Internet-Draft | COSE Algs for Two-Party Signing | February 2025 |
Lundberg & Jones | Expires 22 August 2025 | [Page] |
This specification defines COSE algorithm identifiers used when the signing operation is performed cooperatively between two parties. When performing two-party signing, the first party typically hashes the data to be signed and the second party signs the hashed data computed by the first party. This can be useful when communication with the party holding the signing private key occurs over a limited-bandwidth channel, such as NFC or Bluetooth Low Energy (BLE), in which it is infeasible to send the complete set of data to be signed. The resulting signatures are identical in structure to those computed by a single party, and can be verified using the same verification procedure without additional steps to preprocess the signed data.¶
This note is to be removed before publishing as an RFC.¶
Status information for this document may be found at https://datatracker.ietf.org/doc/draft-lundberg-cose-two-party-signing-algs/.¶
Discussion of this document takes place on the COSE Working Group mailing list (mailto:cose@ietf.org), which is archived at https://mailarchive.ietf.org/arch/browse/cose/. Subscribe at https://www.ietf.org/mailman/listinfo/cose/.¶
Source for this draft and an issue tracker can be found at https://github.com/YubicoLabs/cose-two-party-signing-algs-rfc.¶
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 22 August 2025.¶
Copyright (c) 2025 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.¶
CBOR Object Signing and Encryption (COSE) [RFC9052] algorithm identifiers are used to specify the cryptographic operations performed when creating cryptographic data structures, but do not record all the details of how the cryptography was performed, since those details are typically irrelevant for the recipient. The algorithm identifiers defined by this specification facilitate the cooperation of two parties to perform COSE signing operations together. They are used to specify the division of responsibilities between the two parties. Consumers of the cryptographic data structures thus cooperatively produced do not use these algorithm identifiers; rather, consumers use the normal COSE algorithm identifiers that correspond to the cryptographic operation cooperatively performed together by the two parties.¶
A use case for this is performing a signature operation split between two parties, such as a software application and a discrete hardware security module (HSM) holding the private key. In particular, since the data link between them may have limited bandwidth, it may not be practical to send the entire original message to the HSM. Instead, since most signature algorithms begin with digesting the message into a fixed-length intermediate input, this initial digest can be computed by the software application while the HSM computes the rest of the signature algorithm on the digest.¶
Since different signature algorithms digest the message in different ways and at different stages of the algorithm, there is no generally-applicable way to define such a division point for every possible signature algorithm. Therefore, this document defines algorithm identifiers encoding, for a specific set of signature algorithms, which steps of the signature algorithm are performed by the digester (e.g., software application) and which are performed by the signer (e.g., HSM). In general, the signer holds exclusive control of the signing private key.¶
Note that these algorithm identifiers do not define new "pre-hashed" variants of the base signature algorithm, nor an intermediate "hash envelope" data structure, such as that defined in [COSE-Hash-Envelope]. Rather, these identifiers correspond to existing signature algorithms that would typically be executed by a single party, but split into two stages. The resulting signatures are identical to those computed by a single party, and can be verified using the same verification procedures without additional special steps to process the signed data.¶
However some signature algorithms, for example, PureEdDSA [RFC8032] and ML-DSA [FIPS-204], cannot be split in this way and therefore cannot be assigned two-party signing algorithm identifiers. However, if such a signature algorithm defines a "pre-hashed" variant, such as Ed25519ph [RFC8032] or HashML-DSA [FIPS-204], that "pre-hashed" algorithm can also be assigned a two-party signing algorithm identifier, enabling the hashing step to be performed by the digester and the signing step to be executed by the signer.¶
