About random prefix attacks
Random prefix attacks are when someone sends a lot of traffic to subdomains that are highly unlikely to exist (
abcdefg.example.com), but are still associated with your main domain (
Usually, a DNS query to each random subdomain (or prefix) is not repeated, so it cannot be cached by resolvers or any other proxies and always reaches the authoritative nameservers. Rate limiting or blocking queries based on source IP can introduce a high amount of false positives, since random prefix attacks commonly are conducted via public resolvers. This makes these attacks particularly effective and hard to mitigate.
Queries for nonexistent domains
If the request only involved nonexistent domains, the
NXDOMAIN errors would only be served by the top-level domain (TLD) nameservers for
com.. This means that the queries never reach the authoritative nameservers.
Queries for nonexistent subdomains
These attacks are successful because they target subdomains, which require a response from a domain’s authoritative nameservers.
With an attack against a subdomain of an existing domain, the resolver is forced to fully resolve it against the authoritative nameservers since these random subdomains are likely not cached by the resolver or any other proxy. If an attacker sends enough of these queries, and the authoritative nameservers cannot handle the query load, it will become unresponsive or even fall over, taking all zones it is hosting down, not just the attacked zone.
This attack is difficult to mitigate for a few reasons. From the perspective of the authoritative nameservers, the attacker appears to be Cloudflare (
18.104.22.168) since that is the source of the queries. Blocking Cloudflare is not an option since that will block legitimate traffic.
When we detect an attack, we will temporarily stop querying your upstream nameservers for subdomains, sub-subdomains, and more. Cloudflare will then respond with cached responses (if their TTL has not yet expired).