DomainKeys Identified Mail Signatures v2 (DKIM2)

Introduction DomainKeys Identified Mail v2 (DKIM2) permits a person, role, or organization to document that they have handled an email message by associating a domain name with the message . A public key signature is used to record that they have been able to read the contents of the message and write to it. Verification of claims is achieved by fetching a public key stored in the DNS under the relevant domain and then checking the signature. Message transit from author to recipient is through Forwarders that typically make no substantive change to the message content and thus preserve the DKIM2 signature. Where they do make a change the changes they have made are documented so that these can be "undone" and the original signature validated. When a message is forwarded from one system to another an additional DKIM2 signature is added on each occasion. This chain of custody assists validators in distinguishing between messages that were intended to be sent to a particular email address and those that are being "replayed" to that address. The chain of custody can also be used to ensure that delivery status notifications are only sent to entities that were involved in the transmission of a message. Organizations that process a message can add to their signature a request for feedback as to any opinion (for example, that the email was considered to be spam) that the eventual recipient of the message wishes to share.

DKIM2 Architecture Documents Readers are advised to be familiar with the material in TBA, TBA and TBA which provide the background for the development of DKIM2, an overview of the service, and deployment and operations guidance and advice.

Terminology and Definitions This section defines terms used in the rest of the document. DKIM2 is designed to operate within the Internet Mail service, as defined in . Basic email terminology is taken from that specification. DKIM2 inherits many ideas from DKIM () which, for clarity we refer to in this specification as DKIM1. In addition, some features were influenced by experience from (see ) the experimental ARC protocol (). Syntax descriptions use Augmented BNF (ABNF) . The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in . These words take their normative meanings only when they are presented in ALL UPPERCASE.

Signers Elements in the mail system that sign messages on behalf of a domain are referred to as Signers. These may be MUAs (Mail User Agents), MSAs (Mail Submission Agents), MTAs (Mail Transfer Agents), or other agents such as mailing list "exploders". In general, any Signer will be involved in the injection of a message into the message system in some way. The key point is that a message must be signed before it leaves the administrative domain of the Signer.

Forwarder defines a Relay as transmitting or retransmitting a message but states that it will not modify the envelope information or the message content semantics. It also defines a Gateway as a hybrid of User and Relay that connects heterogeneous mail services. In this document we use the concept of a Forwarder which is an MTA that receives a message and then, as an alternative to delivering it into a destination mailbox, can forward it on to another system in an automated, pre-determined, manner.

Reviser As will be seen, a Forwarder may alter the message content or header fields, in such a way that existing signatures on the message will no longer validate. If so, then a record will be made of these changes. We call a Forwarder that makes such changes a Reviser.

Verifiers Elements in the mail system that verify signatures are referred to as Verifiers. These may be Forwarders, Revisers, MTAs, Mail Delivery Agents (MDAs), or MUAs. It is an expectation of DKIM2 that a recipient of a message will wish to verify some or all signatures before determining whether or not to accept the message or pass it on to another entity.

Signing Domain A domain name associated with a signature. This domain may be associated with the author of an email, their organization, a company hired to deliver the email, a mailing list operator, or some other entity that handles email. What they have in common is that at some point they had access to the entire contents of the email and were in a position to add their signature to the email.

Whitespace There are two forms of whitespace used in this specification: WSP represents simple whitespace, i.e., a space or a tab character (formal definition in ). FWS is folding whitespace. It allows multiple lines separated by CRLF followed by at least one whitespace, to be joined. The formal ABNF for these are (WSP given for information only):

The definition of FWS is identical to that in except for the exclusion of obs-FWS.

Imported ABNF Tokens The following tokens are imported from other RFCs as noted. Those RFCs should be considered definitive. The following tokens are imported from : "local-part" (implementation warning: this permits quoted strings) "sub-domain" The following tokens are imported from : "field-name" (name of a header field) Other tokens not defined herein are imported from . These are intuitive primitives such as SP, HTAB, WSP, ALPHA, DIGIT, CRLF, etc.

Common ABNF Tokens The following ABNF tokens are used elsewhere in this document:

Note that base64strings are defined in , but that document does not contain any ABNF. Note that a base64string MUST be padded with trailing = characters if needed. Note that the definitions of both textstring and base64string allow for the presence of FWS, which simplifies folding header fields to an allowable line length. FWS within base64strings will be ignored when their value is being used. FWS within a textstring MUST be treated as a single space when this value is used.

Protocol Elements Protocol Elements are conceptual parts of the protocol that are not specific to either Signers or Verifiers. The protocol descriptions for Signers and Verifiers are described in later sections ("Signer Actions" () and "Verifier Actions" () and this section must be read in the context of those sections.

Selectors To support multiple concurrent public keys per signing domain, the key namespace is subdivided using "selectors". Periods are allowed in selectors and are component separators. Periods in selectors define DNS label boundaries in a manner similar to the conventional use in domain names. This will allow portions of the selector namespace to be delegated. ABNF:

The number of public keys and corresponding selectors for each domain is determined by the domain owner. Many domain owners will use just one selector, whereas administratively distributed organizations can choose to manage disparate selectors and key pairs in different regions or on different email servers. Selectors can also be used to delegate a signing authority, which can be withdraw at any time. Selectors also make it possible to seamlessly replace keys on a routine basis by signing with a new selector, while keeping the key associated with the old selector available.

