| Internet-Draft | TLS Trust Anchor Identifiers | September 2026 |
| Beck, et al. | Expires 16 March 2027 | [Page] |
This document defines the TLS Trust Anchors extension, a mechanism for a TLS client or server to select a certificate to present based on the peer's trusted certification authorities. It describes certification authorities more succinctly than the TLS Certificate Authorities extension.¶
This note is to be removed before publishing as an RFC.¶
The latest revision of this draft can be found at https://tlswg.github.io/tls-trust-anchor-ids/draft-ietf-tls-trust-anchor-ids.html. Status information for this document may be found at https://datatracker.ietf.org/doc/draft-ietf-tls-trust-anchor-ids/.¶
Discussion of this document takes place on the Transport Layer Security Working Group mailing list (mailto:tls@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/tls/. Subscribe at https://www.ietf.org/mailman/listinfo/tls/.¶
Source for this draft and an issue tracker can be found at https://github.com/tlswg/tls-trust-anchor-ids.¶
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 16 March 2027.¶
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.¶
TLS [RFC9846] authentication uses X.509 certificates [RFC5280] to associate the authenticating party's TLS key with its application identifiers, such as DNS names. These associations are signed by some certification authority (CA). The peer, or relying party, curates a set of CAs that are trusted to only sign correct associations, which allows it to rely on the TLS to authenticate application identifiers. For a TLS server certificate, the authenticating party is the server and the relying party is the client. For a TLS client certificate, the roles are reversed.¶
An authenticating party may need to interoperate with relying parties that trust different sets of CAs. Section 4.3.4 of [RFC9846] defines the certificate_authorities extension to accommodate this. It allows the authenticating party to provision multiple certificates and select the one that will allow the relying party to accept its TLS key. This is analogous to parameter negotiation elsewhere in TLS.¶
Without a negotiation mechanism, the authenticating party must obtain a single certificate that simultaneously satisfies all relying parties. This is challenging when relying parties are diverse. PKI transitions, including those necessary for user security, naturally lead to relying party diversity, so the result is that service availability conflicts with security and overall PKI evolution:¶
For an authenticating party to use a CA in its single certificate, all supported relying parties must trust the CA. PKI transitions then become difficult when authenticating parties support older, unupdated relying parties. This impacts both new keys from existing CA operators and new CA operators.¶
When a relying party must remove no longer trustworthy CAs, rotate CA keys, add new CAs, or otherwise update to meet new security requirements, it will differ from older versions and potentially other relying parties. This adds to relying party diversity and the challenges that authenticating parties and CAs face. The relying party must then choose between compromising on user security or burdening the rest of the ecosystem, potentially impacting availability in the process.¶
However, certificate_authorities's size is impractical for some applications. Existing PKIs may have many CAs, and existing CAs may have long X.509 names. As of August 2023, the Mozilla CA Certificate Program [MOZILLA-ROOTS] contained 144 CAs, with an average name length of around 100 bytes. Such TLS deployments often do not use trust anchor negotiation at all.¶
To address this, this document introduces Trust Anchor Identifiers (Trust Anchor IDs). There are several parts to this mechanism:¶
Section 4 defines trust anchor IDs, which are short, unique identifiers for X.509 trust anchors, or groups of trust anchors.¶
Section 5 defines a TLS extension that communicates the relying party's requested trust anchors using trust anchor IDs. IDs that represent individual trust anchors can mitigate long X.509 names. IDs that represent groups of trust anchors can mitigate large trust anchor lists.¶
Section 5.6 defines a recovery mechanism that, when the relying party is a TLS client, can mitigate signaling failures. The server provides its available trust anchors alongside its certificate, so that the client can retry on mismatch. This can further mitigate large trust anchor lists by allowing the client to initially omit some trust anchors or use an otherwise too broad trust anchor group. However, this mitigation can come at the cost of additional round trips in some cases.¶
Together, they reduce the size costs of trust anchor negotiation, supporting flexible and robust PKIs for more applications.¶
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.¶
This document additionally uses the TLS presentation language, defined in Section 3 of [RFC9846], and ASN.1, defined in [X680].¶
This document discusses three roles:¶
The party authenticating itself in the protocol. In TLS, this is the side sending the Certificate and CertificateVerify message.¶
The party whom the authenticating party presents its identity to. In TLS, this is the side that validates a Certificate and CertificateVerify message.¶
The service issuing certificates to the authenticating party.¶
Additionally, there are several terms used throughout this document to describe this proposal:¶
A pre-distributed X.509 name and public key that relying parties use to determine whether a certification path is trusted. See Section 6.1.1 of [RFC5280]. Trust anchors are sometimes configured as self-signed certificates.¶
An ordered list of X.509 certificates starting with the target certificate. Each certificate is issued by the next certificate, except the last, which is issued by a trust anchor.¶
TLS certificate selection (see Section 4.5.1.2 of [RFC9846]) combines information from both the authenticating and relying party:¶
The authenticating party is configured with one or more candidate certification paths.¶
The relying party is configured with one or more supported trust anchors.¶
In the TLS handshake, the relying party sends a ClientHello or CertificateRequest message that describes its preferences.¶
Based on this information, the authenticating party selects the best candidate certification path to present.¶
To successfully complete the handshake, the authenticating party must select some path that both:¶
is issued by one of the relying party's trust anchors and¶
satisfies any other constraints in the TLS protocol, such as whether there is a common signature algorithm¶
This document defines a mechanism to evaluate the first condition. In particular, it defines:¶
How the relying party describes its supported trust anchors (Section 5.2)¶
How the authenticating party interprets this description to inform certificate selection (Section 5.3)¶
For server certificates, a recovery mechanism for signaling failure (Section 5.6)¶
A trust anchor ID is a short, unique identifier that represents a trust anchor or a group of trust anchors. When a trust anchor ID represents a group of trust anchors, it is known as a trust anchor group.¶
A trust anchor ID is an object identifier (OID) [X680] under the OID arc of some IANA-registered Private Enterprise Number (PEN) [RFC9371]. For compactness, they are represented as relative object identifiers (see Section 33 of [X680]), relative to the OID prefix 1.3.6.1.4.1. For example, an organization with PEN 32473 might define a trust anchor ID with the OID 1.3.6.1.4.1.32473.1. As a relative object identifier, it would be the OID 32473.1.¶
Depending on the protocol, trust anchor IDs may be represented in one of three ways:¶
