I recall that subdomains are their own record inside a DNS, which would imply that anyone can claim that their server is a non-existent subdomain of the real domain
There are a lot of answers here but I feel they mostly miss OP’s point so I’ll try my own:
What stops a scammer from HTTPS certifying foobar.reputable.com is the trust system.
Anybody can create a certificate on their machine for anything within seconds, even you could create a certificate for www.google.com. The problem is that you, as an issuer, are not trusted by anybody.
Browsers and operating systems are released with a list of issuers that are considered trustworthy, so if you want your certificate to be recognised it has to come from one of these, not from you.
All of these issuers are in the list because they have been individually vetted, and are known to do their due diligence before issuing certificates, so they would not give you that cert unless they know that the bank domain or subdomain belongs to you, and the technical means to achieve this have been explained in other answers.
But if one of these issuers went rogue, or if you hypothetically hacked into their certification authority, then indeed nothing would stop you from obtaining a valid and recognised certificate for foobar.bank.com.
This is why for example Trustcor was removed from this list in 2022: from that position it would be trivial for a certificate authority to allow third parties to spy on people.
And when you are requesting a certificate for foobar.bank.com, your certificate request must come from an authorized email address at bank.com. That is also where your issued certificate would be sent. So, in order to get a certificate from a third party issuer, you have to:
- Control the domain registration at the level just above the TLD (I don’t know how it works for co.uk, probably similar though)
- Have access to a mailbox at the domain, where that mailbox has an address which is authorized to request certificates (this would be configured in the domain registration)
Could a malicious actor compromise that mailbox in a way that allows them to request a certificate and then receive it? It’s not impossible, but it would be a huge effort with a small payout. Honestly, if you’ve got access to that mailbox, you don’t want to give yourself away by making false certreqs through it. You want to just exfiltrate as much data from it as you can. There’s certainly something way more valuable in there.
your certificate request must come from an authorized email address at bank.com
That isn’t true in general. In fact, it can’t be.
It might be policy for most cases from the well-known certificate authorities, but it’s not part of the protocol or anything like that.
If it were, then it would be impossible to set up your mailserver to begin with because you could never get a certificate for mail.bank.com
Yeah, letsencrypt doesn’t do this for example. They do ask for an email address, but that’s just for expiry notices.
They do require you control the domain, and run it on the server the DNS record points to. When using certbot at least.
I was wrong again today.
To claim a subdomain in the DNS system you have to have the domain first.
I recall that subdomains are their own record inside a DNS
Well, not a record, but a zone. A subdomain is its own zone. There are additional DNS records that support a separate zone though.
which would imply that anyone can claim that their server is a non-existent subdomain of the real domain
False. The person wanting control of the subdomain must be delegated control from the parent domain. Owners of the parent domain don’t just hand that out to anyone. The mechanism is called DNS Delegation.
The way DNS works, each dot is authoritative.
So if you want the IPv4 for scam.legitco.com, your computer contacts the authoritative DNS for “com” and asks it for the address for legitco’s DNS. You then contact legitco.com and ask it for scam’s IP. Which it won’t have.
This is simplified, because in reality there’s DNS caching and pooling, but that wouldn’t affect your scenario. Although, cache poisoning IS a thing, as is BGP hijacking where the IP of the DNS itself may get redirected to a different machine.
They’d need a certificate authority to issue the certificate, and the victim’s browser would have to trust that authority.
Edit: and the scammer would need to control the domain DNS server to use the subdomain, like another reply said, so the certificate alone wouldn’t help much.
I’ve been able to downgrade https sites to plain http sites, through a series of loopholes which I won’t go into.
So you’ve… compromised your own security. Grats?
That’s nice, be sure to tell us how it goes when HSTS is enabled
Checks own servers…
Strict-Transport-Security: max-age=63072000; includeSubDomains; preload
Yeah, I’d like to see that…
Something in the way of “an apartment key is useless if you can’t get into the building”.
There were some rather in detail answers already to which I could add. But instead I am going with a more simple answer that is hopefully also good.
Basically, bad actors want to stay undetected if possible. Like staying in dark places with dark clothing and not making noise. trying to get your own subdomain is more like wearing a high Vis jacket, having Christmas lights on you and broadcasting your presence with a loudspeaker with something like “catch me if you can!” on repeat.
Or even simpler: getting detected is bad for bad actors, doing that is one great way to get detected, they know that so they don’t do that.
Or metaphorically: a drop of water in the ocean won’t get noticed, rain in the desert will.
At the end of the day it’s not about what you can do but if you should do that.
