this post was submitted on 06 May 2024
34 points (92.5% liked)

Linux

48331 readers
659 users here now

From Wikipedia, the free encyclopedia

Linux is a family of open source Unix-like operating systems based on the Linux kernel, an operating system kernel first released on September 17, 1991 by Linus Torvalds. Linux is typically packaged in a Linux distribution (or distro for short).

Distributions include the Linux kernel and supporting system software and libraries, many of which are provided by the GNU Project. Many Linux distributions use the word "Linux" in their name, but the Free Software Foundation uses the name GNU/Linux to emphasize the importance of GNU software, causing some controversy.

Rules

Related Communities

Community icon by Alpár-Etele Méder, licensed under CC BY 3.0

founded 5 years ago
MODERATORS
 

I was reading GitLab's documentation (see link) on how to write to a repository from within the CI pipeline and noticed something: The described Docker executor is able to authenticate e.g. against the Git repository with only a private SSH key, being told absolutely nothing about the user's name it is associated with.
If I'm correct, that would mean that technically, I could authenticate to an SSH server without supplying my name if I use a private key?

I know that when I don't supply a user explicitly like ssh user@server or via .ssh/config, the active environment's user is used automatically, that's not what I'm asking.

The public key contains a user name/email address string, I'm aware, is the same information also encoded into the private key as well? If yes, I don't see the need to hand that info to an SSH call. If no, how does the SSH server know which public key it's supposed to use to challenge my private key ownership? It would have to iterate over all saved keys, which sounds rather inefficient to me and potentially unsafe (timing attacks etc.).

I hope I'm somewhat clear, for some reason I find it really hard to phrase this question.

top 23 comments
sorted by: hot top controversial new old
[–] [email protected] 39 points 6 months ago (1 children)

There are instances where the user is implied, but there is always a user. As far as Git goes, the user is almost always git.

[–] [email protected] 4 points 6 months ago (1 children)

I guess this is probably the solution to my riddle. Thanks.

[–] [email protected] 1 points 6 months ago

Yeah either that or they use a custom SSH implementation that just ignores the username because it's not needed for the type of authentication they're doing (like checking the keys of a specific account/project that is already known).

[–] [email protected] 30 points 6 months ago (1 children)

The public key contains a user name/email address string

No it does not. That is just a comment field.

[–] [email protected] 6 points 6 months ago

Thanks for pointing that out.

[–] [email protected] 18 points 6 months ago (1 children)

Technically, you always use a username, however in case of Gitlab that SSH username is always git. When an SSH client connects to server, it offers an authentication method. Gitlab accepts that method only if it is a publickey and the fingerprint of the offered key maps to the known Gitlab user.

[–] [email protected] 4 points 6 months ago (3 children)

It's a blessing and a curse. I have two gitlab accounts on the same server - private and work. I can't use the same key for both as the key is used to distinguish git users, and git doesn't make it easy to select which key you want to use to pull or clone particular repo.

[–] [email protected] 8 points 6 months ago (1 children)

git config core.sshCommand 'ssh -i <path to desired key>'

[–] [email protected] 3 points 6 months ago (1 children)

Hmm, with a similar technique one could even create git command aliases for running git with specific ssh private keys

[–] [email protected] 5 points 6 months ago

Better than that, git config supports conditional includes, based on a repo URL or path on disk. So you can have a gitconfig per organization or whatever, which specifies an sshCommand and thus an ssh key.

[–] [email protected] 5 points 6 months ago (1 children)

Just a thought: why don't you just use two different aliases for the Server in your .ssh/config with your two differing ssh keys, that way you can just use two different "hostnames" that have different ssh keys specified

[–] [email protected] 6 points 6 months ago (1 children)

I could but then i can't simply copy the "clone" url from the projects page without editing it.

I think the best solution for me may be to have a separate user account in the OS for work and personal projects. That way i can just log off at the end of the day and use personal account for personal stuff.

[–] [email protected] 2 points 6 months ago

You might not even have to log out: Just change the user in the terminal: su - user2

[–] [email protected] 4 points 6 months ago

You can tell git to use a specific key for each repo. I have the same situation as you and this is how I handle it.

https://superuser.com/questions/232373/how-to-tell-git-which-private-key-to-use

[–] [email protected] 8 points 6 months ago* (last edited 6 months ago)

Public SSH keys don't contain any user information at all. They could have some metadata for users to easily read, but that can be deleted without repercussions.

I'm no expert, and this is probably how it does NOT work, but if you have a private key, it can generate the public key, so that could be a way to tell the server "this is me", now try me.