Tag=Value Lists DKIM2 uses a simple "tag=value" syntax in the Message-Instance and DKIM2-Signature header fields, as well as in domain signature records (see ). Values are a series of strings containing either plain text or "base64" text (as defined in , Section 6.8). The name of the tag will determine the encoding of each value. Unencoded semicolon (";") characters MUST NOT occur in the tag value, since that separates tag-specs. Formally, the ABNF syntax rules are as follows:

Note that WSP is allowed anywhere around tags. In particular, any WSP after the "=" and any WSP before the terminating ";" is not part of the value; however, WSP inside the value is significant. Tags MUST be interpreted in a case-sensitive manner. Values MUST be processed as case sensitive unless the specific tag description of semantics specifies case insensitivity. Tags with duplicate names MUST NOT occur within a single tag-list; if a tag name does occur more than once, the entire tag-list is invalid. Whitespace within a value MUST be retained unless explicitly excluded by the specific tag description. Tag=value pairs that represent the default value MAY be included to aid legibility. Unrecognized tags MUST be ignored. Tags that have an empty value are not the same as omitted tags. An omitted tag is treated as having the default value; a tag with an empty value explicitly designates the empty string as the value.

JSON This document uses JSON for collections of related information. The JSON objects are then base64 encoded and placed into Tag=Value lists. This is done to avoid significant complexities in parsing Tag=Value lists and to simplify the way in which algorithmic agility is provided. Unrecognised fields within JSON objects MUST be ignored.

Signing and Verification Cryptographic Algorithms DKIM2 supports multiple hashing and digital signature algorithms. One hash function (SHA256) if specified here and two signing algorithms are defined by this specification: RSA-SHA256 and Ed25519-SHA256. Signers and Verifiers MUST implement SHA256. Signers SHOULD implement both RSA-SHA256 and Ed25519-SHA256. Verifiers MUST implement both RSA-SHA256 and Ed25519-SHA256.

The SHA256 Hashing Algorithm The SHA256 hashing algorithm is used to compute body and header hashes as defined in and . The resultant values are stored within Message-Instance header fields.

The RSA-SHA256 Signing Algorithm The RSA-SHA256 Signing Algorithm computes a hash over all the Message-Instance and DKIM2-Signature header fields as described in using SHA-256 (FIPS-180-4-2015) as the hash-alg. That hash is then signed by the Signer using the RSA algorithm (defined in PKCS#1 version 1.5 ) as the crypt-alg and the Signer's private key. The hash MUST NOT be truncated or converted into any form other than the native binary form before being signed. The signing algorithm MUST use a public exponent of 65537. Signers MUST use RSA keys of at least 1024 bits. Verifiers MUST be able to validate signatures with keys ranging from 1024 bits to 2048 bits, and they MAY be able to validate signatures with larger keys.

The Ed25519-SHA256 Signing Algorithm The Ed25519-SHA256 Signing Algorithm computes a hash over all the Message-Instance and DKIM2-Signature fields as described in using SHA-256 (FIPS-180-4-2015) as the hash-alg. It signs the hash with the PureEdDSA variant Ed25519, as defined in Section 5.1 of .

Other Algorithms Other algorithms MAY be defined in the future. Verifiers MUST ignore any hashes or signatures using algorithms that they do not implement.

Key Management Some level of assurance is required that a public key is associated with the claimed Signer. DKIM2 does this by fetching the key from the DNS for the domain specified in the d= field. DKIM2 keys are stored in a subdomain named "_domainkey". Given a DKIM2-Signature field with a "d=" tag of "example.com" and an "s1=" tag of "foo.bar", the DNS query will be for "foo.bar._domainkey.example.com". NOTE: these keys are no different, and are stored in the same locations as those for DKIM1 (). Further details can be found in .

Recipes A set of "recipes" is used to recreate the previous version of the body and/or header fields of a message. The recipes are provided within a JSON object with the schema:

Header Recipes A Header Recipe is an array of instructions applied to the specified header fields with the given header field name. These instructions are applied in order to the message which has been received so as to recreate the message as it was before modifications were made. If there is no "h" field in the JSON object then there was no modification to the header fields. If the "h" field is null (there are no recipes for any header field) then the previous state of the header fields cannot be recreated. Verifiers of the message may be able to determine, by seeing which entity makes this declaration, that this is acceptable to them because, for example, that entity is providing a contractually arranged service. Matching of header field names is always done without regard to case. If a header field name is not present in the JSON object then all header fields with that header field name are to be retained. If there are no "recipes" for a header field name that is present in the JSON object then all instances of that header field are to be removed to reinstate the previous state of the message. Header fields are numbered "bottom up" (the opposite direction to the body lines). That is to say, when walking the header fields from the top of the message to end of the header fields then the last header field instance encountered with any particular header field name is numbered 1, the header field (with the same header field name) above that is numbered 2, and so on. The header fields should be treated as being unwrapped (in the normal manner). That is, all of the physical lines that form a single header field are processed under the same logical number. The recipes are processed in order and the resulting header fields are emitted so that later header field will appear above earlier header fields in the recreated message. If the recipe is of the form "(" start "-" end ")" then this means that (relevant) header field lines from to start to end, inclusive, are to be emitted. The start value of each of these recipes MUST be in ascending order and MUST be greater than the end value of all preceding recipes for this header field name. Otherwise the recipe is treated as a text string to which relevant header field name and a colon is prepended and a CRLF is appended and the resultant string is then emitted. Note that the way in which hashes are calculated (see below) means that no heed needs to be taken of wrapping or the case of the header field name. The text string MUST NOT contain CR or LF characters. If the string is empty then the CRLF will immediately follow the header field name and colon.

Body Recipes A Body Recipe is an array of instructions applied to message body which can recreate the message as it was before modifications were made. If there is no "b" field in the JSON object then there was no modification to the message body. Note that the JSON schema requires either "h" or "b" to be present. If the "b" field is null (there are no recipes) then the previous state of the message body cannot be recreated. Verifiers of the message may be able to determine, by seeing which entity makes this declaration, that this is acceptable to them because, for example, that entity is providing a contractually arranged service. Body lines are numbered "top down" (the opposite direction to the header fields). The first line of the body (immediately after the blank line that indicates that there are no more header fields) is numbered 1. The recipes are processed in order and the resulting body lines fields are emitted so that later lines will appear below earlier lines in the recreated message. If the recipe is of the form "(" start "-" end ")" then this means that (relevant) message body lines from to start to end, inclusive, are to be emitted. The start value of each of these recipes MUST be in ascending order and MUST be greater than the end value of all preceding Body Recipes. Otherwise the recipe is treated as a text string and a CRLF is appended and the resultant string is emitted. The text string MUST NOT contain CR or LF characters. If the string is empty then just a CRLF is emitted.

Message Hash Values A set of cryptographic "hashes" are used to record the current message body and header fields. The hashes are provided within a JSON object with the schema:

To provide for algorithmic dexterity more that one hash value, using a different algorithm MAY be supplied. A verifier MUST check all signatures that it understands and SHOULD treat any failure as invalidating all hashes. The method for computing the header hash value(s) is given in and the method for computing the body hash value in .

SMTP parameters DKIM2 records the MAIL FROM and RCPT TO values to be used when this message is transmitted over an SMTP link from the signing MTA. These values are provided with a JSON object with the schema:

Note that MAIL FROM may be just "<>", for example for a Delivery Status Notification. When a message is intended for more than one recipient then the RCPT TO values provided MAY include all of the recipients so that a single copy of the email MAY be sent to all of the recipients in a single SMTP transaction. Alternatively, multiple copies of the email may be generated so as to not immediately reveal who else received the email. However, if "bcc:" recipients are involved then in order to meet the requirements of Section 3.6.3 each and every bcc recipients MUST NOT revealed to any other message recipient.

Message Signature A set of cryptographic "hashes" are used to record the current message body and header fields. The hashes are provided within a JSON object with the schema:

The selector values subdivides the namespace for the domain being used for signing. The algorithm value is the one used to generate the signature. Verifiers MUST support "RSA-SHA256" for which the string "rsa" is used and "Ed25519-SHA256" for which the string "ed25519" is used; See for a description of these algorithms. To provide for algorithmic dexterity more than further signatures, using different algorithms MAY be supplied. A verifier MUST check all signatures that it understands and SHOULD treat any failure as invalidating all signatures. Since the DNS lookup for the public key will check that the algorithm is correct a different selector MUST necessarily be used for the second signature. The method for computing the signature values is given in

The Message-Instance Header Field A Message-Instance header field documents the current contents of the message and, in the case of a Reviser records any relevant changes that have been made to the incoming message. The Message-Instance header field is a tag-list as described in . Tags on the Message-Instance header field along with their type, encoding and requirement status are shown below. v= The revision number of the Message-Instance header field.

The originator of a message uses the value 1. Further Message-Instance header fields are added with a value one more than the current highest numbered Message-Instance header field. Gaps in the numbering MUST be treated as making the whole message impossible to verify. ABNF:

r= The recipes that will allow recreation of the previous instance of the message.

The r= tag value is the base64 encoded version of the JSON object that contains the recipes that allow the previous instance of the message to be recreated (see {#JSONrecipe}}. ABNF:

h= The hash values for the message

The h= tag value is the base64 encoded version of the JSON object that contains the hashes for the message (see {#JSONrhash}}. ABNF:

The DKIM2-Signature Header Field The signature of the email is stored in a DKIM2-Signature header field. This header field contains all of the signature and key- fetching data. The DKIM2-Signature value is a tag-list as described in . Tags on the DKIM2-Signature header field along with their type, encoding and requirement status are shown below. It will be noted that we have not included a version number. Experience from IMF onwards shows that it is essentially impossible to change version numbers. If it becomes necessary to change DKIM2 in the sort of incompatible way that a v=2 / v=3 version number would support, it is expected that header fields will be labelled as DKIM3 instead. i= The sequence number of the DKIM2-Signature header field.

The originator of a message uses the value 1. Further DKIM2-Signature header fields are added with a value one more than the current highest numbered DKIM2-Signature header field. Gaps in the numbering MUST be treated as making the whole message unsigned. ABNF:

v= The highest numbered Message-Instance header field

This value allows verifiers to determine which entity made a particular revision to the message body or header fields. ABNF:

n= Nonce value

This value, if present, has a meaning to the creator of the signature but MUST NOT be assumed to have any meaning to any other entity. It MAY be used as an index into a database to assist in handling Delivery Status Notifications or for any other purpose. To discourage use of the tag field as an alternative to the use of more appropriate header fields, the length of the textstring MUST NOT exceed 64 characters and implementations SHOULD reject messages where this limit has been ignored. ABNF:

t= Signature Timestamp

The time that this header field was created. The format is the number of seconds since 00:00:00 on January 1, 1970 in the UTC time zone. The value is expressed as an unsigned integer in decimal ASCII. This value is not constrained to fit into a 31- or 32-bit integer. Implementations SHOULD be prepared to handle values up to at least 10^12 (until approximately AD 200,000; this fits into 40 bits). Implementations MAY ignore signatures that have a timestamp in the future. Implementations MAY ignore signatures that are more than 14 days old. ABNF:

d= The domain associated with this signature.

This domain is used to form the query for the public key. The domain MUST be a valid DNS name under which the DKIM2 key record is published. The domain in the d= tag MUST exactly match the rightmost labels of the domain-name part of the mf= tag. That is to say, the domain-name of the mf= tag MUST either match the d= domain exactly or be a sub-domain of d= domain. ABNF:

m= The SMTP parameters for the message

The m= (mail) tag value is the base64 encoded version of the JSON object that contains the MAIL FROM and RCPT TO parameters that will be used when the message is transmitted over SMTP from the signing MTA (see {#JSONsmtp}}. ABNF:

s= The signature value(s) for the message

The s= tag value is the base64 encoded version of the JSON object that contains the signature(s) for the message (see {#JSONsignature}}. ABNF:

f= Flags

Flags serve two purposes; they either report what has been done to the message by the system creating the DKIM2-Signature or they make a request to systems that handle the mail thereafter. Flags are separated by commas, and optional white-space allows systems to add several flags without creating long lines. If a flag value is not recognised it MUST be ignored. The flag values that report things are: "exploded": this message (identified by its unique header hash value (recorded in b1= and perhaps b2=)) is being sent to more than one email address. An MTA which receives a message MAY use this information to help it distinguish between malicious "DKIM replay" and legitimate activity performed by mailing list. If this flag is not present in at least one DKIM2-Signature header field then an MTA MAY assume that only one copy of a particular message (identified by relevant cryptographic hash values) is intended to exist; The flags values that make requests are: "donotexplode": this signer requests that the message not be sent to more than one recipient. A system that, by local policy, ignores this request MUST NOT allow any of the copies it creates to be forwarded on to any MTA outside its control. "donotmodify": this signer requests that the message not be modified from the form in which it is sent. A system that, by local policy, ignores this request MUST NOT allow the message to be forwarded on to any MTA outside its control. "feedback": this signer requests feedback about how this message is handled during delivery and thereafter. This document does not describe what such feedback might be or where it might be delivered. If this flag is absent then feedback is explicitly not required. ABNF:

Computing the Body Hash Although the body hash value will be incorporated into a Message-Instance header field, these header fields are ignored when calculating the header hash value and so the body hash and header hash may be calculated in any convenient order. The body of messages is treated as merely a string of octets. DKIM2 messages MAY be either in plain-text or in MIME format; no special treatment is afforded to MIME content. Message attachments in MIME format MUST be included in the content that is signed. The DKIM2 body hash is calculated in the same manner as DKIM1's "simple" scheme: All empty lines at the end of the message body are ignored. An empty line is a line of zero length after removal of the line terminator. If there is no body or no trailing CRLF on the message body, a CRLF is added. That is "*CRLF" at the end of the body is converted to "CRLF". No other changes are made to the body, which is then processed by the relevant hash algorithm(s). The hash value is converted to base64 form and inserted into (Signers) or compared to (Verifiers) the "bh1=" tag of the Message-Instance header field that is being created. If a second hash is calculated then its base64 representation will be included in the "bh2=" tag.

Preventing Transport Conversions DKIM2's design is predicated on valid input. In order to be signed a message will need to be in "network normal" format (text is ASCII encoded, lines are separated with CRLF characters, etc.). A message that is not compliant with , , and other relevant message format standards can be subject to attempts by intermediaries to correct or interpret such content. See Section 8 of for examples of changes that are commonly made. Such "corrections" may invalidate DKIM2 signatures or have other undesirable effects, including some that involve changes to the way a message is presented to an end user. When calculating the hash on messages that will be transmitted using base64 or quoted-printable encoding, Signers MUST compute the hash after the encoding. Likewise, the Verifier MUST incorporate the values into the hash before decoding the base64 or quoted-printable text. However, the hash MUST be computed before transport-level encodings such as SMTP "dot-stuffing" (the modification of lines beginning with a "." to avoid confusion with the SMTP end-of-message marker, as specified in ). Further, if the message contains local encoding that will be modified before transmission, that modification to canonical form MUST be done before signing. In particular, bare CR or LF characters (used by some systems as a local line separator convention) MUST be converted to the SMTP-standard CRLF sequence before the message is signed. Any conversion of this sort SHOULD be applied to the message actually sent to the recipient(s), not just to the version presented to the signing algorithm. More generally, the Signer MUST sign the message as it is expected to be received by the Verifier rather than in some local or internal form.

