53
  1. Am I correct calling file with .p7b file extension saved as 'Cryptographic Message Syntax Standard - PKCS#7 Certificates (.P7B)' in Windows - a 'PKCS#7 certificate'? Or is it better called 'X.509 certificate saved in PKCS#7 format'?

  2. When would one choose one certificate format over another? Do these formats have any particular strengths or weaknesses?

  3. Adding this question after my first two edits. How is PKCS#7 format different compared to DER/PEM file formats?

Thanks


Edit #1: Firefox under Linux offers me an ability to export some website's certificate as:

  • X.509 Certificate (PEM)
  • X.509 Certificate (DER)
  • X.509 Certificate (PKCS#7)

Does it mean that PKCS#7 here is just a binary file format similar to but distinct from DER? If true then .p7b file is just an X.509 certificate saved in PKCS#7 format (as opposed to PEM or DER formats).


Edit #2: Follow up to my first edit. This page OpenSSL: Documents, pkcs7 suggests that PKCS#7 can be encrypted in either DER or PEM. From that I deduce that PKCS#7 is not a distinct binary file format. Now I'm totally confused.


Edit #3: Ok, I figured the relationship between PEM and DER formats. The Base64 encoded payload of the PEM file is actually data in DER format. So initially the X.509 certificate is encoded in DER format and then optionally you can encode the resulted 'DER encoded certificate' to 'PEM encoded certificate'. I'm still having difficulties fitting PKCS#7 part of the puzzle.


Edit #4: Another piece of information. PKCS#7 seems to be a container that allows to bundle together several X.509 certificates prior to encode them into DER format (which is different from PEM format where you can bundle certificates together in the same file by just pasting them one after another).

  • 1
    Consider posting your edits as an answer to your own question. – CoverosGene Nov 19 '14 at 14:23
56

You've evolved to mostly right, but to add several points and expand on @CoverosGene' answer more than I felt comfortable doing in an edit:

X.509 defines a certificate (and some other things not relevant here) in ASN.1, a (very!) general data structuring method which has several defined encodings, of which DER Distinguished Encoding Representation is quite common and is used here.

PEM format -- for several types of data of which a certificate is only one -- is much as you say just binary (DER) data encoded in base64 (edit) broken into lines normally every 64 chars (but there are variations), plus header and trailer lines consisting of dashes + BEGIN or END + the type of data, in this case CERTIFICATE + dashes. Those lines look redundant to a human but they are expected and mostly required by software. PEM (Privacy Enhanced Mail) was actually a complete standard for secure email that has now been mostly forgotten (see below) except for its encoding format. (edit) As of 2015 there is RFC 7468 describing in detail most use of 'PEM' formats for modern crypto data.

PKCS#7 was defined by RSA (the company, not the algorithm) as a multi-purpose format for encrypted and/or signed data. It was turned over to IETF and evolved into CMS Cryptographic Message Syntax in RFC 2630, then RFC 3369, then RFC 3852, then RFC 5652, hence the wording of the Windows (inetopt) prompt. "PKCS#7" is often used to mean both the original RSA PKCS#7 and the IETF successor CMS, in the same way "SSL" is often used for both the original Netscape protocol and the IETF successor TLS Transport Level Security.

The .p7b or .p7c format is a special case of PKCS#7/CMS: a SignedData structure containing no "content" and zero SignerInfos, but one or more certificates (usually) and/or CRLs (rarely). Way back when this provided a standard way to handle (edit) the set of certificates needed to make up a chain (not necessarily in order).

PKCS#7/CMS is (are?) also ASN.1 and depending on circumstances can be either DER or BER, a closely-related encoding with some very minor differences that most DER decoders handle.

While PKCS#7/CMS like any DER or BER object can be PEM-formatted, I've not seen any implementation other than openssl (edit) it is rare for certs. (Java CertificateFactory can read PKCS7/CMS-certs-only from DER or PEM, but CertPath.getEncoded writes it only to DER.) In contrast both DER and PEM formats for a single cert are common.

PKCS#7/CMS is also used as the basis for S/MIME secure email (multiple rfcs starting from 5751). Basically PEM encoded PKCS#7 into ASCII text which email systems of the 1980s could easily handle, while S/MIME represents CMS as MIME entities which are encoded in several ways modern email systems can handle.

OpenSSL confused matters by implementing, in order: a pkcs7 command which handles the certs-CRLs-only case not full PKCS#7; a crl2pkcs7 command which actually handles CRLs and certs, but again not the rest of PKCS#7; a smime command which actually handles both S/MIME and PKCS#7/CMS for most cases of encrypted and/or signed messages; and a cms command which actually handles both S/MIME and PKCS#7/CMS for a more complete set of cases.

So I would describe the options as: a cert in PEM or DER format; a (single) cert in a PKCS#7 container or for short just p7, and mention PEM only in the rare case it applies; or a cert chain in PKCS#7 or p7. The semantic difference between a single cert and a cert chain is at least as important as the format difference between a cert by itself or in a container.

And this doesn't even reach the widespread confusion between a certificate by itself (for some other entity, most often a CA root or anchor) and the combination of privatekey PLUS certificate -- or usually chain -- that you use to prove your own identity, for example as an SSL/TLS server or when signing S/MIME email. That uses the originally-Microsoft PFX Personal Information Exchange format, or its standardized form PKCS#12 or "p12".

  • 6
    It may be worth mentioning that a PKCS#7/CMS SignedData structure used as a mere bag for certificates is unordered: you can put several certificates, but the format does not keep track of any ordering. Therefore, it can transport a certificate chain exactly as well as IKEA sells furniture. Decoders must still work out who signed who in the chain. – Thomas Pornin Nov 19 '14 at 18:14
  • @Thomas good point, edited. And the same for SignedData that does convey a signature (or several), but that's out of scope for this question. – dave_thompson_085 Nov 21 '14 at 11:48
9

PKCS #7 can be thought of as a format that allows multiple certificates to be bundled together, either DER- or PEM- encoded, and may include certificates and certificate revocation lists (CRLs).

Per RFC2315, PKCS#7 is

a general syntax for data that may have
cryptography applied to it, such as digital signatures and digital
envelopes. The syntax admits recursion, so that, for example, one
envelope can be nested inside another, or one party can sign some
previously enveloped digital data.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.