For use in ASN.1-based protocols, a trust anchor ID's ASN.1 representation is the relative object identifier described above. This may be encoded in DER [X690], or some other ASN.1 encoding. The example ID's DER encoding is the six-octet sequence {0x0d, 0x04, 0x81, 0xfd, 0x59, 0x01}.¶
For use in binary protocols such as TLS, a trust anchor ID's binary representation consists of the contents octets of the relative object identifier's DER encoding, as described in Section 8.20 of [X690]. Note this omits the tag and length portion of the encoding. The example ID's binary representation is the four-octet sequence {0x81, 0xfd, 0x59, 0x01}.¶
For use in ASCII-compatible text protocols, a trust anchor ID's ASCII representation is the relative object identifier in dotted decimal notation. The example ID's ASCII representation is 32473.1.¶
The length of a trust anchor ID's binary representation MUST NOT exceed 255 bytes. It SHOULD be significantly shorter, for bandwidth efficiency.¶
A trust anchor ID representing a single trust anchor SHOULD be allocated by the CA operator and be common among relying parties that trust the CA. They MAY be allocated by another party, e.g. when bootstrapping an existing ecosystem, if all parties agree on the ID. In particular, the protocol requires authenticating and relying parties to agree, and the authenticating party's configuration typically comes from the CA.¶
A trust anchor ID representing a trust anchor group MAY be allocated by any party. However, to be useful, the group requires agreement between relying parties and authenticating parties. Section 6 discusses defining trust anchor groups in more detail.¶
When embedded in a TLS structure, a trust anchor ID uses the TrustAnchorID structure defined below. The contents of the TrustAnchorID, after the one-byte length prefix, are the binary representation of the trust anchor ID.¶
opaque TrustAnchorID<1..2^8-1>;¶
The trust_anchors extension is defined using the structures below:¶
enum { trust_anchors(TBD), (2^16-1) } ExtensionType;
/* Syntax when sent in ClientHello or CertificateRequest: */
TrustAnchorID RequestedTrustAnchorList<0..2^16-1>;
/* Syntax when sent in Certificate: */
struct {} Empty;
/* Syntax when sent in EncryptedExtensions: */
TrustAnchorID AvailableTrustAnchorList<1..2^16-1>;
¶
A TrustAnchorID structure contains the binary representation of some trust anchor ID, as described in Section 4.¶
When the trust_anchors extension is sent in ClientHello or CertificateRequest, the extension_data is a RequestedTrustAnchorList. It indicates that the sender supports the specified trust anchors or trust anchor groups. The list is unordered, and MAY be empty. Section 5.2 describes how the relying party determines this value. Section 5.3 describes how the authenticating party evaluates this value.¶
When the trust_anchors extension is sent in Certificate, the extension_data MUST be empty. The extension MUST only be sent in the first CertificateEntry. It indicates that the sender sent the certificate because the certificate matched a trust anchor ID sent by the peer. Section 5.5 describes this in detail.¶
When the trust_anchors extension is sent in EncryptedExtensions, the extension_data is an AvailableTrustAnchorList. It indicates individual trust anchors for which the server has a candidate path, in order of most to least preferred by the server. This list MUST NOT be empty. If the server has no available trust anchors to present, it MUST omit the extension. Section 5.6 describes this in detail.¶
Relying parties are configured with:¶
An associated trust anchor ID for each supported trust anchor that participates in this protocol¶
A list of requested trust anchor IDs which, together, describe supported trust anchors¶
A trust anchor's associated ID MUST be the trust anchor ID which represents it. In this document, the ID is expected to be configured separately from the trust anchor for compatibility with existing PKIs. Future certificate profiles MAY define representations where the trust anchor ID is encoded directly in the trust anchor.¶
Relying parties MAY support trust anchors without associated trust anchor IDs, but such trust anchors will not participate in this protocol. Those trust anchors MAY participate in other trust anchor negotiation protocols, such as the certificate_authorities extension.¶
In a TLS connection, the relying party sends its requested trust anchor IDs in the ClientHello message (if a client) or CertificateRequest message (if a server). This communicates a set of supported trust anchors to the authenticating party.¶
The requested trust anchor IDs MAY be determined by collecting the associated IDs of each supported trust anchor. Alternatively, a relying party MAY configure a requested list of IDs for individual trust anchors and IDs for trust anchor groups. Using groups can further reduce the size of messages sent by the relying party, but requires that authenticating parties be configured to recognize them. See also Section 5.3.¶
If the relying party is a client, it is not necessary for the requested trust anchor IDs to be fully accurate. A client MAY omit trust anchors that it trusts or signal trust anchors which it does not trust. This can be useful in several scenarios:¶
The client MAY try to reduce size with a common trust anchor group, but the group contains some untrusted trust anchors. Sending the group would signal the full contents of the group.¶
The client MAY send a (possibly empty) subset of its trust anchors due to fingerprinting risks (see Section 9) or size concerns.¶
The client MAY send trust anchors it does not trust. This can reduce fingerprinting if, e.g., default instances of the client send this value, but an individual user has configured their software to distrust the CA.¶
If the client list is inaccurate, it is possible the server will select an untrusted certificate. The connection will then fail. Clients that send potentially inaccurate lists SHOULD implement the recovery mechanism described in Section 5.6. The associated IDs of individual trust anchors are used in recovery. Recovery requires a round-trip, so clients SHOULD send as accurate a list as feasible.¶
The authenticating party compares the requested trust anchor IDs with its candidate certification paths. To do this, each candidate certification path that participates in this protocol MUST be configured with:¶
The trust anchor ID for the CA that issued this candidate path.¶
The trust anchor groups known to contain the issuing CA. The CA can be contained in a family of related trust anchor groups, such as in Section 6.1. To accomodate this, the IDs of the containing groups are described with a list of trust anchor ID patterns, defined below in Section 5.3.1. Note these patterns specify the IDs of the groups, not their contents.¶
Section 7 defines a format to represent these properties. Section 7.5 defines how to obtain them from ACME [RFC8555].¶
The authenticating party intersects this information with the requested trust anchor IDs to determine if the relying party trusts the issuing CA. A candidate path is said to match the requested trust anchor IDs if either:¶
One of the requested trust anchor IDs is equal to the path's trust anchor ID.¶
One of the requested trust anchor IDs is contained in one of the path's trust anchor group patterns.¶
Authenticating parties MAY have candidate certification paths that do not participate in this protocol and lack these properties. These paths MAY participate in other trust anchor negotiation protocols, such as the certificate_authorities extension, or they MAY be used as a fallback when no matching issuer is found.¶
A trust anchor ID pattern specifies a collection of related IDs. In this document, the IDs matched by a pattern are always the IDs of trust anchor groups. It is a sequence of pairs min and max. min is a non-negative integer and max is either a non-negative integer or infinity. A pattern is said to contain some trust anchor ID if both of the following are true:¶