Computing the Header Fields Hash The header fields hash calculation MUST apply the following steps in the order given. A verifier will need to do the equivalent steps in order to check that the hash they have received is correct. Ignore some header fields When calculating the header field hash "Received" or "Return-Path" header fields MUST be ignored. These are Trace headers as described in and serve only to document details of the SMTP transmission process. When calculating the header field hash any header field with a header field name starting with "X-" MUST be ignored. Currently deployed email systems use these fields as proprietary Trace headers which have no defined meaning for other systems and it considerably simplifies reporting on changes to header fields to ignore them. When calculating the header field hash any "Message-Instance" or "DKIM2-Signature" header fields MUST be ignored. These header fields will be included in the hash value that will be signed by a DKIM2-Signature header field and it simplifies implementations if they are not included twice, especially when determining whether all modifications to a message have been correctly declared. When calculating the header field hash any "DKIM-Signature" header fields and any header fields whose field name starts with "ARC-" MUST be ignored. Not including DKIM1 and ARC signatures means that systems that wish to add other types of signature as well as a DKIM2 signature are free to do this in any convenient order. Convert all header field names (not the header field values) to lowercase. For example, convert "SUBJect: AbC" to "subject: AbC". Unfold all header field continuation lines as described in ; in particular, lines with terminators embedded in continued header field values (that is, CRLF sequences followed by WSP) MUST be interpreted without the CRLF. Implementations MUST NOT remove the CRLF at the end of the header field value. Convert all sequences of one or more WSP characters to a single SP character. WSP characters here include those before and after a line folding boundary. Delete all WSP characters at the end of each unfolded header field value. Delete any WSP characters remaining before and after the colon separating the header field name from the header field value. The colon separator MUST be retained. Place the header fields in alphabetical order by the header field name. If there is more than one header with the same header field name then the header fields are placed in the order in which they occur in the message header, from the top downwards. It is sometimes suggested that some MTAs re-order header fields after they receive an email, if they do then it is their responsibility to recover the original order of any header fields with identical header field names (that are part of a signature calculation) before verifying an existing signature or passing a previously signed message to another MTA that is going to wish to verify a signature. The hash(es) of the concatenated header fields are calculated. The hash value is converted to base64 form and inserted into (Signers) or compared to (Verifiers) the JSON object that will be base64 encoded and placed into the Message-Instance header field that is being created. If further hashes are calculated then their base64 representations will be included in the array structure within the JSON object, see .

The Relaxed Domain Match Algorithm To assist in addressing the "DKIM replay" problem DKIM2 provides a "chain of custody" for every message. This is established by checking that the MAIL FROM value recorded in every DKIM2-Signature header field (except of course the i=1 instance) can be matched with a RCPT TO value of the next lower numbered DKIM2-Signature header field. It is also necessary to check DKIM2-Signature header fields for a match between the signing domain (specified in the d= tag) and the MAIL FROM domain. To allow systems to use existing "bounce-handling" schemes with special subdomains in their MAIL FROM values a "relaxed" approach is taken to the matches between these values. Only the domain part of the MAIL FROM and RCPT TO values is used for these matches The local part (and the @) are ignored. If there is not an exact match between the domain names then labels are removed, one by one from the left hand side of the MAIL FROM domain name and the comparison is repeated. If no labels remain then there is no match.

Signer Actions This section gives the actions that need to be undertaken by the signer of a message. They may be done in any appropriate order.

Add any Necessary Message-Instance Header Fields If a system is generating the initial form of a message or if it a Reviser that has made to changes to the message body and/or header fields then it MUST compute the body hash as described in and the hash of the header fields as described in . If the message does not contain a Message-Instance header field then one MUST be added. If hashing the message body or relevant header fields does not give the same hash values as those recorded in the highest version (v=) Message-Instance header field then a new Message-Instance header field MUST be added. This Message-Instance header field MUST contain "recipes" to be able to recreate the message corresponding to the hash values in the currently highest numbered Message-Instance header field, or a null recipe to indicate that recreating the previous version of the message will not be possible. A system may add more than one Message-Instance header field if it wishes to do so, but the DKIM2 design allows all modifications made by any single system to be documented in a single Message-Instance header field. Note that the first (v=1) Message-Instance header fieldMAY contain "recipes" if it is wished to record any changes made to a message as it enters the DKIM2 ecosystem. All other Message-Instance header fields SHOULD contain at least one "recipe".

Provide a "chain of custody" for the message To construct the DKIM2-Signature header field contains the MAIL FROM and RCPT TO values that will be used when the message is transmitted so these "envelope" values MUST be available to (or deducible by) a Signer. The receiver of a message will check for an exact match (including the local parts of the email addresses) between the MAIL FROM / RCPT TO protocol values and the values in the highest numbered (most recent) DKIM2-Signature header field. It is acceptable for there to be more RCPT TO email addresses recorded in the header field that are actually used in the SMTP conversation, but any RCPT TO value which is used MUST be present. Verifiers will check for a relaxed domain match (see {relaxed-domain-match}) between the signing domain (d=) and the domain in the MAIL FROM value. When the message being signed already has a DKIM2-Signature header field (i.e. it has already been transmitted at least once) then a valid "chain of custody" MUST be apparent when all of the DKIM2-Signature header fields are considered. This "chain of custody" contributes to the way in which DKIM2 tackles "DKIM replay" attacks. The "chain of custody" uses a relaxed domain match (see {relaxed-domain-match}). In any situation where a messages will be forwarded in such a way that the mf= on the outgoing message is such that the "chain of custody" would be broken then the Signer MUST generate an extra DKIM2-Signature header field that causes values to match, i.e. a record must be fabricated that documents the mail being passed from one domain to another. It will be noted that the creation of this extra header field will require the Signer to have access to a DKIM2 private key associated with a domain in the RCPT TO entry. This is often achieved by the Signer creating the private key and never sharing it. The Signer gives the public key (and selector value) to the domain owner who creates an appropriate DNS entry. Alternatively, the Signer creates a public key DNS entry within a part of the DNS that they control and the domain owner merely needs to publish a CNAME pointing at this.

Select a Private Key and Corresponding Selector Value This specification does not define the basis by which a Signer should choose which private key and selector value to use -- this will be a matter of administrative convenience. Distribution and management of private keys is also outside the scope of this document.

Calculate a Signature Value A Signer calculates a hash solely over the Message-Instance and DKIM2-Signature header fields of the message and then signs this hash value. The hashes of the body and other header fields are covered by the hashes in the highest version (v=) Message-Instance header field. Note that the DKIM2-Signature header field contains a signature but does not give the hash value that was signed. This permits flexibility for any future signature schemes where a relevant hash value may not be readily available (or may be inordinately long). The signature algorithm MUST apply the following steps in the order given. Convert all relevant header field names (not the header field values) to lowercase. For example, convert "DKIM2-signature" to "dkim2-signature". Unfold all header field continuation lines as described in ; in particular, lines with terminators embedded in continued header field values (that is, CRLF sequences followed by WSP) MUST be interpreted without the CRLF. Implementations MUST NOT remove the CRLF at the end of the header field value. Convert all sequences of one or more WSP characters to a single SP character. WSP characters here include those before and after a line folding boundary. Delete all WSP characters at the end of each unfolded header field value. Delete any WSP characters remaining before and after the colon separating the header field name from the header field value. The colon separator MUST be retained. Place the header fields in order. First come the Message-Instance header fields in ascending version (v=) order. Second are the DKIM2-Signature header fields in ascending sequence (i=) order. Last of all is an incomplete DKIM2-Signature header field (the one that this system is creating) with all tags present except that the signature value (s=) is null (that is the base64 value is absent). The hash of the concatenated header fields is calculated and this value is then signed using an the appropriate algorithm and the selector value, algorithm identifier and the base64 encoded value of the signature value are placed into the JSON object for the message signature. Any further signatures are calculated and added to the array within the JSON object (see The JSON object holding the signature is then converted to base64 and inserted into the s= tag of the DKIM2-Signature header field which MUST then be placed into the message.

Verifier Actions This section discusses the actions taken by a Verifier. In essence this will involve repeating all the actions taken by a Signer to produce a Message-Instance or DKIM2-Signature header field. To avoid a lot of repetition these actions will not be spelled out in detail. Once a hash value has been calculated it is then compared with the value reported by the Signer, or the Signer's public key is used to determine whether the signature that has been provided is correct.

Output States A verification ends in one of three states, which this document refers to as follows: SUCCESS: a successful verification PERMFAIL: a permanent, non-recoverable error such as a signature verification failure or mismatched hash value TEMPFAIL: a temporary, recoverable error such as a DNS query timeout A verifier MAY cease verifying once a single failure is detected. Verifiers wishing to communicate the results of verification to other parts of the mail system may do so in whatever manner they see fit. For example, implementations might choose to add an email header field to the message before passing it on. Any such header field SHOULD be inserted before any existing DKIM2-Signature or pre-existing authentication status header fields in the header field block. The Authentication-Results: header field () MAY be used for this purpose. It should be noted that any "Authentication-Results" header field will count as a modification to the email if any further DKIM2-Signature header fields are to be generated.

Validation of Tag Fields Implementers MUST meticulously validate the format and values of Message-Instance and DKIM2-Signature header fields. Errors SHOULD be treated as a PERMFAIL (signature syntax error). Being "liberal in what you accept" is an inappropriate strategy. Note, however, that the presence of unknown tags in a DKIM2-Signature header field (or a Message-Instance header field), MUST NOT cause a verification to fail. Verifiers MAY return PERMFAIL ("signature expired") if it is more than 14 days since the timestamp recorded in the "t=" tag of a DKIM2-Signature header field.

