This is a draft document for public discussion. It reflects the personal opinions of the author, and does not necessarily represent the views of mozilla.org staff and the Mozilla Foundation.
Please post comments and questions to the netscape.public.mozilla.crypto newsgroup or the corresponding mozilla-crypto mailing list, or send them to the document author, Frank Hecker.
This section of the Mozilla Certificate FAQ is intended primarily for general Mozilla users who may be unfamiliar with certificates and CAs; it provides general background information important for understanding how Mozilla and related software (e.g., Firefox, Thunderbird, Camino, etc.) uses certificates.
Certificates are used in the context of "public key" cryptography. More specifically, certificates are digitally signed data items combining a public key used for some purpose (e.g., enabling a web server to accept SSL connections) with information about the entity associated with the public key. Certificates are used in at least three functions within Mozilla and related software:
In public key cryptography the communicating entities (e.g., web servers or user email programs) each have a unique pair of cryptographic keys, a "private" key that is kept secret and a "public" key that is made known to others. In the various cryptography-based functions in Mozilla mentioned above public keys are not used in their "raw" form but rather are encapsulated in "certificates". (These are often referred to as "X.509 v3 certificates", from the name of the technical standard defining the certificates' format.)
As noted above, a certificate is a digitally signed bundle of data that includes both the public key for a given entity and various pieces of information about that entity; for example, the certificate for a secure web server includes the domain name of the server, a certificate used for secure email includes the email address of the sending user, and the certificate for a signed Java applet includes the name of the organization or individual who developed and/or distributed the applet.
(The following information is more technical in nature, and can be skipped unless you're interested in knowing more about how certificates actually work.)
Digitally signing a certificate (or any other data) is done using some entity's private key. More specifically, certificate data are signed by taking the bit string representing the data and putting it through a specially-designed "hashing" operation that generates a small fixed-length bit string, and then encrypting that new bit string (the "hash") using some entity's private key to generate the "signature". Hash functions are mathematically designed to ensure that different bit strings will generate different hash values, and public/private key pairs and algorithms are designed to ensure that data encrypted by a private key can be easily decrypted only by the corresponding public key. (In both cases these guarantees are not absolute, but are as good as skilled cryptographers can make them.)
Someone receiving a copy of the data and its signature can verify the data and signature by decrypting the signature using the entity's public key to recover the original hash value, separately hashing the data to generate a new hash value, and then comparing the original hash value and the new hash value. If the recovered hash value and the newly-generated hash value match then the receiver can be reasonably sure that the data received are the same as the data as originally signed, and that the entity whose public key was used to verify the signature is the same entity that did the signing (with the corresponding private key).
Briefly, a Certification Authority or CA is an entity that digitally signs other entities' certificates. (A Certification Authority is also sometimes referred to as a Certificate Authority.)
As mentioned previously, a certificate contains an entity's public key, and is digitally signed using some entity's private key. A "self-signed" certificate is one in which the private key used to digitally sign the certificate is the private key corresponding to the public key in the certificate itself; in other words, the entity signing the certificate is the same entity whose public key is in the certificate (hence the term "self-signed").
Verifying the digital signature on a self-signed certificate can then be done using the public key in the certificate itself, as described above. If the signature is valid then we can be reasonably sure that the public key in the certificate has not been corrupted in any way, and that any other data found in the certificate is as originally put there by the entity associated with the certificate and public key.
It is also possible for a certificate to be signed using some other private key belonging to a third party (i.e., an entity other than the one whose certificate it is). In this case verifying the information in the certificate requires having the public key for that third party. The third party's public key can itself be distributed in the form of a certificate, and that certificate can in turn by signed either by the third party itself (as a self-signed certificate) or by some other party entirely.
In the scheme used by Mozilla, certificates for entities such as web servers, email users, or code developers are typically not self-signed but rather are signed by third parties (organizations or individuals) known as "Certification Authorities" or "CAs" for short. (CAs are in turn considered to be part of what is commonly referred to as a Public Key Infrastructure or PKI — although in fact it is possible to have a PKI that does not use CAs.) By signing the data in certificates CAs are assumed to be in some way vouching for the information contained in the certificate data.
For example, a certificate used for a secure web server normally contains the domain name used to connect to the web server, and by signing such a certificate a CA is assumed to be vouching for the fact that the entity operating the web server (the entity that controls the server's private key corresponding to the public key in the certificate) actually controls the domain name associated with the server. Similarly, a certificate used for secure email should contain the email address of the person or organization that controls the corresponding email account, with the CA signing the certificate assumed to have verified that that is the case, and a certificate used for a digitally signed applet (or other executable code) should contain the name of the developer or distributor of the applet (again, assumed to be verified by the CA).