The number of components of the trust anchor ID, as a relative OID, is equal to the number of pairs in the pattern.¶
Each component of the trust anchor ID, as a relative OID, is between min and max, inclusive, of the corresponding pair in the pattern.¶
A trust anchor ID pattern is represented as a byte string by concatenating the min and max values of each pair, in order. Each min or max value is encoded as follows:¶
Infinity is encoded as a single byte, 0x80.¶
A non-negative integer is encoded as described in paragraph 8.19.2 of [X690]. That is, each value is encoded in variable-length, big-endian, base-128 encoding. Each base-128 digit is in the seven least significant bits of each byte. The most significant bit of each byte is unset for the final byte and set for all other bytes. Values are encoded in the fewest number of non-zero bytes needed.¶
A trust anchor ID pattern can be represented in text as follows:¶
Represent each min and max pair as:¶
Concatenate the representations of each pair, separating each by ".".¶
The byte string representation of an OID component is order-preserving by length and then lexicographic comparison, so the following procedures can be used to check if an ID is contained in the pattern:¶
To remove an encoded base-128 integer from a byte string, in:¶
If in is empty, fail the procedure. There are no more values in in.¶
If the first byte of in is 0x80, fail the procedure. The value was not minimally encoded.¶
Find the earliest byte of in whose most-significant bit is unuset.¶
If not found, fail the procedure. The value was truncated.¶
Remove and return the prefix of in which ends at the found byte.¶
To compare two encoded base-128 integers, a and b:¶
Compare a's length to b's length. If they are not equal, return the result of the comparison.¶
Return the result of lexicographically comparing a and b. Bytes in a and b are intepreted as integers from 0 to 255.¶
To check if a trust anchor ID pattern, pattern, contains a trust anchor ID id, both in their byte representations:¶
While id is not empty:¶
Remove an encoded base-128 integer from id. Let v be the value removed.¶
Remove an encoded base-128 integer from pattern. Let min be the value removed.¶
Compare v and min as described above. If v is less than min, fail the procedure.¶
If pattern is not empty and the next byte of pattern is 0x80, remove this byte and continue to the next loop iteration.¶
Otherwise, remove an encoded base-128 integer from pattern. Let max be the value removed.¶
Compare max and v as described above. If max is less than v, fail the procedure.¶
If pattern is not empty, fail the procedure. Otherwise, the procedure succeeds.¶
For example, 32473.{123-456}.{789-} is a pattern that matches three-component IDs, where the first component must be 32473, the second must be between 123 and 456, and the final component must be at least 789. The byte string representation is:¶
// component[0].min = 32473 0x81, 0xfd, 0x59, // component[0].max = 32473 0x81, 0xfd, 0x59, // component[1].min = 123 0x7b, // component[1].max = 456 0x83, 0x48, // component[2].min = 789 0x86, 0x15, // component[2].max = infinity 0x80,¶
It contains the following IDs:¶
It does not contain any of the following IDs:¶
32473.123 (too few components)¶
32473.123.789.0 (too many components)¶
32474.123.789 (first component out of range)¶
32473.500.789 (second component out of range)¶
32473.123.700 (third component out of range)¶
Appendix A provides more extensive test vectors.¶
This document extends TLS certificate selection (Section 4.5.1.2 of [RFC9846]) as follows:¶
If the ClientHello or CertificateRequest contains a trust_anchors extension, the authenticating party SHOULD send a certification path that matches the requested trust anchor IDs, as described in Section 5.3. See Section 5.5 for additional requirements in this case.¶
If the ClientHello or CertificateRequest contains both trust_anchors and certificate_authorities, certification paths that satisfy either extension's criteria MAY be used. This additionally applies to future extensions which play a similar role.¶
If no certification paths satisfy either extension, the authenticating party MAY return a handshake_failure alert, or send some fallback certificate, without considering trust_anchors or certificate_authorities.¶
Sending a fallback allows the authenticating party to retain support for relying parties that do not implement any form of trust anchor negotiation. In this case, the authenticating party must find a sufficiently ubiquitous trust anchor, if one exists. However, only those relying parties need to be considered in this ubiquity determination. Updated relying parties may continue to evolve without restricting fallback certificate selection. Section 7.3 describes a RECOMMENDED mechanism for determining fallbacks.¶
When the authenticating party is a server, Section 5.6 describes an additional requirement for servers that implement this protocol.¶
If, and only if, the authenticating party sends a certification path that matches the relying party's trust_anchors extension, the authenticating party MUST send an empty trust_anchors extension in the first CertificateEntry of the Certificate message.¶
In this case, the certificate_list flexibility described in Section 4.5.1 of [RFC9846] no longer applies. The certificate_list MUST contain a complete certification path, correctly ordered and with no extraneous certificates. That is, each certificate MUST certify the one immediately preceding it, and the path's trust anchor MUST certify the final certificate.¶
If a relying party receives this extension in the Certificate message, it MAY choose to disable path building [RFC4158] and validate the peer's certificate list as a pre-built certification path. Doing so avoids the unpredictable behavior of path-building, and helps ensure CAs and authenticating parties do not inadvertently provision incorrect paths.¶
If the relying party is a client, it MAY, as described in Section 5.2, request extra trust anchors or omit trusted ones. To accommodate this, this section defines a protocol for recovering from signaling failure in server certificate selection.¶
When receiving a ClientHello with trust_anchors, the server collects all candidate certification paths which:¶
Have a trust anchor ID, and¶
Satisfy the conditions in Section 4.5.1.2 of [RFC9846], with the exception of certificate_authorities, and any future extensions that play a similar role¶
If this collection is non-empty, the server MUST send a trust_anchors extension in EncryptedExtensions, containing the corresponding trust anchor IDs in preference order.¶
If a client requests extra trust anchors or omits trusted ones, it SHOULD implement the following recovery mechanism:¶
If the client receives either a connection error or an untrusted certificate, the client looks in the server's EncryptedExtensions for a trust anchor ID that it trusts. If there are multiple, it selects an option based on the server's preference order and its local preferences. It then makes a new connection to the same endpoint, requesting only the selected trust anchor ID in the ClientHello trust_anchors extension. If the EncryptedExtensions had no trust_anchors extension, or no match was found, the client returns the error to the application.¶
Clients SHOULD retry at most once per connection attempt.¶
This mechanism allows the connection to recover from a certificate selection failure, at additional latency cost.¶
This mechanism also allows servers to safely send fallback certificates that may not be as ubiquitously acceptable. Without some form of trust anchor negotiation, servers are limited to selecting certification paths that are ubiquitously trusted in all supported clients. This often means sending extra cross-certificates to target the lowest common denominator at a bandwidth cost. If the ClientHello contains trust_anchors, the server MAY opportunistically send a less ubiquitous, more bandwidth-efficient path based on local heuristics, with the expectation that the client will retry when the heuristics fail.¶
A trust anchor ID is typically much smaller than the corresponding X.509 name. Depending on the number of trust anchors, this can be sufficient to efficiently represent relying party state.¶
PKIs where further size savings are needed can use trust anchor groups (Section 4). Trust anchor groups require additional coordination within a PKI, but they can further reduce relying party message sizes by allowing one ID to signal multiple trust anchors. To be usable, a trust anchor group must:¶
be known to and sent by relying parties (see Section 5.2); and¶
configured with candidate paths in authenticating parties (see Section 5.3), ideally provided by the CA during issuance (see Section 7).¶
This document does not prescribe how to define trust anchor groups, but gives some general guidance:¶
A trust anchor group specifies a collection of trust anchors, which a relying party can send to represent the contents. For example:¶
A set of root CAs (or intermediate CAs, as in Section 8.5) operated by a CA operator.¶
A set of trust anchors common to large set of relying parties.¶
A set of related application-specific trust anchors, such as a range of Merkle Tree Certificate landmarks [I-D.ietf-plants-merkle-tree-certs].¶
Different group definitions trade off size savings, applicability, and coordination overhead. A group that reflects a single CA operator will cover fewer trust anchors, so a relying party might combine several operators' IDs to describe its trust anchors. However, it is generally usable by relying parties that trust this CA operator. Such a group also requires minimal coordination for the CA operator to provide group inclusion information (Section 5.3) with the certificate.¶
Conversely, a group that reflects a single relying party vendor can potentially be the only ID sent. However, it may be less generally usable when relying parties differ. Groups reflecting multiple relying party vendors are more broadly usable, but may need to be combined with other IDs in a given relying party. For example, a relying party might send a group containing established CAs common to its ecosystem, and individual IDs for its remaining, not yet as common CAs.¶
A client relying party MAY send a group containing CAs it does not trust, however it SHOULD then be prepared to recover (see Section 5.6) in case of signaling failure.¶
The matching process described in Section 5.3 can be implemented generically for any trust anchor group. This allows deployments to tailor their group allocation based on their needs, without requiring software updates in authenticating parties. Where feasible, deployments SHOULD use groups that are more broadly applicable and require lower coordination overhead.¶
Over time, a group may become out-of-date, making it describe current relying parties less effectively. For example, a CA operator may deploy or turn down a CA instance, or a relying party may trust a new CA or distrust an existing CA. Existing trust anchor groups SHOULD NOT be redefined, but the following versioning scheme MAY be used to define updated groups:¶
A versioned sequence of trust anchor groups is identified by a OID arc. Each group has an ID of this OID arc, with a non-negative integer version number component appended. For example, versioned groups using the OID arc 32473.2 would have IDs 32473.2.0, 32473.2.1, 32473.2.2, and so on. When defining a new group version, the version component is incremented.¶
Each candidate path is then configured with the versioned groups that contain it. These groups are described by a trust anchor ID pattern (Section 5.3.1) as follows:¶
Let base be the OID arc that identifies the sequence. Let min be the first version that includes the trust anchor.¶
At issuance, if the trust anchor is no longer in the latest group version, let max be the last version that includes the trust anchor. The pattern is base.{min-max}.¶
At issuance, if the trust anchor is in the latest group version, the pattern is base.{min-}. That is, the last component has a max of 264-1.¶
In the second case, the range contains not-yet-defined group versions, so there is a potential signaling error. Suppose, after issuance, a new group version is defined without the trust anchor. The unlimited upper bound is now incorrect. A relying party might not trust this trust anchor, while sending this new group version. However, the authenticating party will misinterpret the certificate as compatible based on its stale information. Such signaling errors may result in the wrong certificate being selected.¶
This can be mitigated in one several ways:¶
Only pre-existing certificates are impacted. Newly-issued certificates postdate this version and will have the correct upper bound. When the certificate is renewed, group inclusions will be corrected.¶
[SCTNotAfter] describes a trust anchor removal strategy that only impacts newly-issued certificates. In this case, no renewal is needed. Pre-existing group inclusions remain accurate under this strategy.¶
If the authenticating party's preferences place the correct candidate path (issued by a newer trust anchor) ahead of misinterpreted one (issued by the removed trust anchor), the correct candidate will still be chosen.¶
When the relying party is a client, any remaining signaling errors can be corrected with the recovery mechanism described in Section 5.6.¶
As described in Section 5.3, certification paths participating in this mechanism must be configured with a trust anchor ID. This section introduces a RECOMMENDED extensible CertificatePropertyList structure for representing this and other additional properties of a certification path. CertificatePropertyLists may be used as part of authenticating party configuration, and for CAs to communicate additional properties during certificate issuance.¶
The extensibility aims to simplify application deployment as PKI mechanisms evolve. When certificate issuance and application software is updated to pass this structure to the underlying TLS implementation, new properties may be transparently defined without changes to certificate and configuration management.¶
A CertificatePropertyList is defined using the TLS presentation language (Section 3 of [RFC9846]) below:¶
enum {
trust_anchor_id(0),
trust_anchor_groups(1),
trust_anchor_negotiation(2),
(2^16-1)
} CertificatePropertyType;
struct {
CertificatePropertyType type;
opaque data<0..2^16-1>;
} CertificateProperty;
CertificateProperty CertificatePropertyList<0..2^16-1>;
¶
The entries in a CertificatePropertyList MUST be sorted numerically by type and MUST NOT contain values with a duplicate type. Inputs that do not satisfy these invariants are syntax errors and MUST be rejected by parsers.¶
This document defines three properties:¶
trust_anchor_id, defined in Section 7.1¶
trust_anchor_groups, defined in Section 7.2¶
trust_anchor_negotiation, defined in Section 7.3¶
Future documents MAY define other properties for use with other mechanisms. Such a document MUST define the format of the data field and how authenticating parties interpret the property. Authenticating parties MUST ignore properties with unrecognized CertificatePropertyType values.¶
The trust_anchor_id property's data field contains the binary representation of the trust anchor ID of the certification path's trust anchor, as described in Section 5.3. The binary representation is encoded directly into the data field with no additional length prefix.¶
The trust_anchor_groups property's data field contains a TrustAnchorIDPatternList structure, defined below. Its value is the certification path's trust anchor group patterns, as described in Section 5.3 and Section 5.3.1.¶
opaque TrustAnchorIDPattern<0..2^8-1>; TrustAnchorIDPattern TrustAnchorIDPatternList<1..2^16-1>;¶
The trust_anchor_negotiation property's data field MUST be empty.¶
When a candidate certification path has this property, the authenticating party SHOULD NOT select it as a fallback when the path's issuer cannot be matched against the relying party. When a candidate path lacks this property, the authenticating party MAY use it as a fallback. See also Section 5.4.¶
A path without the trust_anchor_negotiation property MAY still participate in this protocol and include the trust_anchor_id and trust_anchor_groups properties. In particular, the authenticating party MAY still choose to condition the path on trust anchor negotiation if it is combining multiple sets of candidate paths, each with their separate determinations about suitable fallbacks. Section 7.5.1 gives an example scenario. This could be implemented either with separate local configuration or by modifying the CertificatePropertyList structures when combining the sets.¶
Section 7.5 discusses how an ACME server might set this property, as well as examples where the authenticating party might override this recommendation.¶
A certification path with its associated CertificatePropertyList may be represented in a PEM [RFC7468] structure in a file of type "application/pem-certificate-chain-with-properties". Files of this type MUST use the strict encoding and MUST NOT include explanatory text. The ABNF [RFC5234] for this format is as follows, where "stricttextualmsg" is as defined in Section 3 of [RFC7468]:¶
certchainwithproperties = 2*stricttextualmsg¶
The first element MUST be the encoded CertificatePropertyList. The second element MUST be an end-entity certificate. Each following element MUST contain a certificate that directly certifies the one preceding it. The certificate representing the trust anchor MUST be omitted from the path.¶
CertificatePropertyLists are encoded using the "CERTIFICATE PROPERTIES" label. The encoded data is a serialized CertificatePropertyList, defined in Section 7.¶
Certificates are encoded as in Section 5.1 of [RFC7468], except DER [X690] MUST be used.¶
The following is an example file with a certification path containing an end-entity certificate and an intermediate certificate. The example CertificatePropertyList encodes:¶
A trust_anchor_id property of 32473.1¶
A trust_anchor_groups property with two patterns:¶
A trust_anchor_negotiation property¶
-----BEGIN CERTIFICATE PROPERTIES----- ACoAAAAEgf1ZAQABABoAGAmRC5ELAgJkgUgNgf1Zgf1ZAwMqgGSBSAACAAA= -----END CERTIFICATE PROPERTIES----- -----BEGIN CERTIFICATE----- MIIBVzCB/6ADAgECAgkAh7Uv5X8pplkwCgYIKoZIzj0EAwIwGjEYMBYGA1UEAwwP SW50ZXJtZWRpYXRlIENBMB4XDTI2MDUwNTIxMzg1NVoXDTI3MDUwNTIxMzg1NVow FjEUMBIGA1UEAwwLZXhhbXBsZS5jb20wWTATBgcqhkjOPQIBBggqhkjOPQMBBwNC AAT5mg5z0464cE7rtEpTeSPFNlRUBjxqycdb4rvNkG3Fbd1R2IRo7zYOi5SP3S7L C4r5Hw+IiDq5X2nQT1w5ympeozIwMDAJBgNVHRMEAjAAMAsGA1UdDwQEAwIHgDAW BgNVHREEDzANggtleGFtcGxlLmNvbTAKBggqhkjOPQQDAgNHADBEAiBRdPrVpQtJ s+J9DFhT1Db6QmIZFfjFFKQ88B0gFezyfAIgSwIxntwrPFYagfK6vPcRpDxG2oLV LkfnP5v1SPjOsMY= -----END CERTIFICATE----- -----BEGIN CERTIFICATE----- MIIBRTCB7KADAgECAgkAkaBeQj6ZErAwCgYIKoZIzj0EAwIwEjEQMA4GA1UEAwwH Um9vdCBDQTAeFw0yNjA1MDUyMTM4MzJaFw0zMTA1MDQyMTM4MzJaMBoxGDAWBgNV BAMMD0ludGVybWVkaWF0ZSBDQTBZMBMGByqGSM49AgEGCCqGSM49AwEHA0IABJEH 0D77iyFv01I/4sEqUaoUel50BBwsWSYrH/LtO6cdGI28NyzMyFuYrE6UCRusgAKo XBmWjHEGJmoDPoAy2t+jIzAhMA8GA1UdEwEB/wQFMAMBAf8wDgYDVR0PAQH/BAQD AgEGMAoGCCqGSM49BAMCA0gAMEUCIBWtPiDwXXEvbgy2+nu/w4MRBNsQ3hbVWyJT ITN+1R6WAiEA2AfGBy3Hz8oYY5wPldIndrXjntCzzSEduB6pEvYQZWo= -----END CERTIFICATE-----¶
The IANA registration for this media type is described in Section 11.2.¶
The format defined in Section 7.4 can be used with ACME's alternate format mechanism (see Section 7.4.2 of [RFC8555]) as follows. When downloading certificates, a supporting client SHOULD include "application/pem-certificate-chain-with-properties" in its HTTP Accept header (Section 12.5.1 of [RFC9110]). When a supporting server sees such a header, it MAY then respond with that format to include a CertificatePropertyList with the certification path. This CertificatePropertyList MAY include trust_anchor_id and trust_anchor_groups properties for use with this protocol, or other properties defined in another document.¶
When the ACME server provides multiple paths, e.g. with ACME's alternate certificate chain mechanism (see Section 7.4.2 of [RFC8555]), the ACME server SHOULD include the trust_anchor_negotiation property on any paths it expects to gate on trust anchor negotiation. It SHOULD omit the property on any paths which are possible fallbacks when no trust anchors match.¶
The authenticating party MAY override this recommendation. In particular, if the authenticating party combines certification paths from two ACME orders, it might only consider some orders as a source for fallback paths.¶
When a path is gated on trust anchor negotiation, this protocol removes the need for heuristics in determining which path to serve to which relying party.¶
There are two CA operators, CA1 and CA2. The authenticating party is configured to request certificates from ACME servers operated by each of CA1 and CA2.¶
When the authenticating party requests certificates from CA1, it receives:¶
Path 1A chains to an older root CA operated by CA1. It does not set trust_anchor_negotiation because CA1 considers this to be a reasonable fallback for legacy relying parties.¶
Path 1B chains to a newer root CA operated by CA1. It sets trust_anchor_negotiation because not all relying parties support it yet.¶
When the authenticating party requests certificates from CA2, it receives:¶
Path 2A chains to a root CA operated by CA2. It does not set trust_anchor_negotiation because CA2 considers this to be a reasonable fallback for legacy relying parties.¶
Path 2B chains to a more specific intermediate CA. It sets trust_anchor_negotiation because not all relying parties preload the intermediate.¶
All paths include trust_anchor_id properties describing their corresponding issuer. The authenticating party's TLS software will consider all four in connections that use the trust_anchors extension.¶
For other connections, the TLS software needs to determine fallback paths. Although both 1B and 2B lack the trust_anchor_negotiation property, the authenticating party knows that CA2 is more ubiquitously trusted among its supported relying parties than CA1. It configures its TLS software to use CA2 as the source of the fallback path, and so only path 2B will be used as fallback.¶
While ACME represents each certification path separately, applications might combine multiple certification paths in one file as part of local configuration. For example:¶
An ACME client might serialize all paths returned from a single order in a file. The TLS server might then be configured to load certificates from the files from each order.¶
A deployment might combine the paths from all ACME orders in a single file. The TLS server might then be configured to load its full certificate configuration from the file.¶
This section extends the PEM representation defined in Section 7.4 for such cases.¶
A list of certification paths is represented in PEM by concatenating their corresponding PEM representations. Each path MUST begin with a CertificatePropertyList, which signals a new path to the decoder. If the path has no properties configured, the corresponding PEM-encoded CertificatePropertyList is as follows:¶
-----BEGIN CERTIFICATE PROPERTIES----- AAA= -----END CERTIFICATE PROPERTIES-----¶
Paths are ordered by the encoder's preference, with the most preferred encoded first. Depending on the application, the decoder might use this preference order, or it might override it with another ordering.¶
This format does not directly represent private keys. However, applications MAY combine this format with private keys in one of several ways:¶
If the application represents paths with the same private key, it can associate all decoded paths with the corresponding private key.¶
If the application represents paths with different private keys, it can first load all available private keys, then match each decoded path with the private key that matches the end-entity certificate's subjectPublicKeyInfo.¶
The following example file contains two certification paths:¶
-----BEGIN CERTIFICATE PROPERTIES----- ACoAAAAEgf1ZAQABABoAGAmRC5ELAgJkgUgNgf1Zgf1ZAwMqgGSBSAACAAA= -----END CERTIFICATE PROPERTIES----- -----BEGIN CERTIFICATE----- MIIBVzCB/6ADAgECAgkAh7Uv5X8pplkwCgYIKoZIzj0EAwIwGjEYMBYGA1UEAwwP SW50ZXJtZWRpYXRlIENBMB4XDTI2MDUwNTIxMzg1NVoXDTI3MDUwNTIxMzg1NVow FjEUMBIGA1UEAwwLZXhhbXBsZS5jb20wWTATBgcqhkjOPQIBBggqhkjOPQMBBwNC AAT5mg5z0464cE7rtEpTeSPFNlRUBjxqycdb4rvNkG3Fbd1R2IRo7zYOi5SP3S7L C4r5Hw+IiDq5X2nQT1w5ympeozIwMDAJBgNVHRMEAjAAMAsGA1UdDwQEAwIHgDAW BgNVHREEDzANggtleGFtcGxlLmNvbTAKBggqhkjOPQQDAgNHADBEAiBRdPrVpQtJ s+J9DFhT1Db6QmIZFfjFFKQ88B0gFezyfAIgSwIxntwrPFYagfK6vPcRpDxG2oLV LkfnP5v1SPjOsMY= -----END CERTIFICATE----- -----BEGIN CERTIFICATE----- MIIBRTCB7KADAgECAgkAkaBeQj6ZErAwCgYIKoZIzj0EAwIwEjEQMA4GA1UEAwwH Um9vdCBDQTAeFw0yNjA1MDUyMTM4MzJaFw0zMTA1MDQyMTM4MzJaMBoxGDAWBgNV BAMMD0ludGVybWVkaWF0ZSBDQTBZMBMGByqGSM49AgEGCCqGSM49AwEHA0IABJEH 0D77iyFv01I/4sEqUaoUel50BBwsWSYrH/LtO6cdGI28NyzMyFuYrE6UCRusgAKo XBmWjHEGJmoDPoAy2t+jIzAhMA8GA1UdEwEB/wQFMAMBAf8wDgYDVR0PAQH/BAQD AgEGMAoGCCqGSM49BAMCA0gAMEUCIBWtPiDwXXEvbgy2+nu/w4MRBNsQ3hbVWyJT ITN+1R6WAiEA2AfGBy3Hz8oYY5wPldIndrXjntCzzSEduB6pEvYQZWo= -----END CERTIFICATE----- -----BEGIN CERTIFICATE PROPERTIES----- AAA= -----END CERTIFICATE PROPERTIES----- -----BEGIN CERTIFICATE----- MIIBojCCAUigAwIBAgIBAjAKBggqhkjOPQQDAjAcMRowGAYDVQQDDBFJbnRlcm1l ZGlhdGUgQ0EgMjAeFw0yNjA5MTEyMjA3MzJaFw0yNzA5MTEyMjA3MzJaMBYxFDAS BgNVBAMMC2V4YW1wbGUuY29tMFkwEwYHKoZIzj0CAQYIKoZIzj0DAQcDQgAErvaU F7iXvurpBgeG5eCx8cMmcOKb11Nvlk/dCdcAelIcvAAHABGc8cVSkjGlNGQhgeCm MBKQipIiDskIhIZoRaOBgDB+MB0GA1UdDgQWBBTw8ysZe1gMI/OX0Jx9Y/0yq4Q6 zjAfBgNVHSMEGDAWgBT9JPmvVv2aTEDF/R+XTZzy9iMRWjAPBgNVHRMBAf8EBTAD AQH/MBYGA1UdEQQPMA2CC2V4YW1wbGUuY29tMBMGA1UdJQQMMAoGCCsGAQUFBwMB MAoGCCqGSM49BAMCA0gAMEUCIQCAbiJcNrPnAr0N9oBJ70ikytGQxTQLEfdMF3Id dRHp/QIgNFQIR4pV/CxvnbJnqUYySx7NgynEBj4v9fndOj8+Mvw= -----END CERTIFICATE----- -----BEGIN CERTIFICATE----- MIIBhDCCASugAwIBAgIBATAKBggqhkjOPQQDAjAUMRIwEAYDVQQDDAlSb290IENB IDIwHhcNMjYwOTExMjIwNzMyWhcNMzEwOTEwMjIwNzMyWjAcMRowGAYDVQQDDBFJ bnRlcm1lZGlhdGUgQ0EgMjBZMBMGByqGSM49AgEGCCqGSM49AwEHA0IABAqUU6fc WctyRGFMz3CGQwUCb5pPhi7imSamipwIrQopqOOUqTr27RLa0CSQwL/87OH/Yxc8 1jp3cC3qdEp4sAmjZjBkMB0GA1UdDgQWBBT9JPmvVv2aTEDF/R+XTZzy9iMRWjAf BgNVHSMEGDAWgBSQ1f8odAe5s7NU91kxX2mPDd8xezASBgNVHRMBAf8ECDAGAQH/ AgEAMA4GA1UdDwEB/wQEAwIBBjAKBggqhkjOPQQDAgNHADBEAiA1VrVfvq1QtS5v gZYh1yEIL8wV863GEE2C6/zSB7TzaAIgTUBrpMo56XIb+Wez1CPWtqYFd2a6NvJx IKzgi/++xTs= -----END CERTIFICATE-----¶
trust_anchors, like certificate_authorities, implements trust anchor negotiation. That is, it allows an authenticating party to incorporate relying party trust anchors into certificate selection. trust_anchors allows a wider range of TLS applications to use trust anchor negotiation, notably those that would be unable to use certificate_authorities due to size or privacy limitations.¶
Without trust anchor negotiation, authenticating parties are limited to CAs in the intersection of all supported relying parties. However, trust anchors can vary significantly between different relying party implementations and different versions of a single relying party implementation, particularly as PKIs evolve to meet user security needs.¶
As security-positive PKI changes increase variance, this intersection shrinks. This leads to a conflict between user security and service availability. When the authenticating party cannot serve a certificate in the intersection, either the relying party must risk user security by not changing the PKI, or the authenticating party must degrade service availability by dropping support for some relying parties.¶
The rest of this section discusses uses cases for trust anchor negotiation.¶
When one relying party trusts a new CA, other relying parties, such as older ones, may not yet trust it. Trust anchor negotiation allows an authenticating party to negotiate a certificate from the newer CA with relying parties that do trust it, while continuing to negotiate another certificate with relying parties that do not. This allows PKI transitions to progress smoothly. Connections can make use of, for example, a new CA's stronger signature algorithms, stronger validation practices, better automation, or more efficient certificate sizes, without interruptions to other connections.¶
Without negotiation, the authenticating party is limited to its relying parties' intersection and must wait for every supported relying party to be updated before the transition even begins. This wait could often take many years. In some cases, such as with IoT devices, relying parties may never receive updates.¶
In some contexts, other fields can provide a partial signal. For example, post-quantum-capable relying parties may be detected with the signature_algorithms and signature_algorithms_cert extensions. However, this relies on all post-quantum CAs being added at roughly the same time and that they are sufficiently interchangeable to be negotiated with these extensions. Trust anchor negotiation directly addresses this problem and allows for both gradual and possibly heterogeneous deployment of post-quantum CAs across relying parties.¶
When CAs are determined to be untrustworthy, relying parties must remove them to mitigate the risk to user security. Over time, this shrinks their intersection with older relying parties. Without negotiation, the result is authenticating parties have fewer and fewer CA choices available. Even determining the intersecting CAs can be difficult. Often, the only option is to try the new certificate and monitor errors. For authenticating parties that serve many diverse relying parties, this is a disruptive and risky process.¶
Trust anchor negotiation removes this constraint. If an authenticating party's CA is distrusted, it can use a new CA in addition to the existing one. The addition does not risk outages for older relying parties and may be chosen from a wider set of CAs, as it only needs to be compatible with the relying parties that distrusted the other CA.¶
Over time, the authenticating party can monitor which certificates it serves, and re-evaluate which CA or CAs to use. For example, it may find the new CA was sufficient, or that older relying parties have since all been updated. However, user security depends on the relying party's trust anchors, not the authenticating party's choice of CA, so this can occur asynchronously, based on serving needs and costs, rather than delay the response to a security incident.¶
Despite the severity of root CA private key compromise and the benefits of routinely rotating cryptographic key material, such rotation in PKIs is often very rare. In 2023, the oldest root in [CHROME-ROOTS] and [MOZILLA-ROOTS] was 25 years old, dating to 1998.¶
Key rotation in PKIs used in TLS is challenging, as it combines the challenges described in both Section 8.1 and Section 8.2. Without trust anchor negotiation, authenticating parties cannot switch to the new root as long as any supported older relying party requires the old root. That, in turn, means relying parties cannot distrust the old root, leaving them vulnerable.¶
Trust anchor negotiation offers a smooth transition for CA key rotation. The CA can provide certification paths for the old and new roots. The authenticating party can then serve both paths without impacting older relying parties. New relying parties can then distrust the old root.¶
The mechanisms in this document can aid PKI transitions beyond key rotation. For example, a CA operator may generate a postquantum root CA and issue from the classical and postquantum roots concurrently. The authenticating party will then, transparently and with no configuration change, serve both. As in Section 8.3, newer relying parties can then remove the classical roots, while older relying parties continue to function.¶