Fetching the Public Key The public key of a signature is needed to complete the verification process. Details of key management and representation are described in and . The Verifier MUST validate the key record and MUST ignore any public key records that are malformed. When validating a message, a Verifier MUST perform the following steps in a manner that is semantically the same as performing them in the order indicated; in some cases, the implementation may parallelize or reorder these steps, as long as the semantics remain unchanged: The Verifier retrieves the public key as described in using the "d=" tag, and the selector from within the JSON object in the "s1=" tag. If there is more than one signature within the JSON object then these steps are repeated for each one. If the query for the public key fails to complete, the Verifier MAY seek a later verification attempt by returning TEMPFAIL ("key unavailable"). If the query for the public key fails because the corresponding key record does not exist, the Verifier MUST return PERMFAIL ("no key for signature"). If the query for the public key returns multiple key records, then the return PERMFAIL ("more than one key returned" since this is not permitted by ). If the result returned from the query does not adhere to the format defined in the DKIM key specification , the Verifier MUST ignore the key record and return PERMFAIL ("key syntax error"). Verifiers are urged to validate the syntax of key records carefully to avoid attacks. In particular, the Verifier MUST ignore keys with a version code ("v=" tag) that they do not implement. If the public key data (the "p=" tag) is empty, then this key has been revoked and the Verifier MUST treat this as a failed signature check and return PERMFAIL ("key revoked"). There is no defined semantic difference between a key that has been revoked and a key record that has been removed. If the public key data is not suitable for use with the algorithm specified in the DKIM-Signature header field, the Verifier MAY immediately return PERMFAIL ("inappropriate key algorithm"). However, the tag fields in the public key record that would cause this to occur are now deprecated so DKIM2 implementations MAY ignore these tag fields altogether. If the "h=" tag exists in the public key record and the hash algorithm implied by the type of signature being checked is not included in the contents of the "h=" tag, the Verifier MUST ignore the key record and return PERMFAIL ("inappropriate hash algorithm").

Perform the Signature Verification Calculation Given a Signer and a public key, verifying a signature consists of actions semantically equivalent to the following steps. Prepare a canonicalized version of the Message-Instance and DKIM2-Signature header fields as described in . Note that this canonicalized version does not actually replace the original content. Based on the algorithm indicated in the JSON object holding the signatures compute the signatures of the canonical copy. Then verify that the there is a match with the hash value from the JSON object. If the signature does not validate, the Verifier SHOULD ignore the signature and return PERMFAIL ("signature did not verify"). Otherwise, the signature has correctly verified. If there is more than one signature provided then they MUST all be checked if the verifier is able to do so. If any signature fails then an error SHOULD be reported. If some signatures fail and other pass then the error that is reported should provide that information (e.g. PERMFAIL "RSA signature passed, elliptic curve signature failed") The reasoning for requiring that all signatures pass is that if a signature scheme has recently become deprecated because it is known to be cryptographically flawed then Signers will use a second (unbroken) signature scheme. However, such a Signer may still provide the other signature for the benefit of Verifiers that have yet to upgrade -- reasoning perhaps that attacks are too expensive to be a very significant security issue. A Verifier that determines that one signature passes whilst the other fails may well be in a position to prevent an attack.

Check Most Recent Signature and Hashes for the Message A Verifier SHOULD check the validity of the most recently applied (highest numbered i= value) DKIM2-Signature header field and the associated (v=) Message-Instance before accepting an email. If this check does not pass then a Delivery Status Notification for the email MUST NOT be generated thereafter -- hence the best strategy, if the email is not wanted, is to reject it (with a 5xx error code) whilst the relevant SMTP conversation is still ongoing. If the check gives a TEMPFAIL result then a 4xx error code SHOULD be used to allow the sending MTA to understand the situation. A Verifier SHOULD check that the MAIL FROM value in the most recent DKIM2-Signature header field is identical to the MAIL FROM values of the SMTP protocol interaction that delivered the email to the Verifier. A Verifier SHOULD also check that all of the RCPT TO values from the SMTP protocol occur in the most recent DKIM2-Signature header field. The values MUST BE put into lower-case before doing these checks. Note that these check MUST NOT use the relaxed domain match algorithm. A Verifier SHOULD check that there is a relaxed domain match (see {relaxed-domain-match}) between the signing domain of the most recently applied DKIM2-Signature header field and the mf= value in that header field. A Verifier SHOULD also check the chain of custody for the message (see {chain-of-custody}) is valid, using a relaxed domain match (see {#relaxed-domain-match}). Should checking these signatures (all but the most recently applied) give the status TEMPFAIL then it may be possible for the Verifier to determine the validity of the signature at a later time. It the TEMPFAIL status continues to occur, or if a PERMFAIL is encountered then this MAY be reported to the sending MTA by means of a Delivery Status Notification. However the non-validating email MUST NOT be forwarded to any MTA outside of the current organisation.

Checking the Message-Instance Header Fields The highest numbered (v=) Message-Instance header field SHOULD be checked to determine that the message body has not been altered since the body hash was calculated. If the message has been modified since its original creation then the Message-Instance header fields will enable a Verifier to determine whether or not all the changes made are correctly recorded by using the "recipes" to construct each preceding version of the message. Note that if it is only the first form of the message is of interest then all the "recipes" can be applied in turn and only one hash value checked -- the correctness of the intermediate hash values are not relevant to this assessment.

Checking the DKIM2-Signature Header Fields However, in order to check the chain of custody, to assess whether the message has been exploded, to pick out "feedback" requests to be honoured or to assign reputation to Revisers then all of the DKIM2-Signature header fields will have to checked for validity. The TBA document explores these issues in more detail.

