DeveloperWiki:Package Signing Proposal for Pacman
This is a proposal for the package signing feature for Pacman. Here we'll gather ideas and commitments, so the implementation will be guided by this document.
- 1 Introduction
- 2 Web of Trust - Simple introduction
- 3 Pacman's keyring
- 4 Arch Key Signing keys
- 5 Package signing by developers
- 6 Installation of KSK by the users
- 7 Package verification
- 8 Affected tools
- 9 Final comments
Package signing is a long asked feature for Pacman. The goal of this implementation is to guarantee that a package was created by the person that claims to have made it.
For that to work, we'll use GnuPG as the tool to sign, verify and manage the group of keys that are trusted.
Web of Trust - Simple introduction
A web of trust is the concept used by OpenPGP (and GnuPG) for the management of trust policies in its public key system. As there's no concept of a central authoritative entity, there's the need of some kind of verification of the public keys that we accept in our keyring. Keys can be signed by other users, indicating that they trust in the veracity of the key. But the verification can become cumbersome, as the number of keys exchanged increases. To solve this, the comcept of Web of Trust was introduced. If I trust enough in a friend, I can sign his key in my keyring and tell to gpg that I also trust in the keys that he signs. So, if I verify a file that was signed by a person that my friend trusts, gpg will accept the signature as valid. When I import the third person's public key, gpg will mark it as a valid key.
In GnuPG, there'are four levels of trust in a public key:
- I didn't said anything about this key yet
- i don't trust this public key to sign other keys (this just affects the signature keys, not the signatures of files)
- I have a little trust in this public key to sign other keys
- I trust completely in this public key, as if it was my own
So, GnuPG can be configured to accept a key as valid if it has 3 marginaly trusted key signatures or 1 fully trusted key signature (this is the default). Or it can be any other combination, if properly configured.
The following is a suggestion for the use of GnuPG to sign and manage the pacman's keyring, together with the creation of a web of trust for the Arch Developers. It is very similar to what Debian and Fedora do, although there are some differences, because of our way of doing things.
Pacman will have a separated keyring, so the root's keyring will not be affected by keys that are intended for use with package signing. This is already implemented in Allan's pacman git branch for gpg support. The directory will be /etc/pacman.d/gnupg/. There will be the public key database, the trust database and a fake private key database, because GnuPG doesn't work very well without one (according to Debian's apt-key script).
The keyring will be populated based on the keys from the pacman-keyring package. In this package, there will be a file for the valid keys and one for the removed keys, so that the post-install script can revoke keys that may not be trusted anymore (be it for security reasons or because some developer has left the project). There'll be signature files for the two sets of keys, so the updatedb script can check to see if they are valid before updating.
Arch Key Signing keys
There will be 3 keys for the sole purpose of signing other developers' keys (hereafter named KSK, Key Signing Keys). They will be created by 3 developers (hereafter named Key Signers) that will be responsible for the role of signing the other developer's keys (and theirs own too). This procedure is very important and must be done with the certainty that the keys being signed are from the person they plead to be. The confirmation of the fingerprints must be done via a secure channel (skype, phone call, secure email or personally). For example, Debian only trust in keys confirmed personally. We can be a little lenient because the group of developers is reduced.
The Key Signers must keep his KSK secured and must choose a strong passphrase, so that even in the event of the secret keys being stolen, the risk of a real misuse will be little. In such case, there will be some time for the generation of a new set of KSK and the re-signing of the others developers' keys. The Key Signers must also generate a revocation key for each of the KSK and must keep the revocation keys secured (preferably only in printed form or in a thumbdrive stored in some kind of safe box). This is needed because anyone that owns the revocation key can revoke the corresponding key.
To sign the developer's keys, the Key Signers must receive a copy of the public key (through email or from a public key server). After importing the key into his own keyring, the Key Signer must sign the public key with the KSK. After that, he needs to export the signed public key and update the pacman-keyring package (there'll be a script for that task). Only one Key Signer should update the pacman-keyring package at a time to avoid having some key being lost because of overwriting files. There must be some form of coordination between the Key Signers on who will do the task.
With the KSK and the signed developers' keys, the pacman-keyring package will be created and signed with one of the KSK.
Use case: Arch Key Signing Keys creation
- 3 developers will be selected to be the Key Signers
- Each Key Signer creates a public/private key pair in his local keyring with gpg --gen-key
- We need to decide if there'll be an expiration date
- Each Key Signer creates a revocation key, which will be used for revocation of the KSK, in case of compromising of the private key. gpg --gen-revoke <key id>
- GnuPG can generate a revocation key for other person to trigger there the revocation. So, for each KSK, there should be designated revocation keys for the other two Key Signers. They could revoke their KSK and the others too, in case of inavailability of the owner of a KSK.
- The revocation key MUST be kept secure, even more than the private key. This is because any person can revoke a key. The advise from GnuPG is that the revocation key be kept in a thumb drive in a safe or printed, so that there would be no way to access the digital file of the rev. key.
- Each Key Signer sends his public KSK to a predefined key server. gpg --send-keys <key id> --keyserver <url for key server>
- Each Key Signer fetches the other KSKs from the key server and signs them with his own KSK (not his personal key)
- Each Key Signer sends the other KSKs to the server again, to update the signatures.
- Each fetches again the KSKs from the key server, so they get the KSKs with all signatures from the others signers.
Use case: Signing of developer's keys
- If a developer doesn't have a key yet, he generates one with: gpg --gen-key
- If he generated the key in the previous step, he needs to generate a revocation key too. gpg --gen-revoke <key id>
- The same remarks for the revocation key above apply here.