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 section defines divisions of signing algorithm steps between a digester and a signer in a two-party signing protocol, and assigns algorithm identifiers to these algorithm divisions. The digester performs the first part of the divided algorithm and does not have access to the signing private key, while the signer performs the second part of the divided algorithm and has access to the signing private key. For signing algorithms that format the message to insert domain separation tags, as described in Section 2.2.5 of [RFC9380], this message formatting is also performed by the signer.¶
The algorithm identifiers defined in this document MUST NOT appear in COSE structures other than COSE_Key_Ref (see Section 3). They are meant only for coordination between the digester and the signer in a two-party signing protocol. Representations of the keys used and the resulting signatures MUST use the corresponding conventional algorithm identifiers instead. These are listed in the "Base algorithm" column in the tables defining two-party signing algorithm identifiers.¶
Two-party ECDSA [FIPS-186-5] uses the following division between the digester and the signer of the steps of the ECDSA signature generation algorithm [FIPS-186-5]:¶
The signing procedure is defined in Section 6.4.1 of [FIPS-186-5].¶
The digester performs Step 1 of the signing procedure - hashing the message, producing the value H.¶
The message input to the signer is the value H defined in the signing procedure.¶
The signer resumes the signing procedure from Step 2.¶
The following algorithm identifiers are defined:¶
Name | COSE Value | Base algorithm | Description |
---|---|---|---|
ESP256-2p | TBD | ESP256 | ESP256 [fully-spec-algs] divided as defined in Section 2.1 of this document |
ESP384-2p | TBD | ESP384 | ESP384 [fully-spec-algs] divided as defined in Section 2.1 of this document |
ESP512-2p | TBD | ESP512 | ESP512 [fully-spec-algs] divided as defined in Section 2.1 of this document |
Two-party HashEdDSA [RFC8032] uses the following division between the digester and the signer of the steps of the HashEdDSA signing algorithm [RFC8032]:¶
HashEdDSA is a combination of the EdDSA signing procedure and the PureEdDSA signing procedure. The EdDSA signing procedure is defined in the first paragraph of Section 3.3 of [RFC8032]. The PureEdDSA signing procedure is defined in the second paragraph of Section 3.3 of [RFC8032].¶
The digester computes the value PH(M)
defined in the EdDSA signing procedure.¶
The message input to the signer is the value PH(M)
defined in the EdDSA signing procedure.
This value is represented as M
in the PureEdDSA signing procedure.¶
The signer executes the PureEdDSA signing procedure,
where the value denoted M
in the PureEdDSA signing procedure
takes the value denoted PH(M)
in the EdDSA signing procedure.¶
PureEdDSA [RFC8032] cannot be divided in this way since such a division would require that the digester has access to the private key.¶
The following algorithm identifiers are defined:¶
Name | COSE Value | Base algorithm | Description |
---|---|---|---|
Ed25519ph-2p | TBD | Ed25519ph | Ed25519ph [fully-spec-algs] divided as defined in Section 2.2 of this document (NOTE: Ed25519ph not yet registered) |
Ed448ph-2p | TBD | Ed448ph | Ed448ph [fully-spec-algs] divided as defined in Section 2.2 of this document (NOTE: Ed448ph not yet registered) |
Two-party HashML-DSA [FIPS-204] uses the following division between the digester and the signer of the steps of the HashML-DSA.Sign algorithm:¶
The signing procedure is defined in Section 5.4.1 of [FIPS-204].¶
The digester computes the value PHM defined in Steps 10 to 22 of the signing procedure.¶
The message input to the signer is the value PHM defined in the signing procedure.
The additional ctx input must also be transmitted to the signer.
This may, for example, be done using the ctx (-1)
parameter of a COSE_Key_Ref
with kty (1): Ref-ML-DSA (TBD)
(see Section 4.1 and Section 4.2).¶
The signer executes all steps of the signing procedure except the Steps 13, 16, 19 or similar that compute the value PHM. Note in particular, that the signer generates the value rnd in Steps 5-8 and constructs the value M' in Step 23.¶
The "pure" ML-DSA version [FIPS-204] cannot be divided in this way
because of how the embedding of the ctx and tr values is constructed
in ML-DSA.Sign
and ML-DSA.Sign_Internal
.
A division like the one above for HashML-DSA would move control of this embedding from the signer to the digester.
