[aur-general] [AUR4] Support of multiple ssh public keys

newer
[aur-general] Signoff report for...

older
[aur-general] Signoff report for...

Yuki Chiba

11 Jun 2015 11 Jun '15

7:18 p.m.

Hi, I prefer to use individual ssh public keys on different machines, but I could register only single key in my profile on aur4.archlinux.org. Is there any chance to support multiple ssh public keys like github[1] or Bitbucket[2]? [1] https://help.github.com/articles/generating-ssh-keys/#step-4-add-your-ssh-ke... [2] https://confluence.atlassian.com/display/BITBUCKET/Add+an+SSH+key+to+an+acco... Best regards, Yuki Chiba -- Yuki Chiba yuki.from.akita@gmail.com

Show replies by date

Giancarlo Razzolini

11 Jun 11 Jun

7:53 p.m.

...

I prefer to use individual ssh public keys on different machines, but I could register only single key in my profile on aur4.archlinux.org. Yes, you can only register one at this moment.

Is there any chance to support multiple ssh public keys like github[1] or Bitbucket[2]? I think that this is not as much relevant with AUR, because it's just instructions on how to build your package, some ocasional patch and the support files (install, systemd service, etc). The real development should be done on te services you mentioned, github, etc. I don't think

Em 11-06-2015 16:18, Yuki Chiba escreveu: the aur devs shouldn't sweat too much to implement this. You can, if you wish, register another user in the AUR, give that user co-maintainer rights. A dirty way to have more than one ssh key. Cheers, Giancarlo Razzolini

Lukas Fleischer

8:44 p.m.

On Thu, 11 Jun 2015 at 21:18:11, Yuki Chiba wrote:

...

Hi,

I prefer to use individual ssh public keys on different machines, but I could register only single key in my profile on aur4.archlinux.org.

Just out of curiosity: Is there any reason for using different SSH keys per machine (for the same server)? Or is it really just a preference?

...

Is there any chance to support multiple ssh public keys like github[1] or Bitbucket[2]?

That would not be too hard to implement. I might add it before the final 4.0.0 release.

...

[1] https://help.github.com/articles/generating-ssh-keys/#step-4-add-your-ssh-ke... [2] https://confluence.atlassian.com/display/BITBUCKET/Add+an+SSH+key+to+an+acco...

Best regards, Yuki Chiba -- Yuki Chiba yuki.from.akita@gmail.com

Remi Gacogne

8:56 p.m.

On 06/11/2015 10:44 PM, Lukas Fleischer wrote:

...

Just out of curiosity: Is there any reason for using different SSH keys per machine (for the same server)? Or is it really just a preference?

I can't speak for Yuki, but I do it in order to be able to revoke only the related keys if a machine is compromised/stolen/lost (FDE and strong passphrases only buy you some time to do it).

Giancarlo Razzolini

9:59 p.m.

Em 11-06-2015 17:56, Remi Gacogne escreveu:

...

(FDE and strong passphrases only buy you some time to do it). In the case of stolen/lost, it buy you a lot of time. Or you are aware of some cryptanalisys development I'm not aware of.

Now, if your machine is compromised, then I think that you might have bigger worries than the keys used to publish some packages on AUR. Cheers, Giancarlo Razzolini

David Kaylor

12 Jun 12 Jun

1:58 a.m.

On Thu, Jun 11, 2015 at 5:59 PM, Giancarlo Razzolini wrote:

...

Em 11-06-2015 17:56, Remi Gacogne escreveu:

...
(FDE and strong passphrases only buy you some time to do it).

In the case of stolen/lost, it buy you a lot of time. Or you are aware of some cryptanalisys development I'm not aware of.

Now, if your machine is compromised, then I think that you might have bigger worries than the keys used to publish some packages on AUR.

Cheers, Giancarlo Razzolini

That's certainly true, but it's not the point. Seperate, individually revokable keys are a good idea if someone will be submitting from multiple machines. And it would help protect AUR down the line. So if it's fairly easy to implement, like Lukas said, +1 on that.

Øyvind Heggstad

11:27 a.m.

On Thu, 11 Jun 2015 21:58:33 -0400 David Kaylor wrote:

...

On Thu, Jun 11, 2015 at 5:59 PM, Giancarlo Razzolini wrote:

...
Em 11-06-2015 17:56, Remi Gacogne escreveu:

...
(FDE and strong passphrases only buy you some time to do it).