Verifying a typical web server, email user, or developer certificate then requires having the public key for the CA that signed the certificate. The CA's public key is itself distributed in the form of a certificate; this "CA certificate" is in turn digitally signed either by some other CA or by the CA itself (as a self-signed certificate). In the former case the CA is referred to as an "intermediate" CA; in the latter case the CA is referred to as a "root" CA, and its certificate is a "root CA certificate".
In general it is possible to have multiple root CAs; each root CA can then "issue" certificates directly for web servers, email users, developers, etc., by digitally signing the data in those certificates, or can issue certificates to one or more intermediate CAs, which then issue certificates in turn.
Note that in theory CAs are not necessary in order to support cryptography-based functions like secure web browsing, etc., and in fact there are systems like PGP that do not use CAs in the sense defined above. (PGP uses a separate "web of trust" system in which PGP users sign each others' keys.) However the main cryptography-based functions in Mozilla — secure web browsing, secure email, and digitally signed code objects — do assume the use of CAs, including root CAs.
(Mozilla's built-in secure email function is based on the S/MIME standard; at this time Mozilla does not natively support PGP-based email except through the externally-developed Enigmail plugin.)
Mozilla as distributed includes various CA certificates by default, in order to reduce the amount of configuration users have to do before they can use Mozilla for these cryptographic-based functions.
As discussed in the answer to the previous question, in order to verify a certificate for a web server, email user, or code developer, Mozilla must thus have the certificate for the CA that issued (i.e., digitally signed) the certificate being verified. If the CA is an intermediate CA then Mozilla must also have the certificate for the CA that issued the intermediate CA's certificate, in order to verify that certificate as well. This other CA may be a root CA or yet another intermediate CA; in the latter case yet another CA will be involved, and so on.
Mozilla continues verifying certificates until it comes to a point where it needs a root CA certificate, corresponding to the root CA that issued the original web server, etc., certificate or that issued an intermediate CA's certificate. Since root CA certificates are self-signed, Mozilla can verify such a certificate using the public key in the root CA certificate itself, and if that verification completes successfully then the process is done.
In some cases Mozilla can get the CA certificate(s) from the same place and in the same way as it got the original certificate; for example, if a web server presents its own certificates to Mozilla then it could also present the needed CA certificate(s) as well, including the root CA certificate and any intermediate CA certificates. (Such a set of linked certificates is known as a "certificate chain".)
However it is also convenient for Mozilla to keep its own copies of certificates, including root CA certificates in particular. Among other things, Mozilla cam mark a given root CA certificate as being valid for verifying certain types of certificates, and as not being valid to verify other types of certificates.
For example, a particular root CA may issue certificates only for web servers, not for email users or code developers; in the Mozilla certificate database this root CA's certificate could be marked as being valid only for verifying web server certificates. If Mozilla receives a email user certificate issued by this root CA (or by an intermediate CA under the root CA) it would then raise an error condition and alert the user; on the other hand web server certificates issued by the root CA (or an intermediate CA under it) would be verified by Mozilla without error and with no need for user intervention.
This process of marking root CA certificates as being valid for verifying certain types of certificates is commonly known as "trusting" the root CA, and the special flags associated with each root CA certificate are known as "trust bits".
If Mozilla or related software did not already have a copy of a given root CA certificate then it would be unable to automatically determine whether certificates issued by that root CA (or subordinate CAs) should be accepted or not, and would have to prompt the user as to what to do. Most users don't know what CAs are or don't possess the necessary information to properly decide what Mozilla should do. To prevent these typical Mozilla users from having to deal with this issue, Mozilla and related software includes a pre-loaded set of default root CA certificates, with the trust bits set appropriately.
These pre-loaded root CA certificates are distributed with Mozilla and related software in the form of a shared library installed on users' systems along with the rest of the software executable code. They can therefore be updated when new versions of the software are release.
(For anyone interested, the pre-loaded CA certificates are included in the following files:
libnssckbi.dlllibnssckbi.soContents/MacOS/libnssckbi.dynlibVersion 0.3, February 4, 2005. Updated and corrected.
Version 0.2, February 9, 2004. Further emphasized that this represents personal opinions only at this time.