Interpret Results/Apply Local Policy It is beyond the scope of this specification to describe what actions the recipient of an email performs, but mail carrying valid DKIM2 signatures gives the recipient opportunities that unauthenticated email would not. Specifically, an authenticated email provides predictable information by which other decisions can reliably be managed, such as trust and reputation. Conversely, it is hard to assign trust or reputation to unauthenticated email. If an MTA wishes to reject messages where signatures are missing or do not verify, the handling MTA SHOULD use a 550/5.7.x reply code. Where the Verifier is integrated within the MTA and it is not possible to fetch the public key, perhaps because the key server is not available, a temporary failure message MAY be generated using a 451/4.7.5 reply code, such as: 451 4.7.5 Unable to verify signature - key server unavailable Temporary failures such as inability to access the key server or other external service are the only conditions that SHOULD use a 4xx SMTP reply code. In particular, cryptographic signature verification failures MUST NOT provoke 4xx SMTP replies.

Delivery Status Notifications in the DKIM2 ecosystem In the DKIM2 ecosystem, when a message cannot be delivered then this is reported to the sending machine by means of an return code or, if the SMTP session has completed, by generating a Delivery Status Notification (DSN, as defined in . A DSN MUST be addressed to the MTA that sent the message. This prevents "backscatter" by passing failures back along the chain of MTAs that were in involved in passing the message forwards. This is achieved by using the mf= tag from the highest numbered DKIM2-Signature field. If this field is null ("mf=<>") then a DNS MUST NOT be sent.

DSN contents As set out in , the DSN has a top-level MIME part of type multipart/report. Among other things, that MIME part must contain a MIME part of type message/rfc822 that holds either the original message exactly as it was submitted by the sending system or just the header fields of that message. All relevant DKIM2-Signature header fields (and Message-Instance header fields if the message body is supplied) MUST verify. The DSN itself MUST have appropriate Message-Instance and DKIM2-Signature fields, noting that the MAIL FROM to be used will be null ("<>"). If the message body has been truncated (rather than omitted altogether) then in order to allow verification of the DNS contents a Message-Instance header field MUST be added to the message with a body recipe of "z".

Bounce propagation A Forwarder which receives a DSN MAY decide to propagate this DSN to the MAIL FROM address used to deliver the message to it (which can be found in the relevant DKIM2-Signature header field). The DSN SHOULD be handled in the usual way, with Message-Instance header fields documenting any changes and a DKIM2-Signature field with an incremented hop count value added. The Forwarder MAY alternatively decide to reconstruct the message (or just the message header fields) as they were when the message was delivered to the Forwarder and construct a DSN using that information. The information in Message-Instance header fields can be used to achieve this. The resultant DSN is sent to the MAIL FROM address from the now highest numbered DKIM2-Signature header field. Doing this will ensure that details of where the message was forwarded to will not be revealed to the previous hop.

Authentication of inbound bounce notifications When a system receives a DKIM2 signed bounce notification, and the included original message is also DKIM2 signed, it SHOULD verify that this message (or just the header fields if the body is not present) has not been altered. This means: 1) The DSN's DKIM2-Signature will have a signing domain that is aligned with the recipient of the message that is being returned. The recipient's address is located in the rt= tag of the last (highest i= tag) DKIM2-Signature in the returned message. 1) The last (highest i= tag) DKIM2-Signature header field of the returned message will be one that was generated by the system receiving the bounce notification, determined by examining the d= and mf= tags of that DKIM2-Signature header field. 1) The header fields of the embedded message (in the message/rfc822 MIME part) can be verified. If the message body is present then that can also be verified by inspecting the Message-Instance header field(s). If the verification fails then the DSN MUST NOT be propagated any further. If verification has been performed prior to accepting the DSN from the sender the DSN SHOULD be rejected with a 550/5.7.x return code. If the verification cannot be completed because of a temporary issue (with DNS lookups) then a 4xx return code should be used.

EAI () considerations for DKIM2 TBA

IANA Considerations TBA

Security Considerations TBA

Changes from Earlier Versions draft-clayton-dkim2-spec-07 Mapped various fields in Message-Instance and DKIM2-Signature into JSON. Removed restriction on no recipe for v=1 Message-Instance. Allowed indefinite number of algorithms rather than just 2. draft-clayton-dkim2-spec-06 Changed r= to rb= and r.fieldname to rh=fieldname because the tags in a tag value field may not use "-" characters. Also fixed the ABNF. Incorporated Allen Robinson's BOUNCE draft. Assorted other very minor fixes. draft-clayton-dkim2-spec-05 Assorted fixes, with particular thanks to Hannah Stern. Clarified that there are two types of rt/mt matches performed. Specified that recipes must not contain overlaps. Set out the need for matching mf= and d= and put the relaxed domain match algorithm into its own section. Set out that in practice all DKIM2-Signature header fields will need to be checked. draft-clayton-dkim2-spec-04 Added a definition of a Reviser. Incorporate the Message-Instance scheme previously found in ALGEBRA. Recast the text relating to hashes and signatures accordingly. Changed t= back to just digits. draft-clayton-dkim2-spec-03 Removed the pp= proposal, and briefly explained how signers often handle private keys on behalf of domain owners. Changed t= to be human-readable. Fixed description of body canonicalisation to match DKIM1/relaxed. draft-clayton-dkim2-spec-02 Further clarifications and tidying up; alignment of ALGEBRA description with the new MailVersion header field-name; addition of h= tag field. Also added the pp= mechanism to address forwarders who do not have private keys to hand to make the rt/mf/rt chains validate. draft-clayton-dkim2-spec-01 Significant re-ordering of sections and removal of repetitious material. Relax the matching algorithm between rt= and mf= [[This section to be removed by RFC Editor]]