157

I have two files, id_rsa and id_rsa.pub. What command can be used to validate if they are a valid pair?

1
  • I'll confirm Michuelnik's answer; it saved me from having to make a new key pair, thanks. ssh -v helps a lot too.
    – Krista K
    Dec 30, 2013 at 5:06

5 Answers 5

200

I would prefer the ssh-keygen -y -e -f <private key> way instead of the accepted answer of How do you test a public/private DSA keypair? on Stack Overflow.

ssh-keygen -y -e -f <private key> takes a private key and prints the corresponding public key which can be directly compared to your available public keys. (Hint: beware of comments or key-options.)

(How the hell is it doing that? I can only hope the public key is encoded directly or indirectly in the private key...)

I needed this myself and used the following Bash one-liner. It should output nothing if the keys belong together. Apply a little -q to the diff in scripts and diff only sets the return code appropriately.

PRIVKEY=id_rsa
TESTKEY=id_rsa.pub
diff <( ssh-keygen -y -e -f "$PRIVKEY" ) <( ssh-keygen -y -e -f "$TESTKEY" )
9
  • 6
    "How the hell is it doing that".. The private key is generated from randomness so we all have a unique key. If you apply the repeatable one-way RSA algorithm on this private key, youll get the unique public key. There is no algorithm that you can perform on the public key thatll get a unique private key.
    – Sirch
    Sep 24, 2013 at 14:29
  • 1
    @Sirch: I thought the decision which key is private and which one is public is pure random since the two keys are equal. What one key encrypts can only be decrypted with the other. And if one key could be obtained from the other this all would not work out.
    – Michuelnik
    Sep 25, 2013 at 11:42
  • 3
    @Michuelnik You can derive the public key from the private key. You cant derive the private key from the public key. Were not talking about the material that it encrypts.
    – Sirch
    Sep 25, 2013 at 13:35
  • 1
    @Michuelnik imho, the question is off topic on SO and on-topic here (and/or Superuser). imho, it shouldn't be marked as duplicate and instead flagged there for migration. But it's covered on both so I like the more complete sharing of information.
    – Krista K
    Dec 30, 2013 at 20:39
  • 23
    As long as id_rsa.pub exists, ssh-keygen -y -e -f id_rsa will not check id_rsa at all but just return the value from id_rsa.pub. So e.g. if you echo 5 > id_rsa to erase the private key, then do the diff, the diff will pass! Also, running ssh-keygen -yef foo where foo is not a valid key (and has no corresponding foo.pub) will block waiting for user input, so be careful using this in a script.
    – user68519
    Jul 10, 2015 at 22:45
80

Depending on where you get the public key file you are testing, the accepted answer may give false positive results. This is because of the behavior described in the comment by @drewbenn. Specifically, when the -e option is used with the private key file as the -f option parameter, it simply parrots (but reformats) what's in the associated public key file.

In other words,

ssh-keygen -y -f id_rsa

(apparently) generates the public key value, and

ssh-keygen -y -e -f id_rsa

simply and outputs (and reformats) the key in the existing id_rsa.pub whatever it is.

In my case, I have to verify that the pair has not been corrupted. So, I decided to compare the following:

ssh-keygen -y -f id_rsa | cut -d' ' -f 2

with

cut -d' ' -f 2 id_rsa.pub

Therefore:

diff <(cut -d' ' -f 2 id_rsa.pub) <(ssh-keygen -y -f id_rsa | cut -d' ' -f 2)

Perhaps this is not as flexible, but it is better for my needs. Maybe it helps someone else.

3
  • 9
    This really should replace the accepted answer or at least surpass it in terms of upvotes.
    – thomanski
    Oct 5, 2016 at 15:01
  • 4
    Thank you! This command doesn't work with keys with a passphrase, it doesn't ask that interactively. I extracted the two () command contents to files and diff-ed them, that works. Oct 18, 2019 at 10:10
  • 1
    @YaroslavNikitenko - I had the same problem. The answer by Olivier, which is similar but rearranged, removes the problem. Nov 26, 2021 at 12:08
14

The easiest is to compare fingerprints as the public and private keys have the same. Visual comparison is pretty easy by putting the two commands on same line:

ssh-keygen -l -f PRIVATE_KEY; ssh-keygen -l -f PUBLIC_KEY

Programmatically, you'll want to ignore the comment portion so

diff -s <(ssh-keygen -l -f PRIVATE_KEY | cut -d' ' -f2) <(ssh-keygen -l -f PUBLIC_KEY | cut -d' ' -f2)
4
  • This gives the result right away. Dec 29, 2021 at 13:45
  • 1
    This holds the same limit as mentioned in the answer above: it will not calculate fingerprint from the private key if public key file exists. Instead, it will just print out the fingerprint of whatever is in public key file. Therefore, you may have public key different from private key and the test will still pass.
    – raspy
    Jun 7, 2022 at 13:41
  • This CLI sequence assumes of course that the key does not require a password to decrypt. Sep 13, 2022 at 23:12
  • @JasonHemann considering what raspy wrote, this two-command-one-line would not even ask for a password if matching .pub file exists. Sep 21, 2023 at 7:44
6

If they're on your local system, stick id_rsa.pub in your $HOME/.ssh/authorized_keys and ssh to localhost using the id_rsa key. If it works, then they match.

cat $HOME/.ssh/id_rsa.pub >> $HOME/.ssh/authorized_keys
ssh -i $HOME/.ssh/id_rsa localhost
2
  • 1
    What if it's still prompting you for a password? Permissions of authorized_keys are correct, the public key has been copied correctly in the authorized_keys, I also did the diff mentioned in the verified answer and keys belong together.
    – Vasiliki
    Sep 30, 2020 at 15:39
  • This might be a little intrusive. The answer by Olivier is not. Nov 26, 2021 at 12:04
1

This is a very old thread, but I found the answers a little bit overengineered. So I just created yet another oneliner :)

[[ `ssh-keygen -yf /path/to/private_key_file` == `cat /path/to/public_key_file` ]] && echo "True" || echo "False"

There might be some edge cases because of the comments but it works fine for me since I never mess with them.

Hope it helps.

P.S.: And yes, please change that accepted answer. False positives might be problematic.

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .