]> g10 Code GmbH

Germany wk@gnupg.org

sec Network Working Group Internet-Draft This specification describes a service to locate OpenPGP and LibrePGP keys by mail address using a Web service and the HTTPS protocol. It also provides a method for secure communication between the key owner and the mail provider to publish and revoke the public key.

Introduction This memo describes a method to associate OpenPGP keys with a mail address and how to look them up using a web service with a well-known URI. In addition a mail based protocol is given to allow a client to setup such an association and to maintain it.

Notational Conventions The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in .

Web Key Directory A major use case for OpenPGP is the encryption of mail. A common difficulty of sending encrypted mails to a new communication partner is to find the appropriate public key of the recipient. Unless an off-channel key exchange has been done, there are no easy ways to discover the required key. The common practice is to search the network of public key servers for a key matching the recipient's mail address. This practise bears the problem that the keyservers are not able to give a positive confirmation that a key actually belongs to the mail addresses given in the key. Further, there are often several keys matching a mail address and thus one needs to pick a key on good luck. This is clearly not a secure way to setup an end-to-end encryption. Even if the need for a trusted key for an initial mail message is relinquished, a non-authenticated key may be a wrong one and the actual recipient would receive a mail which she can't decrypt, due to the use of a wrong key. Methods to overcome this problem are

• sending an initial unencrypted message with the public key attached,
• using the OpenPGP DANE protocol to lookup the recipients key via the DNS.

The first method has the obvious problems of not even trying to encrypt the initial mail, an extra mail round-trip, and problems with unattended key discovery. The latter method works fine but requires that mail providers need to set up a separate DNS resolver to provide the key. The administration of a DNS zone is often not in the hands of small mail installations. Thus an update of the DNS resource records needs to be delegated to the ISP running the DNS service. Further, DNS lookups are not encrypted and missing all confidentially. Even if the participating MUAs are using STARTTLS to encrypt the mail exchange, a DNS lookup for the key unnecessarily identifies the local-part of the recipients mail address to any passive eavesdroppers. This memo specified a new method for key discovery using an encrypted https connection.

Key Discovery Although URIs are able to encode all kind of characters, straightforward implementations of a key directory may want to store the local-part of a mail address directly in the file system. This forbids the use of certain characters in the local-part. To allow for such an implementation method the URI uses an encoded form of the local-part which can be directly mapped to a file name. OpenPGP defines its User IDs, and thus the mail address, as UTF-8 strings. To help with the common pattern of using capitalized names (e.g. "Joe.Doe@example.org") for mail addresses, and under the premise that almost all MTAs treat the local-part case-insensitive and that the domain-part is required to be compared case-insensitive anyway, all upper-case ASCII characters in a User ID are mapped to lowercase. Non-ASCII characters are not changed. The so mapped local-part is hashed using the SHA-1 algorithm. The resulting 160 bit digest is encoded using the Z-Base-32 method as described in , section 5.1.6. The resulting string has a fixed length of 32 octets. There are two variants on how to form the request URI: The advanced and the direct method. Implementations MUST first try the advanced method. Only if an address for the required sub-domain does not exist, they SHOULD fall back to the direct method. A non-responding server does not mean that the fall back should be carried out. The advanced method requires that a sub-domain with the fixed name openpgpkey is created and queried. The URI is constructed by concatenating these items:

• The scheme https://,
• the string openpgpkey,
• the domain-part,
• the string /.well-known/openpgpkey/,
• the domain-part in lowercase,
• the string /hu/,
• the above constructed 32 octet string,
• the unchanged local-part as a parameter with name l using proper percent escaping.

An example for such an advanced method URI to lookup the key for Joe.Doe@Example.ORG is: (line has been wrapped for rendering purposes) The direct method requires no additional DNS entries and constructs the URI by concatenating these items:

• The scheme https://,
• the domain-part,
• the string /.well-known/openpgpkey/hu/,
• the above constructed 32 octet string,
• the unchanged local-part as a parameter with name l using proper percent escaping.

Example for a direct method URI: (line has been wrapped for rendering purposes) Sites which do not use the advanced method but employ wildcard DNS for their sub-domains MUST make sure that the openpgpkey sub-domain is not subject to the wildcarding. This can be done by inserting an empty TXT RR for this sub-domain. The HTTP GET method MUST return the binary representation of the OpenPGP key for the given mail address. The key needs to carry a User ID packet () with that mail address. Note that the key may be revoked or expired - it is up to the client to handle such conditions. To ease distribution of revoked keys, a server may return revoked keys in addition to a new key. The keys are returned by a single request as concatenated key blocks. The server MUST accept the HTTP HEAD method to allow a client to check for the existence of a key. The server SHOULD use "application/octet-stream" as the Content-Type for the data but clients SHOULD also accept any other Content-Type. The server SHOULD NOT return an ASCII armored version of the key. The server MUST serve a Policy Flags file as specified in , "Policy Flags". That file is even required if the Web Key Directory Update Protocol is not supported. The benefit of the advanced method is its greater flexibility in setting up the Web Key Directory in environments where more than one mail domain is hosted. DNS SRV resource records, as used in earlier specifications of this protocol, posed a problem for implementations which have only limited access to DNS resolvers. Note that the direct method allows to provide a Web Key Directory without any DNS changes.

Web Key Directory Update Protocol To put keys into the key directory a protocol to automate the task is desirable. The protocol defined here is entirely based on mail and the assumption that a mail provider can securely deliver mail to the INBOX of a user (e.g. an IMAP folder). Note that the same protocol may also be used for submitting keys for use with OpenPGP DANE. In the following sections the term "target key" denotes the to be published key, the term "submission key" the key associated with the submission-address of the mail provider. The string "WELLKNOWN" denotes the first part of an URI specific for a domain. In the examples the domain "example.org" is assumed, thus: (line has been wrapped for rendering purposes) or if the sub-domain openpgpkey does not exist (direct method): We assume that the user already created a key for her mail account alice@example.org. To install the key at her provider's Web Key Directory, she performs the following steps:

1. She retrieves a file which contains one line with the mail address used to submit the key to the mail provider. See , "Submission Address", for the syntax of that file. For a mail address at the domain "example.org" the URI of the file is WELLKNOWN/submission-address
2. She sends her key using SMTP (or any other transport mechanism) to the provider using the submission address and key format as specified by PGP/MIME.
3. The provider checks that the received key has a User ID which matches an account name of the provider.
4. The provider sends an encrypted message containing a nonce and the fingerprint of the key to the mail account of the user. Note that a similar scheme is used by the well known caff(1) tool to help with key signing parties.
5. A legitimate user will be able to decrypt the message because she created the key and is in charge of the private key. This step verifies that the submitted key has actually been created by the owner of the account.
6. The user sends the decrypted nonce back to the submission address as a confirmation that the private key is owned by her and that the provider may now publish the key. The confirmation mail to the provider MUST be encrypted using the provider's public key as retrieved using the key lookup protocol described above.
7. The provider receives the nonce, matches it with its database of pending confirmations and then publishes the key. Finally the provider sends a mail back to the user to notify her of the publication of her key.

The message data structures used for the above protocol are specified in the following sections.

The Submission Address The address of the submission file is The file consists of exactly one line, terminated by a LF, or the sequence of CR and LF, with the full mail address to be used for submission of a key to the mail provider. For example the content of the file may be

The Submission Mail The mail used to submit a key to the mail provider MUST comply to the PGP/MIME specification (, section 7), which states that the Content-Type must be "application/pgp-keys", there are no required or optional parameters, and the body part contains the ASCII-armored transferable Public Key Packets as defined in , section 11.1. The mail provider MUST publish a key capable of signing and encryption for the submission-address in the Web Key Directory or via DANE. The key to be published MUST be submitted using a PGP/MIME encrypted message (, section 4). The message MUST NOT be signed (because the authenticity of the signing key has not yet been confirmed). After decryption of the message at the mail provider a single "application/pgp-keys" part, as specified above, is expected.

The Confirmation Request The mail provider sends a confirmation mail in response to a received key publication request. The message MUST be sent from the submission-address of the mail provider to the mail address extracted from the target key. The message needs to be a PGP/MIME signed message using the submission key of the provider for the signature. The signed message MUST have two parts: The first part MUST have "text" as its Content-Type and can be used to explain the purpose of the mail. For example it may point to this specification and explain on how to manually perform the protocol. By evaluating the Content-Language header of the request the provider may send back a localized version. The second part MUST have a Content-Type of "application/vnd.gnupg.wkd" and carry an OpenPGP encrypted message in ASCII Armor format. If the protocol version is unknown or less than 5 the Content-Type "application/vnd.gnupg.wks" MUST be used for backward compatibility. The message MUST be encrypted to the target key and MUST NOT be signed. After decryption a text file in the Web Key data format must be yielded. That data format consists of name-value pairs with one name-value pair per LF or CR+LF terminated line. Empty lines are allowed and will be ignored by the receiver. A colon is used to terminate a name. In a confirmation request the following names MUST be send in the specified order:

• "type": The value must be "confirmation-request".
• "sender": This is the mailbox the user is expected to sent the confirmation response to. The value must match the mailbox part of the "From:" address of this request. Exactly one address MUST be given.
• "address": The value is the addr-spec part of the target key's mail address. The value SHOULD match the addr-spec part of the recipient's address. The value MUST be UTF-8 encoded as required for an OpenPGP User ID.
• "fingerprint": The value is the fingerprint of the target key. The fingerprint is given in uppercase hex encoding without any interleaving spaces.
• "nonce": The value is a string with a minimum length of 16 octets and a maximum length of 64 octets. The string must entirely be made up of random ASCII letters or digits. This nonce will be sent back to the mail provider as proof that the recipient is the legitimate owner of the target-key.