In the case of stolen/lost, it buy you a lot of time. Or you are aware of some cryptanalisys development I'm not aware of.

Now, if your machine is compromised, then I think that you might have bigger worries than the keys used to publish some packages on AUR.

Cheers, Giancarlo Razzolini

That's certainly true, but it's not the point. Seperate, individually revokable keys are a good idea if someone will be submitting from multiple machines. And it would help protect AUR down the line. So if it's fairly easy to implement, like Lukas said, +1 on that.

Easiest way to attack a password protected private key: Just put a keylogger on the target. This is why we need u2f/similar support everywhere :/

Remi Gacogne

8:15 a.m.

On 06/11/2015 11:59 PM, Giancarlo Razzolini wrote:

...

In the case of stolen/lost, it buy you a lot of time. Or you are aware of some cryptanalisys development I'm not aware of.

I am not, but everything depends on your threat model. If you are targeted via an "evil-maid", or a cold-boot attack, FDE may be doomed. In addition to that, passphrase-protection on SSH keys has been weak for a long time, because a single MD5(IV || passphrase) is applied to generate the AES key used to encrypt the SSH key [1]. OpenSSL 6.5 introduced a new KDF [2] using bcrypt, enabled by default for ed25519 keys but not for RSA keys, so you may want to upgrade your keys to use the new KDF manually.

...

Now, if your machine is compromised, then I think that you might have bigger worries than the keys used to publish some packages on AUR.

Agreed :) [1] https://martin.kleppmann.com/2013/05/24/improving-security-of-ssh-private-ke... [2] http://www.tedunangst.com/flak/post/new-openssh-key-format-and-bcrypt-pbkdf

Giancarlo Razzolini

7:15 p.m.

Em 12-06-2015 05:15, Remi Gacogne escreveu:

...

I am not, but everything depends on your threat model. If you are targeted via an "evil-maid", or a cold-boot attack, FDE may be doomed. Which is why I use secureboot + TPM + this: https://aur.archlinux.org/packages/mkinitcpio-chkcryptoboot/ and this: https://aur.archlinux.org/packages/chkboot/

And I have plans to port the tails memory erase to archlinux.

...

In addition to that, passphrase-protection on SSH keys has been weak for a long time, because a single MD5(IV || passphrase) is applied to generate the AES key used to encrypt the SSH key [1].

I'm aware of this, which is why I use Keepass + Keeagent, so not only my key is encrypted inside the keepass database, it also has a very long passphrase.

...

OpenSSL 6.5 introduced a new KDF [2] using bcrypt, enabled by default for ed25519 keys but not for RSA keys, so you may want to upgrade your keys to use the new KDF manually.

I rotate my keys at least twice a year. And now that keeagent supports ed25519 keys, I probably will rotate more often. Now, for the AUR, if it's a simple implementation, then I don't see why not do it. Just I don't see much benefit in compartimentalizing your ssh keys too much. Perhaps of for work and one personal will do. Cheers, Giancarlo Razzolini. Cheers,

Yuki Chiba

11 Jun 11 Jun

11:58 p.m.

Lukas Fleischer writes:

...

On Thu, 11 Jun 2015 at 21:18:11, Yuki Chiba wrote:

...
Hi,

I prefer to use individual ssh public keys on different machines, but I could register only single key in my profile on aur4.archlinux.org.

Just out of curiosity: Is there any reason for using different SSH keys per machine (for the same server)? Or is it really just a preference?

I want to exclude the possibility that my secret keys are leaked while I try to transfer them to different machines by some ways (e.g. a use of USB sticks). I know that it rarely happen, but I want to avoid the dangerousness which can be avoided.

...

...
Is there any chance to support multiple ssh public keys like github[1] or Bitbucket[2]?

That would not be too hard to implement. I might add it before the final 4.0.0 release.

I'm very glad if my request is implemented on AUR4. Thank you for your correspondence. Best regards, Yuki Chiba -- Yuki Chiba yuki.from.akita@gmail.com

3245

Age (days ago)

3246

Last active (days ago)

List overview

Download

9 comments

6 participants

participants (6)

David Kaylor
Giancarlo Razzolini
Lukas Fleischer
Remi Gacogne
Yuki Chiba
Øyvind Heggstad