Hey I wasn't so far off!

[–] [email protected] 5 points 6 months ago* (last edited 6 months ago)

EDIT: Noticed you're talking about Gitlab in the question, and I responded about Github, but I'm certain that gitlab does everything the same way, because that's all the technology is capable of. (I have no way to test the ssh -T command at the end for gitlab, though, so ymmv.)

To clear up some minor confusion here:

  1. Github knows nothing about your private key. There's very little metadata stored in the private key, and github.com has access to none of it. That includes email address or identity.
  2. Github has identity information stored for you, and then, separately, you uploaded a public key. The public key also contains no information about you, but github knows it's part of your account. Additionally, github enforces a requirement that your public key can't be uploaded to any other account, for the reason I'm about to state below.
  3. Github has an index built of everyone's public keys (or more likely their digests, although the technical details of the index are not something known to me--and it doesn't matter). When it sees an authentication request, it looks up the digest in the index, which maps to a user account.

At this point it already knows who is trying to authenticate. Once your authentication request succeeds with your public key (the usual challenge-response handshake associated with asymmetric cryptography), github interacts with your ssh client (most likely git) applying the permissions of your user and your user account.

BTW, github has a documented method for testing the handshake without doing any git operations:

ssh -T [email protected]

Depending on your ssh config, you might also need to supply -i some_filename.pem to this. Github will reply with

Hi aarkon! You've successfully authenticated, but GitHub does not provide shell access.

and then close the connection.

Note that the test authentication uses the username git and, again, contains no information about who you are. It's all just looked up on github's side.

[–] [email protected] 1 points 6 months ago

It would have to iterate over all saved keys, which sounds rather inefficient to me and potentially unsafe (timing attacks etc.)

sshd only checks for matches in the user's authorized_keys file, not system wide.

[–] [email protected] 1 points 6 months ago (1 children)

When authenticating with git over SSH, the private key should be considered secret and well protected.

That means the corresponding public key that was uploaded to the git server is enough to authenticate and no username is required. However, a password protected privare key is possible and extra layers of security can be added to the authentication mechanism.

As far as resource based attacks based on public key searching, I doubt many servers have significant enough public keys on a single host to even notice. Most repos are siloed and have specific teams/individuals assigned to them, so only a small number of public keys even gets loaded.

I dont know enough about the server side mechanics to be sure, but imo the attack surface is pretty small.

[–] [email protected] 1 points 6 months ago (1 children)

That means the corresponding public key that was uploaded to the git server is enough to authenticate and no username is required.

A username is required to authenticate with an SSH server. A public key alone is not enough.

[–] [email protected] 2 points 6 months ago (1 children)

While true, in most cases the username is simply "git" and not a specific username tied to the pub/priv keypair

[–] [email protected] 1 points 6 months ago

The post you originally replied to was misunderstanding how the username is located when authenticating with a server.

Original post:

The public key contains a user name/email address string, I’m aware, is the same information also encoded into the private key as well?

Your reply would be creating more confusion, because you implied that no username is required.

Your reply:

That means the corresponding public key that was uploaded to the git server is enough to authenticate and no username is required.

I am just clarifying if the original poster read your comment and was led to believe they wouldn't need a username. It is, in fact, required. As you expressed, it's usually "git" when connecting to a a git server, but it doesn't have to be.

[–] [email protected] 1 points 6 months ago

If I'm correct, that would mean that technically, I could authenticate to an SSH server without supplying my name if I use a private key?

Yes.

The public key contains a user name/email address string, I'm aware, is the same information also encoded into the private key as well? If yes, I don't see the need to hand that info to an SSH call. If no, how does the SSH server know which public key it's supposed to use to challenge my private key ownership?

Most of this can be found reading through different Git docs, whether from GitHub, GitLab, Codeberg, Gitea, etc. When using Git you can use different keys for different repos/forges and each has a defined pair, similar to accessing different SSH servers that require specific key pairs. I do understand your questions, but I lack the finesse to explain it since I really only use SSH and Git for my blog and not for anything too complicated.

[–] [email protected] 0 points 6 months ago

I only have experience with Bitbucket, and absolutely none if this may be applicable to you, but we have to generate a key with certain parameters (a minimum) for them to work, and the public key has to be input into a field on your account. So while you do not need to "provide" a username to perform git commands, it is set up in your account as your private key. The command to gen the key is: ssh-keygen -t rsa -m PEM -b 4096 -C "[email protected]"

Once you put your public key into your bitbucket account, using that key will mark all changes you make to you. Is this what you are talking about or am I just off in left field?