Oh this is interesting. Often bug bounty’s claim a published private key wasn’t in use or sensitive. This gives researchers the ability to call BS on that.
One thing that you'll see happen is, something is supplied with an example or test key and at some point somebody who doesn't understand what's going on need a "Private key". Huh. Where can I get a "Private key"? Oh, here's one, I'll use that. Sometimes it's in the context where it was found, but sometimes far away from that.
For example, back in March 2020, somebody on m.d.s.policy wondered why seemingly unrelated Web PKI certs had the same private key. So I stared at the certs and I came to the following conclusion:
This company Paessler makes a Windows tool called PRTG with a web interface. It is supplied with a demo certificate. So you set up the tool on a Windows server and it basically works, but the certificate isn't trustworthy.
Some people will click "Ignore" and press on. This is horribly insecure but what's new?
However, some people will decide they need to get a Certificate. Getting a certificate requires a Private Key. Fortunately, PRTG is supplied with a Private Key so no need to go learn how to make one yourself.
So, the people who made PRTG didn't screw up here in the sense that they really did make a test or demo key, and from their point of view obviously it isn't sensitive since everybody has the same key.
Unfortunately their users may not realise and come to depend on this key as "their" private key, and so in this sense it is sensitive, just not for the people who made it.
--
Regardless of who minted the key you have, and whether they understood its importance, in the Web PKI the correct thing to do is always the same and should be pretty easy:
For Let's Encrypt you can actually do this mechanically, their API will accept proof you know the private key as grounds to revoke the certificate even though that's not listed above.
If the CA refuses to revoke a certificate you've genuinely proved you have the key for (follow any instructions carefully) or just goes silent with no action for a prolonged period, report this problem to m.d.s.policy yourself. You should not need to send the private key itself to anybody as "proof" you have it, the whole point of private keys is that you can prove you know this key without revealing it.
Committing private keys is definitely the wrong thing to do. But if the system is talking to something, then generating new keys won’t help either. What you need is some kind of system to deploy or retrieve secrets. There are various different solutions that exist for this.
Most private keys in Git repositories seem to be test data. But why are those test private keys sometimes used for other things? Probably just people lazily copying from ~/.ssh/idrsa or copying to ~/.ssh/idrsa.
Keys (at least rsa,ec,ssh ones) don’t expire - certs do. Also you’re not required to set expiration at all and probably should not in test for the very reason you mentioned. Unless you’re testing expiration validation of course in which case cert will be intentionally expired.
I'm somewhat surprised how many keygen type tools don't support ways to do that without putting a passphrase on a command line. Gpg is nice, with --passphrase-fd.
Congrats to Dylan and team on this release; incredibly cool - more companies, especially those for whom software is a functional organization but not the primary one -- need to take key leaks seriously.
Many are for tests and don't go to anything. Some go to really important things though, even among the test keys, and this tool tells you that instantly for billions of keys.
The public key can be derived from the private key
They then check both certificate transparency to see if the public key matches any certificates that have been generated, and to see if it's used by a github user (will this public key let me in to a github repo)
If neither, then it's not sensitive (well it might be, but only like finding a key on the floor in the street is -- won't do you much good without knowing where you can use it)
In the first case, if you have the private key, you can spoof the website
In the second case, if you have the private key, you not only have push access to the repos that user has (which could be quite wide ranging), but also you're likely able to get into many servers via SSH, as developers tend to use the same ssh key for github and for server access
What their latest software does is take your key and check it against these sources,
Now private keys have a further layer of protection - the passphrase. Turns out the majority of passphrases belonging to the leaked private keys are trivial ones.
Many leaked keys will unlikely to be used anywhere, but it turns out many more are.
