4 ms·

Ask HN: Is sharing .ssh/authorized_keys a security issue?

8 points by kator 14y ago

I noticed a .ssh/authorized_keys in a github dotfiles collection recently and wondered if people think this is dangerous?

kator 14y ago
My gut tells me this is TMI for whomever might want to hack a site. Just knowing what keys are allowed to ssh w/o a password seems like a great target. Not sure what others think?
- chilton 14y ago
  Under normal circumstances it should be impossible to generate the private key that corresponds to public key. This is because a public key is mathematically derived from it's private key in a way that both forces a brute force test of the entire key domain as the only attack and requires that the each individual test is expensive. In general sharing public keys should be safe. There are be exceptions: If the key generator is somehow flawed the brute force search can be narrowed. Also, the rate you can test is constantly increasing so in 1995 512 bit RSA keys were "safe" but I wouldn't trust them now. A realistic attacker probably isn't going to attack the public key but if the key was generated by "rob@example.com" he may try attacks against publicly visible machines in example.com to obtain the correct private key. That's probably the worst case scenario and it also may not bear fruit. Don't forget that you can put a pass phrase on your private key to enhance your own security.
patio11 14y ago
If that account had id_rsa in it then I'd send them a courtesy email, since that is a private key, but otherwise publishing your public keys is not worrisome. Consider the threat environment: I have a laptop with a private/public key pair that gets me my main login on all services I run. It is possible that Server #1 can get compromised, immediately compromising the public key on it. I do not desire Server #2 to get compromised. No problem -- with only the public key, they can't compute my private key. Also, when I authenticate with them, I don't have to divulge my private key -- we can play a little game over the wire and my system provides an answer which would have been computationally unfeasible to calculate without having my private key handy. There might be some minor data leakage if you name your keys something descriptive. Mine are typically no more sensitive than email addresses.
ProblemFactory 14y ago
The .ssh/authorized_keys contains public keys. Technically, these can be completely public, as it is infeasible to compute the corresponding private key to get access. However, keeping it secret may be useful, in case a hacker compromises another computer that you own with private keys on it. If the authorized_keys file is published, they immediately know which private keys are useful for remote logins. If the authorized_keys file contains source-machine names and usernames in the comment, it gives the hackers a list of your other machines to attempt compromising to obtain the private keys. If the authorized_keys file uses from="address" restrictions, publishing it gives the hackers a list of allowed source networks for remote logins. If they try to log in with the correct private key from the wrong network, they are denied access (and may assume the key is invalid). In summary, it's technically not a security risk, but there is no reason to publish it and make life easier for hackers if they do happen to stumble on one of your private keys.

8 points by kator 14y ago

I noticed a .ssh/authorized_keys in a github dotfiles collection recently and wondered if people think this is dangerous?

kator 14y ago

My gut tells me this is TMI for whomever might want to hack a site. Just knowing what keys are allowed to ssh w/o a password seems like a great target. Not sure what others think?

chilton 14y ago
Under normal circumstances it should be impossible to generate the private key that corresponds to public key. This is because a public key is mathematically derived from it's private key in a way that both forces a brute force test of the entire key domain as the only attack and requires that the each individual test is expensive. In general sharing public keys should be safe. There are be exceptions: If the key generator is somehow flawed the brute force search can be narrowed. Also, the rate you can test is constantly increasing so in 1995 512 bit RSA keys were "safe" but I wouldn't trust them now. A realistic attacker probably isn't going to attack the public key but if the key was generated by "rob@example.com" he may try attacks against publicly visible machines in example.com to obtain the correct private key. That's probably the worst case scenario and it also may not bear fruit. Don't forget that you can put a pass phrase on your private key to enhance your own security.

patio11 14y ago

If that account had id_rsa in it then I'd send them a courtesy email, since that is a private key, but otherwise publishing your public keys is not worrisome. Consider the threat environment: I have a laptop with a private/public key pair that gets me my main login on all services I run. It is possible that Server #1 can get compromised, immediately compromising the public key on it. I do not desire Server #2 to get compromised. No problem -- with only the public key, they can't compute my private key. Also, when I authenticate with them, I don't have to divulge my private key -- we can play a little game over the wire and my system provides an answer which would have been computationally unfeasible to calculate without having my private key handy. There might be some minor data leakage if you name your keys something descriptive. Mine are typically no more sensitive than email addresses.

ProblemFactory 14y ago

The .ssh/authorized_keys contains public keys. Technically, these can be completely public, as it is infeasible to compute the corresponding private key to get access. However, keeping it secret may be useful, in case a hacker compromises another computer that you own with private keys on it. If the authorized_keys file is published, they immediately know which private keys are useful for remote logins. If the authorized_keys file contains source-machine names and usernames in the comment, it gives the hackers a list of your other machines to attempt compromising to obtain the private keys. If the authorized_keys file uses from="address" restrictions, publishing it gives the hackers a list of allowed source networks for remote logins. If they try to log in with the correct private key from the wrong network, they are denied access (and may assume the key is invalid). In summary, it's technically not a security risk, but there is no reason to publish it and make life easier for hackers if they do happen to stumble on one of your private keys.