This same procedure may also be used to transition between newer, more size-efficient signature algorithms, as they are developed.¶
In many PKIs, root CAs issue shorter-lived intermediate certificates which, in turn, issue end-entity certificates. This comes at a bandwidth cost: the TLS handshake includes an extra certificate, which includes a public key, signature, and X.509 metadata. Post-quantum signature algorithms will dramatically increase this cost. ML-DSA-65 [FIPS204], for example, has a total public key and signature size of 5,261 bytes.¶
Trust anchor negotiation can avoid this size cost. Relying parties predistribute intermediate CAs and configure them as short-lived trust anchors. Authenticating parties can then send shorter paths to those relying parties.¶
More generally, a CA operator provides authenticating parties with two certification paths: a longer path ending at a long-lived root and shorter path the other ending at a short-lived root. Relying parties trust both the long-lived root and the most recent short-lived root. The authenticating party sends the shorter path when possible, falling back to the longer path when the relying party’s short-lived root is stale.¶
An authenticating party may need to support relying parties with different, potentially conflicting requirements. For example, in contexts where online revocation checks are expensive, unreliable, or privacy-sensitive, user security is best served by short-lived certificates. In other contexts, long-lived certificates may be more appropriate for, e.g., systems that are offline for long periods of time or have unreliable clocks.¶
Trust anchor negotiation allows these conflicts to be resolved by different trust anchors where necessary. This avoids the need to compromise on user security or service availability.¶
An authenticating party may obtain certification paths from multiple CAs for redundancy. If one CA is compromised and removed from newer relying parties, the TLS server software will be able to gracefully serve a backup certification path, avoiding the immediate breakage that would otherwise be caused by this removal.¶
To reduce security risk from misissued certificates, relying parties sometimes employ public key pinning [RFC7469]. Pinning effectively reduces a relying party's trust anchor list to a subset of the original set.¶
As other relying parties in the PKI evolve, the pinning relying party limits the authenticating party to satisfy both the pinning constraint and newer constraints in the PKI. This can lead to conflicts if, for example, the pinned CA is distrusted by a newer relying party. The authenticating party is then forced to either break the pinning relying party, or break the newer ones.¶
Trust anchor negotiation reduces this conflict, provided the pinning relying party negotiates with its reduced trust anchor list. The authenticating party can then use a certificate from the pinned CA with the pinning relying party, and another CA with other relying parties.¶
The trust_anchors extension is analogous to the certificate_authorities extension (Section 4.3.4 of [RFC9846]), but more size-efficient. Like certificate_authorities, trust_anchors reveals some information about the relying party's trust anchors. However, unlike certificate_authorities, trust_anchors allows a relying party to only reveal a trust anchor in response to the authenticating party's list, which reduces the fingerprinting exposure. This section provides guidance for a relying party to configure this mechanism, based on its privacy goals.¶
When using this extension, a relying party's trust anchors may be divided into three categories:¶
Trust anchors whose IDs the relying party never sends, but still trusts. These are trust anchors that do not participate in this mechanism.¶
Trust anchors whose IDs the relying party sends conditionally, i.e. only if the server offers them. For example, the relying party may indicate support for a trust anchor if its ID is listed in the server's HTTPS/SVCB record or trust anchor list in EncryptedExtensions.¶
Trust anchors whose IDs the relying party sends unconditionally, i.e. independently of the authenticating party's behavior.¶
Each of these categories carries a different fingerprinting exposure:¶
Trust anchors that do not participate are not revealed by this extension. However, they have some fingerprinting exposure due to being trusted. Given a certification path, an authenticating party can probe whether the relying party trusts the trust anchor by seeing if the relying party accepts it.¶
Trust anchor IDs sent in response to the authenticating party can only be observed actively. That is, the authenticating party could vary its list and observe how the client responds, in order to probe for the client's trust anchor list. This is similar to the exposure of trust anchors not participating in this extension, except that the trust anchor can be probed by only knowing the trust anchor ID.¶
Trust anchor IDs sent unconditionally can be observed passively. This mode is analogous to the certificate_authorities extension. Relying parties SHOULD NOT unconditionally advertise trust anchor lists that are unique to an individual user. Rather, unconditionally-advertised lists SHOULD be empty or computed only from the trust anchors common to the relying party's anonymity set (Section 3.3 of [RFC6973]).¶
Relying parties SHOULD determine which trust anchors participate in this mechanism, and whether to advertise them unconditionally or conditionally, based on their privacy goals.¶
Additionally, a relying party that computes the trust_anchors extension based on prior state may allow observers to correlate across connections. Relying parties SHOULD NOT maintain such state across connections that are intended to be uncorrelated.¶
If the authenticating party is a server, the trust_anchors extension in EncryptedExtensions enumerates the trust anchors for the server's available certification paths. (See Section 5.6.) This assumes these trust anchors are not sensitive. Servers SHOULD NOT use this mechanism to negotiate certification paths with sensitive trust anchors.¶
In servers that host multiple services, this protocol only enumerates certification paths for the requested service. If, for example, a server uses the server_name extension to select services, this list is expected to be filtered by server_name. This ensures that co-located services are not revealed.¶
The above does not apply if the authenticating party is a client. This protocol does not enumerate the available certification paths for a client.¶
If the authenticating party has provisioned certification paths with incorrect trust anchor IDs, it may negotiate inaccurately and send an untrusted path to the relying party when another candidate would have been trusted. This will not result in the untrusted path becoming trusted, but the connection will fail.¶
Both the trust_anchors and certificate_authorities (Section 4.3.4 of [RFC9846]) extensions implement trust anchor negotiation, so security considerations are largely unchanged from certificate_authorities. This section discusses security considerations for trust anchor negotiation in general.¶
PKI-based TLS authentication depends on the relying party's certificate policies. If the relying party trusts an untrustworthy CA, that CA can intercept TLS connections made by that relying party by issuing certificates associating the target name with the wrong TLS key.¶
This attack vector is available with or without trust anchor negotiation. The negotiation mechanism described in this document allows certificate selection to reflect a relying party's certificate policies. It does not determine the certificate policies themselves. Relying parties remain responsible for trusting only trustworthy CAs, and untrustworthy CAs remain a security risk when trusted.¶
