| Internet-Draft | did:x509 | February 2026 |
| Riechert, et al. | Expires 10 August 2026 | [Page] |
This document defines the did:x509 decentralized identifier method, which enables a direct, resolvable binding between X.509 certificate chains and compact issuer identifiers (DID string). In particular, the did:x509 identifier format in this documents comes with a CWT Claims definition. In general, this identifier is a compact and interoperable mechanism for certificate-based identification by combining a certificate fingerprint with optional policies for subject names, subject alternative names, extended key usage, and issuer information. It is especially useful for policy evaluation and reference in transparency services and similar systems requiring cryptographic binding to certificate material.¶
This Informational document is published as an Independent Submission to improve interoperability with Microsoft's architecture. It is not a standard nor a product of the IETF.¶
This note is to be removed before publishing as an RFC.¶
Source for this draft and an issue tracker can be found at https://github.com/henkbirkholz/draft-birkholz-did-x509.¶
This Internet-Draft is submitted in full conformance with the provisions of BCP 78 and BCP 79.¶
Internet-Drafts are working documents of the Internet Engineering Task Force (IETF). Note that other groups may also distribute working documents as Internet-Drafts. The list of current Internet-Drafts is at https://datatracker.ietf.org/drafts/current/.¶
Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress."¶
This Internet-Draft will expire on 10 August 2026.¶
Copyright (c) 2026 IETF Trust and the persons identified as the document authors. All rights reserved.¶
This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (https://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Revised BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Revised BSD License.¶
This document aims to define an interoperable and flexible decentralized identifier ([DID]) format for COSE messages that transport or refer to X.509 certificates using [RFC9360]. The did:x509 identifier format implements a direct, resolvable binding between a certificate chain and a compact issuer string. It can be used in a COSE Header CWT Claims map as defined in [RFC9597].¶
Including a certificate chain directly in configuration or in policy is often impractical. This is due to its size, and to the frequency at which some elements, particularly the leaf, are refreshed. Relying on a partial certificate chain (e.g., a root certificate and some intermediary certificates) is similarly unwieldy. While stable, the level of granularity afforded by a partial certificate chain may not be sufficient to distinguish several identities that are not equivalent for the purpose of policy.¶
Combining authority pinning with attribute assertions is a precise and stable way of capturing identities as a constrained set of certificates. Their representation as compact and durable identifier strings enables the formulation of readable policy (e.g. "request.issuer == 'did:x509...'"), for example in the context of transparency ledger registration.¶
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 BCP 14 [RFC2119] [RFC8174] when, and only when, they appear in all capitals, as shown here.¶
In this document, CDDL ([RFC8610], [RFC9165]) is used to describe the data formats, and ABNF (defined in [RFC5234]) to describe identifiers.¶
The reader is assumed to be familiar with the vocabulary and concepts defined in [I-D.ietf-scitt-architecture].¶
Rego is a descriptive query language used to define policies in a precise and unambiguous way. This document uses Rego ([REGO]) to define the parsing and validation logic for did:x509 identifiers. The Rego code snippets provided in this document can be evaluated using any Rego v1 runtime, but there is no expectation that implementations use the Rego language.¶
Per [RFC8792], line breaks may be present in the figures of this document to stay within the line-length limits of this document's format.¶
The did:x509 ABNF definition defined below uses the syntax defined in [RFC5234] and the corresponding definitions for ALPHA and DIGIT.
[DID] contains the definitions for idchar and pct-encoded in Section 3.1.¶
idchar = ALPHA / DIGIT / "." / "-" / "_" / pct-encoded pct-encoded = "%" HEXDIG HEXDIG¶
=============== NOTE: '\' line wrapping per RFC 8792 ================
did-x509 = "did:x509:" method-specific-id
method-specific-id = version ":" ca-fingerprint-alg ":" ca-\
fingerprint 1*("::" predicate-name ":" predicate-value)
version = 1*DIGIT
ca-fingerprint-alg = "sha256" / "sha384" / "sha512"
ca-fingerprint = base64url
predicate-name = 1*ALPHA
predicate-value = *(1*idchar ":") 1*idchar
base64url = 1*(ALPHA / DIGIT / "-" / "_")
Implementations of this specification MUST indicate a version value of 0.¶
ca-fingerprint-alg is one of sha256, sha384, or sha512.
ca-fingerprint is chain[i].fingerprint[ca-fingerprint-alg] with i > 0, that is, either an intermediate or root CA certificate.
predicate-name is a predicate name and predicate-value is a predicate-specific value.
:: is used to separate multiple predicates from each other.¶
The following sections define the predicates and their predicate-specific syntax.¶
Validation of predicates is defined using policies written in the Rego language ([REGO]), rather than pseudo-code. This is to avoid ambiguity and to make it possible for a reader to evaluate the logic automatically.¶
The inputs to the resolution process are the DID string itself and the x509chain DID resolution option, which carries a comma-separated base64url-encoded X.509 certificate chain.
