5 ms·

Any regards to cache poisoning attacks? They're actually easier at the subdomain level, because you can keep asking for <rand>._dv.example.com to prevent cache

Any regards to cache poisoning attacks? They're actually easier at the subdomain level, because you can keep asking for <rand>._dv.example.com to prevent cache hits, all the while spamming back NS delegation answers for _dv to the cache servers with a long TTL. DNSSEC of course solves that, but a lot of people are going to roll their eyes at that solution and a lot of things don't support it. Just curious.

elliottinvent 4y ago
Thanks for this really interesting feedback – especially that the subdomain method could create more opportunities for cache poisoning. Interested to explore how a cache poisoning attack might be easier on a Domain Verification protocol TXT record compared to a DNS-01 TXT record. Given the scenario: an attacker is attempting to claim a domain (example.com) with service provider (SP1) who uses a DNS revolver (DR1) I don’t immediately see how the subdomain approach makes a cache poisoning attack easier. Couldn’t the attacker similarly keep asking DR1 for <rand>._acme-challenge.example.com and spam back NS delegation answers for _acme-challenge.example.com, with the goal that upon cache poisoning success, they could fraudulently claim example.com with SP1? I may have overlooked something here, if so please explain. With either method, DNSSEC would definitely seem the solution as you’ve suggested.
- silisili 4y ago
  I'm not well versed in ACME but in DNS, unfortunately. Yes, it would have the same issue if it works as described. I've read a whitepaper on the subject I described and the time to poison cache goes way, way down. Basically, anyone could pull it off. I think a solution here in the validation space is just avoiding the caching server altogether. If the validating server does something like dig +trace, it avoids this problem, especially if it prefers TCP.

elliottinvent 4y ago

Thanks for this really interesting feedback – especially that the subdomain method could create more opportunities for cache poisoning. Interested to explore how a cache poisoning attack might be easier on a Domain Verification protocol TXT record compared to a DNS-01 TXT record. Given the scenario: an attacker is attempting to claim a domain (example.com) with service provider (SP1) who uses a DNS revolver (DR1) I don’t immediately see how the subdomain approach makes a cache poisoning attack easier. Couldn’t the attacker similarly keep asking DR1 for <rand>._acme-challenge.example.com and spam back NS delegation answers for _acme-challenge.example.com, with the goal that upon cache poisoning success, they could fraudulently claim example.com with SP1? I may have overlooked something here, if so please explain. With either method, DNSSEC would definitely seem the solution as you’ve suggested.

silisili 4y ago
I'm not well versed in ACME but in DNS, unfortunately. Yes, it would have the same issue if it works as described. I've read a whitepaper on the subject I described and the time to poison cache goes way, way down. Basically, anyone could pull it off. I think a solution here in the validation space is just avoiding the caching server altogether. If the validating server does something like dig +trace, it avoids this problem, especially if it prefers TCP.