In Chapter 1 a procedure was given to validate your correspondents' public keys: a correspondent's key is validated by personally checking his key's fingerprint and then signing his public key with your private key. By personally checking the fingerprint you can be sure that the key really does belong to him, and since you have signed they key, you can be sure to detect any tampering with it in the future. Unfortunately, this procedure is awkward when either you must validate a large number of keys or communicate with people whom you do not know personally.
GnuPG addresses this problem with a mechanism popularly known as the web of trust. In the web of trust model, responsibility for validating public keys is delegated to people you trust. For example, suppose
Alice has signed Blake's key, and
Blake has signed Chloe's key and Dharma's key.
In practice trust is subjective. For example, Blake's key is valid to Alice since she signed it, but she may not trust Blake to properly validate keys that he signs. In that case, she would not take Chloe's and Dharma's key as valid based on Blake's signatures alone. The web of trust model accounts for this by associating with each public key on your keyring an indication of how much you trust the key's owner. There are four trust levels.
Nothing is known about the owner's judgement in key signing. Keys on your public keyring that you do not own initially have this trust level.
The owner is known to improperly sign other keys.
The owner understands the implications of key signing and properly validates keys before signing them.
The owner has an excellent understanding of key signing, and his signature on a key would be as good as your own.
The GnuPG key editor may be used to adjust your trust in a key's owner. The command is trust. In this example Alice edits her trust in Blake and then updates the trust database to recompute which keys are valid based on her new trust in Blake.
alice% gpg --edit-key blake pub 1024D/8B927C8A created: 1999-07-02 expires: never trust: q/f sub 1024g/C19EA233 created: 1999-07-02 expires: never (1) Blake (Executioner) <blake@cyb.org> Command> trust pub 1024D/8B927C8A created: 1999-07-02 expires: never trust: q/f sub 1024g/C19EA233 created: 1999-07-02 expires: never (1) Blake (Executioner) <blake@cyb.org> Please decide how far you trust this user to correctly verify other users' keys (by looking at passports, checking fingerprints from different sources...)? 1 = Don't know 2 = I do NOT trust 3 = I trust marginally 4 = I trust fully s = please show me more information m = back to the main menu Your decision? 3 pub 1024D/8B927C8A created: 1999-07-02 expires: never trust: m/f sub 1024g/C19EA233 created: 1999-07-02 expires: never (1) Blake (Executioner) <blake@cyb.org> Command> quit [...] |
The web of trust allows a more elaborate algorithm to be used to validate a key. Formerly, a key was considered valid only if you signed it personally. A more flexible algorithm can now be used: a key K is considered valid if it meets two conditions:
it is signed by enough valid keys, meaning
you have signed it personally,
it has been signed by one fully trusted key, or
it has been signed by three marginally trusted keys; and
the path of signed keys leading from K back to your own key is five steps or shorter.
Figure 3-1 shows a web of trust rooted at Alice. The graph illustrates who has signed who's keys. The table shows which keys Alice considers valid based on her trust in the other members of the web. This example assumes that two marginally-trusted keys or one fully-trusted key is needed to validate another key. The maximum path length is three.
When computing valid keys in the example, Blake and Dharma's are always considered fully valid since they were signed directly by Alice. The validity of the other keys depends on trust. In the first case, Dharma is trusted fully, which implies that Chloe's and Francis's keys will be considered valid. In the second example, Blake and Dharma are trusted marginally. Since two marginally trusted keys are needed to fully validate a key, Chloe's key will be considered fully valid, but Francis's key will be considered only marginally valid. In the case where Chloe and Dharma are marginally trusted, Chloe's key will be marginally valid since Dharma's key is fully valid. Francis's key, however, will also be considered marginally valid since only a fully valid key can be used to validate other keys, and Dharma's key is the only fully valid key that has been used to sign Francis's key. When marginal trust in Blake is added, Chloe's key becomes fully valid and can then be used to fully validate Francis's key and marginally validate Elena's key. Lastly, when Blake, Chloe, and Elena are fully trusted, this is still insufficient to validate Geoff's key since the maximum certification path is three, but the path length from Geoff back to Alice is four.
The web of trust model is a flexible approach to the problem of safe public key exchange. It permits you to tune GnuPG to reflect how you use it. At one extreme you may insist on multiple, short paths from your key to another key K in order to trust it. On the other hand, you may be satisfied with longer paths and perhaps as little as one path from your key to the other key K. Requiring multiple, short paths is a strong guarantee that K belongs to whom your think it does. The price, of course, is that it is more difficult to validate keys since you must personally sign more keys than if you accepted fewer and longer paths.
Figure 3-1. A hypothetical web of trust
trust | validity | ||
---|---|---|---|
marginal | full | marginal | full |
Dharma | Blake, Chloe, Dharma, Francis | ||
Blake, Dharma | Francis | Blake, Chloe, Dharma | |
Chloe, Dharma | Chloe, Francis | Blake, Dharma | |
Blake, Chloe, Dharma | Elena | Blake, Chloe, Dharma, Francis | |
Blake, Chloe, Elena | Blake, Chloe, Elena, Francis |
[1] | GnuPG overloads the word ``trust'' by using it to mean trust in an owner and trust in a key. This can be confusing. Sometimes trust in an owner is referred to as owner-trust to distinguish it from trust in a key. Throughout this manual, however, ``trust'' is used to mean trust in a key's owner, and ``validity'' is used to mean trust that a key belongs to the human associated with the key ID. |