Comment by Arainach
11 hours ago
Not if Quad9 is using DNSSEC, no.
What's the specific threat you're envisioning? If it involves Quad9 themselves being malicious, what would DNSSEC on the forwarding prevent?
This page explains how all of this works in detail: https://quad9.net/news/blog/quad9-enables-dnssec-on-all-serv...
DNSSEC validation on your forwarder would prevent a maliciously modified record from Quad9 (or others) from being accepted locally - i.e. "Quad9 can poison my DNS".
I've always been of two minds on this. On one hand, that concern is beyond any reasonable level of security/performance/reliability tradeoff for most any user. At the same time, it is a bit of a shame DNS doesn't have a more scalable & performant approach to security which can just always be done without having to consider it a tradeoff, however minor in practice.
You are rely here on the assumption what your resolver already knows what the zone is DNSSEC signed. If your forwarder or resolver strips that information?..
At a high level, one of 3 things happens:
1. The forwarder gets a response claiming the record is supposed to be DNNSEC signed from the parent (recursively traversing from the root). The forwarder checks the signature of this claim. If the signature is valid, the forwarder continues on to validate the signature of the record and checks its validity to know if the info was secure. If the signature is invalid, the forwarder knows any information any information is not able to be validated as secure
- Somewhere during the recursive checks through the root, the forwarder gets an unsigned (no DNSSEC) or invalidly signed (e.g. your stripped response) response. The forwarder knows any information is not able to be validated as secure.
- A claim for lack of configuration or support of DNSSEC records comes back. The forwarder knows any information is not able to be validated as secure.
So you always know whether or not the information was secure, it's just if it was insecure you don't know if it's because it was just never secured or if someone tried to tamper with it. And that should make sense, an insecure message is by definition one which you can't tell if it has been tampered with.
Wait, I must be misunderstanding you, because if you're resolving off Quad9, they can definitely poison your DNSSEC-signed records. Between a stub resolver and a recursor DNSSEC collapses down to a single "yes it was signed" bit in the header.
To protect yourself from an upstream resolver using DNSSEC, you need to be doing something akin to a full recursive lookup yourself. This is a flaw in the DNSSEC design and a reason why DoH took off instead.
I don't know the details but knot-resolver asks for DS records of the domain being looked up when forwarding to avoid needing to do a full recursive lookup to validate DNSSEC. As I recall from what I've read this works almost everywhere, including Quad9, but not Google DNS due to a bug that they claim to have identified three years ago but haven't fixed. But as I understand it this is not how recursive resolvers validate so it is extra data that the forwarding server needs to request and cache.
> To protect yourself from an upstream resolver using DNSSEC, you need to be doing something akin to a full recursive lookup yourself. This is a flaw in the DNSSEC design and a reason why DoH took off instead.
What? I don't see how you can call that a flaw in DNSSEC when DoH is no better in this regard; it doesn't even attempt to protect against a malicious recursive resolver. The only way to do that is to validate DNSSEC on the client.
DNSSEC and DoH provide different security services. But to get the benefit of DNSSEC, you need to resolve recursively. DoH works for stub resolvers. That's all I'm saying.