The receiver of that message is expected to verify the outer signature and disregard the entire message if it can't be verified or has not been signed by the key associated with the submission address. After the message has been verified the receiver decrypts the second part of the signed message, checks that the "fingerprint" matches the target key, checks that the "address" matches a User ID of the target key, and checks the other constrains of the request format. If any constraint is not asserted, or the fingerprint or User ID do not match the target key, or there is no pending publication requests (i.e. a mail recently sent to the submission address), the user MAY be notified about this fake confirmation attempt. In other cases the confirmation request is legitimate and the MUA shall silently send a response as described in the next section. The rationale for the outer signature used with this request is to allow early detection of spam mails. This can be done prior to the decryption step and avoids asking the user to enter a passphrase to perform the decryption for a non-legitimate message. The use of a simple encrypted attachment, instead of using PGP/MIME encryption, is to convey the Content-Type of that attachment in the clear and also to prevent automatic decryption of that attachment by PGP/MIME aware clients. The MUA may in fact detect this confirmation request and present a customized dialog for confirming that request.

The Confirmation Response A response to a confirmation request MUST only be send in the positive case; there is no negative confirmation response. A mail service provider is expected to cancel a pending key submission after a suitable time without a confirmation. The mail service provider SHOULD NOT retry the sending of a confirmation request after the first request has been send successfully. The user MUST send the confirmation response from her target mail address to the "from" address of the confirmation request. The message MUST be signed and encrypted using the PGP/MIME Combined format (, section 6.2). The signing key is the target key and the encryption key is the key associated with the provider's submission address. The server MAY verify the signature and silently discard the message if it is not found to be valid; The Content-Type used for the plaintext message MUST match the Content-Type of the request. The format is the same as described above for the Confirmation Request. The body must contain four name-value pairs in this order:

• "type": The value must be "confirmation-response".
• "sender": The value is the value of the "sender" parameter from the confirmation request.
• "address": The value is the value of the "address" parameter from the confirmation request.
• "nonce": The value is the value of the "nonce" parameter from the confirmation request.

The server MUST check the "nonce" against the one sent in the request and publish the key only if the nonce matches and has not been re-used.

Policy Flags For key generation and submission it is useful to tell the client about certain properties of the mail provider in advance. This can be done with a file at the URL A site supporting the Web Key Directory MUST serve this file; it is sufficient if that file has a zero length. Clients may use this file to check for Web Key Directory support. The file contains keywords and optionally values, one per line with each line terminated by a LF or the sequence of CR and LF. Empty lines and lines starting with a '#' character are considered comment lines. A keyword is made up of lowercase letters, digits, hyphens, or dots. An underscore is allowed as a name space delimiter; see the end of this section. The first character must be a letter. Keywords which are defined to require a value are directly followed by a colon and then after optional white space the value. Clients MUST use case-insensitive matching for the keyword. Currently defined keywords are:

• "mailbox-only": The mail server provider does only accept keys with only a mailbox in the User ID. In particular User IDs with a real name in addition to the mailbox will be rejected as invalid.
• "dane-only": The mail server provider does not run a Web Key Directory but only an OpenPGP DANE service. The Web Key Directory Update protocol is used to update the keys for the DANE service. The use of this keyword is deprecated.
• "auth-submit": The submission of the mail to the server is done using an authenticated connection. Thus the submitted key will be published immediately without any confirmation request.
• "protocol-version": This keyword can be used to explicitly claim the support of a specific version of the Web Key Directory update protocol. This is in general not needed but implementations may have workarounds for providers which only support an old protocol version. If these providers update to a newer version they should add this keyword so that the implementation can disable the workaround. The value is an integer corresponding to the respective draft revision number.
• "submission-address": An alternative way to specify the submission address. The value is the addr-spec part of the address to send requests to this server. If this keyword is used in addition to the submission-address file, both MUST have the same value.

More keywords will be defined in updates to this I-D. There is no registry except for this document. For experimental use of new features or for provider specific settings, keywords MUST be prefixed with a domain name and an underscore.