To evaluate the reference Rego code shown below, the DID and certificate chain have to be passed to a Rego runtime as a JSON document: {"did": "<DID>", "chain": <CertificateChain>}, where did is the DID string and chain is the parsed representation of the certificate chain derived from the x509chain resolution option.¶
Core Rego policy:¶
parse_did(did) :=
[ca_fingerprint_alg, ca_fingerprint, predicates] if {
prefix := "did:x509:0:"
startswith(did, prefix) == true
rest := trim_prefix(did, prefix)
parts := split(rest, "::")
[ca_fingerprint_alg, ca_fingerprint] := split(parts[0], ":")
predicates_raw := array.slice(parts, 1, count(parts))
predicates := [y |
some i
s := predicates_raw[i]
j := indexof(s, ":")
y := [substring(s, 0, j), substring(s, j+1, -1)]
]
}
valid if {
[ca_fingerprint_alg,
ca_fingerprint,
predicates] := parse_did(input.did)
ca := [c | some i; i != 0; c := input.chain[i]]
ca[_].fingerprint[ca_fingerprint_alg] == ca_fingerprint
valid_predicates := [i |
some i
[name, value] := predicates[i]
validate_predicate(name, value)
]
count(valid_predicates) == count(predicates)
}
The overall Rego policy is assembled by concatenating the core Rego policy with the Rego policy fragments in the following sections, each one defining a validate_predicate function.¶
Some of the predicates that are defined in subsequent sections require values to be percent-encoded. Percent-encoding is specified in Section 2.1 of [RFC3986]. All characters that are not in the allowed set defined below must be percent-encoded:¶
allowed = ALPHA / DIGIT / "-" / "." / "_"
Note that most libraries implement percent-encoding in the context of URLs and do NOT encode ~ (%7E).¶
predicate-name = "subject"
predicate-value = key ":" value *(":" key ":" value)
key = label / oid
value = 1*idchar
label = "CN" / "L" / "ST" / "O" / "OU" / "C" / "STREET"
oid = 1*DIGIT *("." 1*DIGIT)
<key>:<value> are the subject name fields in chain[0].subject in any order. Key repetitions are not allowed. Values must be percent-encoded.¶
Example:¶
did:x509:0:sha256:WE..jk::subject:C:US:ST:Texas:L:Austin:O:Example¶
Rego policy:¶
validate_predicate(name, value) := true if {
name == "subject"
items := split(value, ":")
count(items) % 2 == 0
subject := {k: v |
some i
i % 2 == 0
k := items[i]
v := urlquery.decode(items[i+1])
}
count(subject) >= 1
object.subset(input.chain[0].subject, subject) == true
}
predicate-name = "san" predicate-value = san-type ":" san-value san-type = "email" / "dns" / "uri" san-value = 1*idchar
san-type is the SAN type and must be one of email, dns, or uri. Note that dn is not supported.¶
san-value is the SAN value, percent-encoded.¶
The pair [<san_type>, <san_value>] is one of the items in chain[0].extensions.san.¶
Example:¶
did:x509:0:sha256:WE..jk::san:email:bob%40example.com¶
Rego policy:¶
validate_predicate(name, value) := true if {
name == "san"
[san_type, san_value_encoded] := split(value, ":")
san_value := urlquery.decode(san_value_encoded)
[san_type, san_value] == input.chain[0].extensions.san[_]
}
predicate-name = "eku"
predicate-value = eku
eku = oid
oid = 1*DIGIT *("." 1*DIGIT)
eku is one of the OIDs within chain[0].extensions.eku.¶
Example:¶
did:x509:0:sha256:WE..jk::eku:1.3.6.1.4.1.311.10.3.13¶
Rego policy:¶
validate_predicate(name, value) := true if {
name == "eku"
value == input.chain[0].extensions.eku[_]
}
predicate-name = "fulcio-issuer" predicate-value = fulcio-issuer fulcio-issuer = 1*idchar
fulcio-issuer is chain[0].extensions.fulcio_issuer, without leading https://, percent-encoded.¶
Example:¶
did:x509:0:sha256:WE..jk::fulcio-issuer:accounts.google.com::san:email:bob%40example.com¶
Example 2:¶
did:x509:0:sha256:WE..jk::fulcio-issuer:issuer.example.com::san:uri:https%3A%2F%2Fexample.com%2Focto-org%2Focto-automation%2Fworkflows%2Foidc.yml%40refs%2Fheads%2Fmain¶
Rego policy:¶
=============== NOTE: '\' line wrapping per RFC 8792 ================
validate_predicate(name, value) := true if {
name == "fulcio-issuer"
suffix := urlquery.decode(value)
concat("", ["https://", suffix]) == input.chain[0].extensions.\
fulcio_issuer
}
This illustrates what a typical DID document ([DID-DOCUMENT]), describing the DID subject and the methods it can use to authenticate itself, can look like once resolved:¶
{
"@context": "https://www.w3.org/ns/did/v1",
"id": "did:x509:0:sha256:hH..GE::subject:CN:Example",
"verificationMethod": [
{
"id": "did:x509:0:sha256:hH..GE::subject:CN:Example#key-1",
"type": "JsonWebKey2020",
"controller": "did:x509:0:sha256:hH..GE::subject:CN:Example",