- The developer exports his public key to a file or sends it to a public key server. gpg --export <key id> > key.gpg or gpg --send-key <key id> --keyserver <url for key server>
- The developer contacts one of the Key Signers and sends him his public key file or instructs him to fetch it from the key server. Is important to negotiate a secure channel for the exchange of fingerprint. It should not be through email. Maybe IM, Skype or through ssh session with some Arch server as temporary storage.
- The Key Signer signs the developer's public key with his owned KSK.
- There'll be a script to help the management of the files used to populate pacman's key ring, so the new key will be added to the set of valid keys.
- The pacman-keyring package will be rebuilt and signed with a KSK (not with a common developer's key). The procedure is described in the use case for package signing.
Package signing by developers
When a developer builds a new package, makepkg will have the options to sign the package too, with the developer's own key (not the KSK, if the developer owns one). At this point, there are three options (that we should choose now) for the format of the signed/signature pair:
- detached signature external to the package
- the package will stay unchanged and there'll be a new file for the signature.
- detached signature internal to the package
- makepkg would generate a detached signature, but would tar the package and the signature into a new file, so that both are always toghether (Debian and RPM based distros do that way). This would have a bigger impact on all developer tools and pacman itself.
- attached signature
- the signature would contain the signed file, and pgp would be used to extract the signed file. Just like the one above, this would require lots of changes on the tools.
The cheaper approach is obviously the first option. It will not require lots of changes, but there'll be some. Maybe the convenience of the latter two would compensate for the trouble of changing the tools? This needs to be decided upon.
- A package is rebuilt and signed with makepkg --sign-key <key id> [other options]
- Developer uses tool to upload the package and signature to his staging area
- Developer repeats this until there are no more packages to build
- Developer uses tool to copy his packages from staging area to the final repository and sign the repo.db. The script will do the following:
- check if the repo.db is locked. Just proceed if unlocked.
- lock the repo.db
- call repo-add to add all the new packages
- generate a sha1 hash of the repo.db, on the server
- scp the hash from the server to the local machine of the developer
- sign the local hash with the developer's personal key. The signature will be an attached and ascii armored one, to simplify the handling latter.
- scp the signature back to the server
- unlock the repo.db
Installation of KSK by the users
This is a very sensitive part and must be done with caution, at the risk of driving all the work moot. The KSK should be manually verified by the user before pacman can accept the pacman-keyring package. I believe that the following would be the workflow:
- the KSK and the corresponding fingerprints would be available in several channels of Arch: Home page, git repository, forums, public key servers, etc.
- the user downloads the KSK and import them (with a new tool, called pacman-key, named after apt-key)
- pacman-key shows the fingerprint of each KSK and asks for the approval of the user
- if the user confirms, pacman imports the keys to its keyring, setting them as fully trusted
After this, the pacman-keyring can be installed and the public keys of the developers can be added by the post-install script. The trustdb will be updated automatically.
- The user must download the pacman-key tool (or it could be provided in a version of pacman without signing capabilities).
- User fetches the KSKs from a key server. The id's can be discovered on a new item or forum or mail list post. pacman-key --fetch <KSK1 id> <KSK2 id> <KSK3 id>
- If the user downloaded the KSKs, they must be imported to pacman's keyring. Otherwise, they are already there. pacman-key --import <file with KSKs>
- User must trust explicitly each KSK with pacman-key trust <KSK id>
- The KSK fingerprint will be shwon to the user. He must guarantee that he consults more than one secure source for comparison, such as Arch news item, Forum post, Mail list archive post, maybe we should have a file in git too.
- The user must type 'trust' in the prompt. GnuPG will as the level of trust that the user assigns tho the key. He should answer 5 (ultimately) for the KSKs. Any value other than 5 will not make the KSK effective to assign trust to other keys signed by it.
- The user types 'save' to end the edit session of GnuPG
- When a version of pacman able to verify signatures is available, pacman will update itself and install pacman-keyring first, as a dependency. The pacman-key tool will be used by --post-install script from pacman-keyring and will update the real pacman keyring with the keys in this package. The web of trust will be updated accordingly.
When pacman downloads a package, the signature will be downloaded together, if applicable. Pacman will inform the user if the package's signature is not valid and stop. There'll no possibility to install a signed package with an invalid signature.
- Pacman downloads the repo.db and its corresponding signature. The following should happen:
- pacman verifies if the signature date is older than n days (we must decide this). If older, stop the procedure and ask the user to change mirrors or report the bug (the lack of new versions of the repo.db will be a bug)
- pacman verifies the signature to see if it is valid.
- pacman computes the sha1 hash from the repo.db file locally.
- pacman compares the hash with the one in the signature. The hash on the signature is readable because the signature is ascii armored to easy this step. If the hashes don't match, stop the process.
- pacman downloads the packages and its signatures
- Before the installation (or update), each package is verified to see if the signature is correct and is signed by a trusted key. If the package has a signature and it is not valid, it should not be installed. One weak point is that the user could delete the signature file and we would lose track that the package was signed. Maybe we could indicate this in the repo.db, so that the user will not be able to fool pacman.
There should be options to choose the key to sign a package. The key will be always from the keyring of the user building the package.
This is a subject that is being discussed and this text will be updated soon.
Allan's branch is already verifying signatures, but there'are some easy changes needed to point to the right keyring.
Pacman-key is a new tool, responsible for the management of Pacman's keyring. It is a shell script and is needed because gpgme is not able to handle some operations on keyrings, so we can't change pacman to handle them too. The operations available will be, amongst others:
- approval of KSK
- importing/exporting of keys
- fetching keys from a key server
- changing level of trust for keys
- removal of keys
- update of trustdb
We believe that this suggestions are feasible and will bring a new level of quality to Arch Linux. The gpg branch of pacman git repository of Allan is in a good position in relation of what I suggested above.
The discussions will take place on pacman-dev mailing list and this document will be updated as the decisions are made.