Security Considerations The use of SHA-1 for the mapping of the local-part to a fixed string is not a security feature but merely used to map the local-part to a fixed-sized string made from a well defined set of characters. It is not intended to conceal information about a mail address. The domain name part of the mail address is not part of the hash to avoid problems with internationalized domain names. Instead a separate URL is required for each domain name. To make it a bit harder to test for published keys, the server responsible to serve the WELLKNOWN directory SHOULD NOT create an index file for that directory or any sub-directory. The mail provider MUST make sure to publish a key in a way that only the mail address belonging to the requested user is part of the User ID packets included in the returned key. Other User ID packets and their associated binding signatures MUST be removed before publication. Confirmation requests MUST only be send for such to be published User ID. It is further recommended that a client filters a received key or a key send for a publication requests so that only the specific User ID with the mail address of the provider is imported or send. A client MUST follow redirects requested by the server as long as the intermediate and final target are also using the https scheme. A client MUST check that the received key has the requested User-Id and MUST drop all other User-Ids found in the received key. A client MUST NOT accept a HTTP authentication challenge (HTTP code 401) because the information in the Web Key Directory is public and needs no authentication. Allowing an authentication challenge has the problem to easily confuse a user with a password prompt and tricking him into falsely entering the passphrase used to protect his private key or to login to his mail provider. The use of DNS SRV records as specified in former revisions of this document reduces the certainty that a mail address belongs to a domain. For example an attacker may change the target to a host in a sub-domain under their control and thus gain full control over all keys.

IANA Considerations

Well-Known URI IANA is requested to assign a well-known URI in the "Well-Known URIs" registry as defined by : URI suffix: openpgpkey Change controller: IETF Specification document: This

Acknowledgments The author would like to acknowledge the help of the individuals who kindly voiced their opinions on the GnuPG mailing lists, in particular, the help of Bernhard Reiter and Guilhem Moulin.

Normative References In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. This document describes how the OpenPGP Message Format can be used to provide privacy and authentication using the Multipurpose Internet Mail Extensions (MIME) security content types described in RFC 1847. [STANDARDS-TRACK] This document is maintained in order to publish all necessary information needed to develop interoperable applications based on the OpenPGP format. It is not a step-by-step cookbook for writing an application. It describes only the format and methods needed to read, check, generate, and write conforming packets crossing any network. It does not deal with storage and implementation questions. It does, however, discuss implementation issues necessary to avoid security flaws. OpenPGP software uses a combination of strong public-key and symmetric cryptography to provide security services for electronic communications and data storage. These services include confidentiality, key management, authentication, and digital signatures. This document specifies the message formats used in OpenPGP. [STANDARDS-TRACK] This memo defines a path prefix for "well-known locations", "/.well-known/", in selected Uniform Resource Identifier (URI) schemes. In doing so, it obsoletes RFC 5785 and updates the URI schemes defined in RFC 7230 to reserve that space. It also updates RFC 7595 to track URI schemes that support well-known URIs in their registry. This document defines ZRTP, a protocol for media path Diffie-Hellman exchange to agree on a session key and parameters for establishing unicast Secure Real-time Transport Protocol (SRTP) sessions for Voice over IP (VoIP) applications. The ZRTP protocol is media path keying because it is multiplexed on the same port as RTP and does not require support in the signaling protocol. ZRTP does not assume a Public Key Infrastructure (PKI) or require the complexity of certificates in end devices. For the media session, ZRTP provides confidentiality, protection against man-in-the-middle (MiTM) attacks, and, in cases where the signaling protocol provides end-to-end integrity protection, authentication. ZRTP can utilize a Session Description Protocol (SDP) attribute to provide discovery and authentication through the signaling channel. To provide best effort SRTP, ZRTP utilizes normal RTP/AVP (Audio-Visual Profile) profiles. ZRTP secures media sessions that include a voice media stream and can also secure media sessions that do not include voice by using an optional digital signature. This document is not an Internet Standards Track specification; it is published for informational purposes.

Sample Protocol Run The following non-normative example can be used by implementors as guidance. Note that GnuPG version 2.1.12 supports the key discovery described in version -00 of this document (auto-key-locate method "wkd"). Version 2.1.16 can run the protocol described in this document but is also able to run the protocol version specified by -01. For backward compatibility this example uses the Content-Type as required for versions of this protocol prior to -04; if the client knows that the server support -04 "vnd.gnupg.wkd" should be used.

Sample Keys This is the provider's submission key: This is the target key to be published:

Sample Messages The first message triggers the publication requests. The server decrypts this message to and returns this confirmation request The client decrypts this PGP/MIME message as creates this response and sends it PGP/MIME Combined signed and encrypted to the server

Changes Since -14

• Deprectated dane-only keyword.