I would definitely however spend effort into verifying a host key that changes unexpectedly.
My attempts to convince them to use the same key came to naught, so instead I use one of the IP addresses.
I could alternately erase the known hosts entry on each transfer. That would probably have been preferable.
I also got a shell on it when I attempted ssh, so you can guess the care that is taken with it.
I was actually personally a victim of such (unsuccessful) attack in the tor network. SSH login to my hidden service complained about a wrong fingerprint. The same happened when I tried again. After tor restart, problem disappeared. I assume this was an attempt at SSH mitm by one of exit nodes?
Or something like that.
Good idea. That way when your CA private key leaks (the key which we never ever rotate, of course) the bad guys can compromise the whole fleet and not just one server. Bonus points if the same CA is also used for authenticating users.
As with X.509, any serious usage will involve a hardware security module, so that compromise of the CA host does not allow the key to be leaked. You'd still have a very bad day, but it can be mitigated.
I do think it's a fairly significant flaw that SSH CA doesn't support intermediate CA's (or at least didn't last time I looked into it) to enable an offline root CA.
>Bonus points if the same CA is also used for authenticating users.
The SSH CA mechanism can be used for both Host and User auth, yes.
Keeping in mind, in a real use case this would be tied to something like active directory / LDAP, so you can automate issuance of ssh keys to users and hosts.
Systems configured to trust the SSH CA can trust that the user logging in is who they say they are because the principal has already been authenticated and vouched for by the identity provider, no more manually managing known_hosts and authorized_keys, or having to deal with Trust On First Use or host key changed errors.
You can also set the CA's endorsement of the issued keys to fairly short lifetimes, so you can simplify your keymat lifecycle management a great deal - no worrying about old keys lying around forever if the CA only issues them as valid for an hour / day / etc. .
Overall I think you still come out ahead on security.
If you aren't bothering to verify then they do not need to trick you at all.
In DayJob we have a lot of clients send feeds and collect automated exports via SFTP, and a few to whom we operate the other way (us pulling data via SFTP or pushing it to their endpoint). HTTPS based APIs are very common and becoming more so, but SFTP is still big in this area (we offer some HTTPS APIs, few use them instead of SFTP).
One possible exploit route, for a malicious actor playing a long and targetted game, that could affect us:
1. Attacker somehow poisons our DNS, or that of a specific prospective client of ours, sending traffic for sftp.ourname.tld to their server, and has access to our mail.
2. Initially they just forward traffic so key verification works for existing users. They monitor for some time to record host addresses that already access the server, so when they start intervening they can keep just forwarding connections from those addresses, so those users see no warnings (and are unaffected by the hack).
3. When they do start to intercept connections from hosts not already on the list make above instead of forwarding everything, existing users are unaffected¹ but new users coming in from entirely different addresses now go to their server and if they are not verifying the key will happily send information through it², authenticating with the initial user+pass we sent or PKI using the public key they sent, with the malicious server connecting through to ours to complete the transfers.
4. Now wait and collect data as no one realises there is a MitM, and later use any PII or other valuable information for ransom/extortion purposes.
Of course there are ways to mitigate this attack route. For one: source address whitelisting, supported by OpenSSH's key based auth as the acceptable source list can be included with the public key so only specific sources can use that key for auth. But they client would have to make effort to do this, and if they aren't going to make the effort to verify the host key then they aren't going to make other efforts either.
We do have some clients who verify the host properly and/or give us source addresses to limit connections to when they provide a public key, we work with financial institutions who are appropriately paranoid about their data and the data of their customers, some even use PGP for data in transit (and in case it is ever stored where it shouldn't be) for an extra level of paranoia. But most do none of this. Most utterly ignore our strong suggestion that they use keys, or change passwords in case of email breach, instead using the password we mail them before first connection for eternity.
--------
[1] none of our clients are likely to be sending files from dynamic source addresses, at most the source might move around a v4/24 or v6/64, currently I don't think all of them connect from a single IPv4 address, I've had one recently let us know (months in advance) that their source address will be changing.
[2] it can connect to us and send the data
Literally happens every single damn day and literally nobody on the face of this earth ever gives a shit.
Host keys are the stupidest idea in the history of computer so-called "security".
Mine change maybe once every couple of years, if I do a full reinstall without copying over the old host key. And then I know exactly why it changed.
Nobody knows how the hell the host keys are generated in the first place. Don't worry about it.
> And then I know exactly why it changed.
Really? What is a "full" reinstall as opposed to a "non-full" reinstall, and how much exactly reinstall do I need for my host keys to change?
If my host keys were changing regularly, I would worry about it. There's no legitimate reason for that to be happening, since I'm not regularly wiping the drive and reinstalling, nor am I regularly manually deleting the host keys (the other way they get regenerated).