This would break the domain separation enforced by the embedding
and possibly enable signature malleability attacks or protocol confusion attacks.¶
The following algorithm identifiers are defined:¶
Name | COSE Value | Base algorithm | Description |
---|---|---|---|
HashML-DSA-44-2p | TBD | HashML-DSA-44 | HashML-DSA-44 TODO: divided as defined in Section 2.3 of this document (NOTE: HashML-DSA-44 not yet registered) |
HashML-DSA-65-2p | TBD | HashML-DSA-65 | HashML-DSA-65 TODO: divided as defined in Section 2.3 of this document (NOTE: HashML-DSA-65 not yet registered) |
HashML-DSA-87-2p | TBD | HashML-DSA-87 | HashML-DSA-87 TODO: divided as defined in Section 2.3 of this document (NOTE: HashML-DSA-87 not yet registered) |
While keys used by many other algorithms can usually be referenced by a single atomic identifier,
such as that used in the kid
parameter in a COSE_Key object or in the unprotected header of a COSE_Recipient,
some signature algorithms use additional parameters to the signature generation
beyond the signing private key and message to be signed.
For example, ML-DSA [FIPS-204] has the additional parameter ctx
and ARKG-Derive-Secret-Key
[ARKG] has the parameters kh
and info
, in addition to the private key.¶
While these additional parameters are simple to provide to the API of the signing procedure in a single-party context, in a two-party context these additional parameters also need to be conveyed from the digester to the signer. For this purpose, we define new COSE key types, collectively called "COSE key reference types". This enables defining a unified, algorithm-agnostic protocol between the digester and the signer, rather than requiring a distinct protocol for each signature algorithm for the sake of conveying algorithm-specific parameters.¶
A COSE key reference is a COSE_Key object whose kty
value is defined to represent a reference to a key.
The kid
parameter MUST be present when kty
is a key reference type.
These requirements are encoded in the CDDL [RFC8610] type COSE_Key_Ref
:¶
COSE_Key_Ref = COSE_Key .within { 1 ^ => $COSE_kty_ref ; kty: Any reference type 2 ^ => any, ; kid is required any => any, ; Any other entries allowed by COSE_Key }¶
The following CDDL example represents a reference to an ML-DSA-65 key,
which uses the AKP
key type [COSE-ML-DSA],
along with the value of the ctx parameter to ML-DSA.Sign [FIPS-204]:¶
{ 1: TBD, ; kty: Ref-AKP ; kid: Opaque identifier of the AKP key 2: h'92bc2bfa738f5bb07803fb9c0c58020791acd29fbe253baa7a03ac84f4b26d44', 3: TBD, ; alg: ML-DSA-65 ; ctx argument to ML-DSA.Sign -1: 'Example application info', }¶
The following CDDL example represents a reference to a key derived by ARKG-P256ADD-ECDH
[ARKG]
and restricted for use with the ESP256 [fully-spec-algs] signature algorithm:¶
{ 1: -65538, ; kty: Ref-ARKG-derived ; kid: Opaque identifier of ARKG-pub 2: h'60b6dfddd31659598ae5de49acb220d8 704949e84d484b68344340e2565337d2', 3: -9, ; alg: ESP256 ; ARKG-P256ADD-ECDH key handle ; (HMAC-SHA-256-128 followed by SEC1 uncompressed ECDH public key) -1: h'ae079e9c52212860678a7cee25b6a6d4 048219d973768f8e1adb8eb84b220b0ee3 a2532828b9aa65254fe3717a29499e9b aee70cea75b5c8a2ec2eb737834f7467 e37b3254776f65f4cfc81e2bc4747a84', ; info argument to ARKG-Derive-Private-Key -2: 'Example application info', }¶
This section registers the following values in the IANA "COSE Key Types" registry [IANA.COSE]:¶
These registrations add the following choices to the CDDL [RFC8610] type socket $COSE_kty_ref
:¶
$COSE_kty_ref /= -1 ; Value TBD $COSE_kty_ref /= -2 ; Value TBD $COSE_kty_ref /= TBD ; Value TBD¶
This section registers the following values in the IANA "COSE Key Type Parameters" registry [IANA.COSE]:¶
-00¶
Initial individual draft¶