"publicKeyJwk": {
"kty": "RSA",
"n": "s9..WQ",
"e": "AQAB"
}
}
],
"assertionMethod": [
"did:x509:0:sha256:hH..GE::subject:CN:Example#key-1"
],
"keyAgreement": [
"did:x509:0:sha256:hH..GE::subject:CN:Example#key-1"
]
}
CertificateChain = [2*Certificate] ; leaf is first
Certificate = {
fingerprint: {
; base64url-encoded hashes of the DER-encoded certificate
sha256: base64url, ; FIPS 180-4, SHA-256
sha384: base64url, ; FIPS 180-4, SHA-384
sha512: base64url ; FIPS 180-4, SHA-512
},
issuer: Name, ; RFC 5280, Section 4.1.2.4
subject: Name, ; RFC 5280, Section 4.1.2.6
extensions: {
? eku: [+OID], ; RFC 5280, Section 4.2.1.12
? san: [+SAN], ; RFC 5280, Section 4.2.1.6
? fulcio_issuer: tstr
; http://oid-info.com/get/1.3.6.1.4.1.57264.1.1
}
}
; X.509 Name as an object of attributes
; Repeated attribute types are not supported
; Common attribute types have human-readable labels (see below)
; Other attribute types use dotted OIDs
; Values are converted to UTF-8
Name = {
; See RFC 4514, Section 3, for meaning of common attribute types
? CN: tstr,
? L: tstr,
? ST: tstr,
? O: tstr,
? OU: tstr,
? C: tstr,
? STREET: tstr,
* OID => tstr
}
; base64url-encoded data, see RFC 4648, Section 5
base64url = tstr
; ASN.1 Object Identifier
; Dotted string, for example "1.2.3"
OID = tstr
; X.509 Subject Alternative Name
; Strings are converted to UTF-8
SAN = rfc822Name / DNSName / URI / DirectoryName
rfc822Name = ["email", tstr] ; Example: ["email", "user@example.com"]
DNSName = ["dns", tstr] ; Example: ["dns", "example.com"]
URI = ["uri", tstr] ; Example: ["uri", "https://example.com"]
DirectoryName = ["dn", Name] ; Example: ["dn", {CN: "Example"}]
This DID method maps characteristics of X.509 certificate chains to identifiers. It allows a single identifier to map to multiple certificate chains, giving the identifier stability across the expiry of individual chains. However, if the policies used in the identifier are chosen too loosely, the identifier may match too wide a set of certificate chains. This may have security implications as it may authorize an identity for actions it was not meant to be authorized for.¶
To mitigate this issue, the certificate authority should publish their expected usage of certificate fields and indicate which ones constitute a unique identity, versus any additional fields that may be of an informational nature. This will help users create an appropriate did:x509 as well as consumers of signed content to decide whether it is appropriate to trust a given did:x509.¶
Typically, a verifier trusts an X.509 certificate by applying chain validation defined in Section 6 of [RFC5280] using a set of certificate authority (CA) certificates as trust store, together with additional application-specific policies.¶
This DID method does not require an X.509 trust anchor store but rather relies on verifiers either trusting an individual DID directly or using third-party endorsements for a given DID, like [VC], to establish trust.¶
By layering this DID method on top of X.509, verifiers are free to use traditional chain validation (for example, verifiers unaware of DID), or rely on DID as an ecosystem to establish trust.¶
While it is acceptable to use a did:x509 identifier as an opaque handle when it has been endorsed through an external trust mechanism, such as a verifiable credential or a trusted registry, implementers MUST NOT parse or interpret individual components of the identifier string for authorization decisions unless the identifier has been resolved against a verified certificate chain.¶
Specifically, extracting and relying upon subject names, organizational information, or other embedded values directly from the identifier string, without performing full resolution and chain validation, is insecure. An attacker could craft a syntactically valid did:x509 identifier containing arbitrary values that do not correspond to any legitimate certificate chain. Only after successful resolution, which includes verification of the CA fingerprint against the provided chain and validation of all policy predicates, can the identifier be considered authentic. Systems that bypass this resolution process and instead parse identifier components directly are vulnerable to impersonation and privilege escalation attacks.¶
RFC Editor: Please replace "RFCthis" with the RFC number assigned to this document.¶
RFC Editor: Some considerations¶
The authors would like to thank list for their reviews and suggestions.¶