As with other TLS parameters, negotiation reduces a conflict between availability and security, which allows PKIs to better mitigate security risks to users. When relying parties in an existing TLS ecosystem improve their certificate policies, trust anchor negotiation helps authenticating parties navigate differences between those relying parties and existing relying parties. Each set of requirements may be satisfied without compatibility risk to the other. Section 8 discusses such scenarios in more detail.¶
Negotiation also reduces pressures on relying parties to sacrifice user security for compatibility. If a relying party does not trust an authenticating party's current CA, connections between the two will fail until either the relying party trusts the CA or the authenticating party uses an already trusted CA. Without trust anchor negotiation, the authenticating party is limited to one certificate, and therefore switching CAs risks compatibility problems with other relying parties. The relying party then faces compatibility pressure to add this CA, even if it deems the CA a security risk. With trust anchor negotiation, the authenticating party can use its existing CA in addition to another CA trusted by the relying party. This allows the ecosystem to improve interoperability without sacrificing user security.¶
Trust anchor negotiation reduces compatibility pressures against authenticating parties serving certificates from a less common CA, as they can be served with other certificates. In some cases, the CA may have been distrusted, but still used to support older relying parties. As discussed in Section 8 and Section 10.2.2, this capability aids PKI transitions that mitigate security risks to users.¶
Even if the CA is untrustworthy, these certificates do not enable the CA to decrypt or intercept the connection. If a certificate asserts the correct information about the authenticating party, notably the correct public key, the authenticating party can safely present it. Issuing a certificate for the authenticating party's public key does not grant the CA access to the corresponding private key. Conversely, if the attacker already has access to the authenticating party's private key, they do not need to be in control of a CA to intercept a connection.¶
Rather, it is the relying party's choice of trusted CAs that determines susceptibility to interception. If the relying party trusts a misbehaving or attacker-controlled CA, the attacker can intercept the connection with a public key certified by that CA, regardless of which CA is used by the intended authenticating party. Conversely, if the relying party does not trust the attacker's CA, the attacker cannot successfully intercept the connection using a public key certified by this CA.¶
Choosing trusted CAs is a complex, security-critical process, the full considerations of which are outside the scope of this document. Relying parties thus SHOULD NOT interpret the authenticating party's choice of CA as an endorsement of the CA. Trusting a CA means trusting all certificates issued by that CA, so it is not enough to observe correct certificates from an authenticating party. An untrustworthy CA may sign one correct certificate, but also sign incorrect certificates, possibly in the future, that can attack the relying party.¶
A network attacker in possession of a misissued certificate could use trust anchor negotiation to differentiate clients and only enable TLS interception with clients that accept the certificate. The network attacker may wish to do this to reduce the odds of detection.¶
However, trust anchor negotiation only impacts detection where this differentiation was not already possible. In TLS, the client offers all its available TLS features, including cipher suites and other extensions, in the TLS ClientHello. Any variation in client TLS policies, related or unrelated to trust anchors, may be used as a fingerprint. Transport properties, such as IP geolocation, may also be used. While fingerprinting's heuristic nature makes broad, legitimate use difficult, a network attacker's single interception service can easily use it for targeted attacks.¶
If the attacker targets any clients that enforce Certificate Transparency [RFC6962], the misissued certificates will need to be publicly logged. In this case, detection is more robust, and client differentiation, with or without trust anchor negotiation, has no significant impact.¶
IANA is requested to create the following entry in the TLS ExtensionType Values registry, originally created in [RFC4366]:¶
| Value | Extension Name | TLS 1.3 | DTLS-Only | Recommended | Reference |
|---|---|---|---|---|---|
| TBD | trust_anchors | CH, EE, CR, CT | N | Y | [this-RFC] |
IANA is requested to create the following entry in the "Media Types" registry, defined in [RFC6838]:¶
application¶
pem-certificate-chain-with-properties¶
None¶
None¶
7bit¶
Carries a cryptographic certificate and its associated certificate chain and additional properties. This media type carries no active content.¶
None¶
[this-RFC, Section 7.4]¶
ACME clients and servers, HTTP servers, other applications that need to be configured with a certificate chain¶
See Authors' Addresses section.¶
COMMON¶
n/a¶
See Authors' Addresses section.¶
IETF¶
IANA is requested to create the "CertificatePropertyType" registry within the "Transport Layer Security (TLS) Extensions" group. The initial entries in the registry are as follows:¶
| Decimal | Description | References |
|---|---|---|
| 0 | trust_anchor_id | [this-RFC] |
| 1 | trust_anchor_groups | [this-RFC] |
| 2 | trust_anchor_negotiation | [this-RFC] |
New values are allocated according to the following process:¶
This section contains test vectors for trust anchor ID patterns (Section 5.3.1). Patterns and IDs are provided in their byte representations in hexadecimal.¶
The following IDs are contained in the pattern 81fd5981fd597b8348861580 (32473.{123-456}.{789-}):¶
81fd597b8615 (32473.123.789)¶
81fd59822c8704 (32473.300.900)¶
81fd598348868d1f (32473.456.99999)¶
81fd59834881ffffffffffffffff7f (32473.456.(264-1))¶
81fd59834882808080808080808000 (32473.456.(264))¶
The following IDs are not contained the pattern 81fd5981fd597b8348861580 (32473.{123-456}.{789-}):¶
81fd597b (32473.123, too few components)¶
81fd597b861500 (32473.123.789.0, too many components)¶
81fd5a7b8615 (32474.123.789, first component out of range)¶
81fd5983748615 (32473.500.789, second component out of range)¶
81fd597b853c (32473.123.700, third component out of range)¶
8081fd597b8615 (invalid ID, not minimally encoded)¶
81fd597b8695 (invalid ID, component was truncated)¶
The following IDs are contained in the pattern 81fd5981fd598280808080808080800182808080808080808003 (32473.{264+1 - 264+3}):¶
81fd5982808080808080808001 (32473.(264+1))¶
81fd5982808080808080808002 (32473.(264+2))¶
81fd5982808080808080808003 (32473.(264+3))¶
The following IDs are not contained the pattern 81fd5981fd598280808080808080800182808080808080808003 (32473.{264+1 - 264+3}):¶
81fd5902 (32473.2)¶
81fd5982808080808080808000 (32473.264)¶
81fd5982808080808080808004 (32473.(264+4))¶
The ID 81fd59 (32473) is not contained in the pattern 81fd59. The pattern is invalid with an odd number of components.¶
The ID 81fd59 (32473) is not contained in the pattern 81fd. The pattern is invalid with a truncated min value.¶
The ID 81fd59 (32473) is not contained in the pattern 81fd5981ffff. The pattern is invalid with a truncated max value.¶
The ID 00 (0) is not contained in the pattern 8042. The pattern is invalid because min cannot be infinity.¶
The authors thank Nick Harper, Ilari Liusvaara, and Emily Stark for many valuable discussions and insights which led to this document. Thanks also to Aaron Gable for providing feedback on ACME extensions.¶