issue I can think of is IP agility in the
face of DDoS attacks -- most booter-style attacks are dumb as rocks, and
null-routing the target IP and moving all the customers on that IP to
another one is the easiest solution.
DNSSEC?
A lot of public sector/government stuff, at least around here, should
have had DNSSEC enabled already.
e-Boks, as being the stuff that all state/municipalities sends
electronic communication through (unless you're excluded from
"electronic mail"):
-> https://dnssec-analyzer.verisignlabs.com/www.e-boks.dk
Sure, there DNSSEC on the actual domain name, but the CNAME
*destination* does not.
Or for another examples:
-> https://dnssec-analyzer.verisignlabs.com/www.nsa.gov
There is also DNSSEC enabled on this domain too, but again, the CNAME
*destination* does not.
Wasn't there once a phrase saying something like "a chain is no stronger
than its weakest link"?
What if the SaaS provider is actually the weakest link?
However, there are many *other* great reasons to get customers to CNAME onto
their SaaS vendors, including:* No need to coordinate routine renumbering events;
* IPv6 support;
* CAA record (SSL cert issuance) support; and
* no doubt a bunch of other reasons I've forgotten for the moment.
Renumbering and CAA record indeed two potential good reasons for using
the CNAME, as they wouldn't require clients to perform any manual
actions on their end.
However, I haven't seen anything pointing the direction that "IPv6
support" and "CNAME" would have anything to do with each other.
In the end, using A/AAAA directly is the matter of knowing what you do,
and if you really do, IPv6 support with or without the CNAME wouldn't
really matter.
Basically, if you sign up for a SaaS that uses your own domain and they
*don't* give you a CNAME target to point at, I'd be very cautious, because
they're either *very* new to the game, or they're probably also
operationally deficient in a lot of other areas, too.
Providing the CNAME, or even requiring the use of it, could also mean
that you should indeed take a close look, at the areas where the SaaS
provider giving you them become "operationally deficient" too.
Hasn't DNS often been criticized of being one common thing that often
make websites slow?
-> https://github.com/PowerDNS/pdns/issues/6874
Real life example from one of the many "SaaS" vendors (in the example,
CDN providers) out there, providing the CNAME, and - obviously depending
on how you look at it, may operate certain things in a very silly way.
My truth? There is too many things out there, making it impossible to
blindly believe that SaaS vendors would always be right, or that their
decisions are always the best.
Your truth? I believe you need to figure out that one yourself.
Just my two cents.
Yes. I didn't think that was something that needed to be explained on NANOG,
though.
- Matt
except - don't forget that the root of a domain (that domain without "www."
or any other label) - cannot have a CNAME as the "A" record - fwiw...Yes. I didn't think that was something that needed to be explained on NANOG,
though.
Given the number of ISPs (and others) that ask ISC to support CNAME at the APEX
to whom we have to politely say:
“No. It is not permitted by this part of RFC 1034.”
<quoted text>
It’s well worth reiterating.