Preventing Key Reuse and Cross‑Key Forgeries in Composite ML-DSA

Status of This Memo

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 15 September 2026.¶

Copyright Notice

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.¶

Table of Contents

1. Introduction

Composite signature schemes (e.g., Composite ML‑DSA) pre-hash the application message and prepend a Prefix, an algorithm‑specific Label, and an application‑provided ctx byte string to form the message representative M', which is then signed by each component primitive. The current composite signature construction:¶

M' :=  Prefix || Label || len(ctx) || ctx || PH( M )

The ctx is an application context of up to 255 bytes.¶

While the existing design already mitigates several cross‑protocol issues via Prefix and Label, and explicitly forbids key reuse, operational reality suggests some deployments may still reuse component keys or attempt to combine component signatures across keys. This opens the door to cross‑key “mix‑and‑match” forgeries (splicing a valid ML‑DSA component from one composite with a valid traditional component from another).¶

This document proposes a minimal, mechanical change: set ctx to a hash of the composite public key. Because the hash depends on the exact public key bytes, both component signatures become bound to the same key material, preventing cross‑key recombination.¶

2. Conventions and Definitions

This document inherits the notation of the Composite ML‑DSA draft (e.g., Prefix, Label, PH, SerializePublicKey, etc.) and the conventional KeyGen/Sign/Verify API of a signature scheme.¶

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.¶

3. Threat Model and Goals

Goal: prevent an adversary from taking valid component signatures produced under different composite keys and combining them into a valid composite signature for a target key. The base document notes SUF‑CMA subtleties and “mix and match” where (M, (mldsaSig1, tradSig2)) could be valid if both were obtained separately.¶

Approach: bind both component signatures to the same exact composite public key by including pkc = H(SerializePublicKey(..)) inside M'. Because pkc changes with any bit of the key, component signatures extracted from different keys no longer verify together.¶

• This does not change the malleability properties of individual primitives (e.g., ECDSA); therefore SUF‑CMA is still not claimed. It primarily removes cross‑key splicing and reduces the impact of accidental key reuse.¶

4. Overview of the Construction

This construction proposes len(pkc) as the length of the len(ctx) and pkc as the ctx value; all signature computation remains as specified in the base document.¶

Let Hash_ctx denote the hash function chosen by the algorithm’s OID (e.g., SHA‑256, SHA‑512, SHAKE256/64). The Public‑Key Context is:¶

pkc = Hash_ctx( SerializePublicKey(mldsaPK, tradPK) )

The new message representative is:¶

M' := Prefix || Label || len(pkc) || pkc || PH(M)

Prefix and Label are unchanged. len(pkc) is encoded as a single unsigned byte, which is sufficient because the mandated hash outputs are ≤ 64 bytes. The PH(M) is the pre‑hash of the application message as in the base document.¶

5. Public‑Key Context (PKC) Routines

This draft defines two convenience routines to compute pkc from either the composite private key or the composite public key.¶

Composite-ML-DSA<OID>.ComputePublicKeyContext(sk) -> pkc

Inputs:
sk: composite private key (serialized as in Section 4.2 of the base spec)

Implicit inputs (from <OID>):
Hash_ctx: the hash function for PKC (same as the algorithm’s PH unless specified otherwise)

Process:

1.  (mldsaSeed, tradSK) = DeserializePrivateKey(sk)
2.  (_, mldsaSK) = ML-DSA.KeyGen_internal(mldsaSeed)        // base spec, seed-based expansion
3.  mldsaPK = ML-DSA.PublicKey(mldsaSK)                     // or derived during step 2
4.  tradPK  = Trad.PublicKey(tradSK)                        // algorithm-specific derivation
5.  pk = SerializePublicKey(mldsaPK, tradPK)                // base spec Section 4.1
6.  pkc = Hash_ctx(pk)
7.  return pkc

Composite-ML-DSA<OID>.ComputePublicKeyContext(pk) -> pkc

Inputs:
pk: composite public key (serialized as in Section 4.1 of the base spec)

Implicit inputs (from <OID>):
Hash_ctx

Process:
1.  pkc = Hash_ctx(pk)
2.  return pkc

6. Algorithms

Both Composite-ML-DSA<OID>.Sign(sk, M, ctx) and Composite-ML-DSA<OID>.Verify(pk, M, s, ctx) remain the same. The only difference is that pkc is passed in as the ctx value.¶

The signature interface remains the same.¶

Note: The application specific ctx argument is ignored with this current design. To keep that property, a future version of this specification could set pkc = HASH (ctx || publickey).¶

7. Serialization and ASN.1 Usage

This document does not change the composite public/private key or signature serialization formats from the base spec—keys and signatures remain concatenations of the component encodings. It also does not change DER wrapping in SPKI/PKCS#8.¶

Because wire compatibility requires peers to know whether ctx is application‑set or PKC‑bound, this document could haved register new algorithm identifiers for each PKC‑bound combination. However, that is not within the scope of this document. This is meant for specific application context use-cases where the preventing key reuse is a desired security property.¶

8. Security Considerations

Key Reuse: The base spec strictly forbids reusing component keys across composite and non‑composite contexts or across composites. Binding ctx to pkc provides a cryptographic backstop: even if component keys were (improperly) reused, cross‑key splicing will fail because pkc differs for each public key instance.¶

Non‑separability: The base construction achieved Weak Non‑Separability (WNS) and a limited form of SNS for ML‑DSA via the mldsa_ctx=Label. PKC‑binding additionally prevents forming (mldsaSig1, tradSig2) under different keys, because both signatures are now bounded to the same pkc. This does not fix primitive‑level malleability (e.g., ECDSA) and therefore does not claim SUF‑CMA. However, for algorithms like EdDSA or Ed448 which are SUF secure, this property should remain.¶

Prefix Guard: Existing Prefix and Label remain unchanged; deployments that implemented the optional Prefix guard for traditional verifiers can keep it as is.¶

Hash Choices: Hash_ctx MUST be the algorithm’s registered pre‑hash function (e.g., SHA‑256, SHA‑512, SHAKE256/64). This keeps implementation complexity minimal and ensures digest sizes fit within the ctx length field.¶

Privacy: pkc reveals nothing beyond what the public key already reveals; it is a hash of public data.¶

9. Implementation Considerations

Signer Access to pk: The signer computes pkc either by deriving tradPK from tradSK, or by keeping a cached copy of pk alongside sk.¶

Interoperability: Because M' changes when this context type is used, peers MUST know that this context will be used. One way to achieve this is for application specific use cases to specify use of this context type as part of the usage.¶

10. IANA Considerations

None¶

11. Normative References

[RFC2119]

Bradner, S., "Key words for use in RFCs to Indicate Requirement Levels", BCP 14, RFC 2119, DOI 10.17487/RFC2119, March 1997, <https://www.rfc-editor.org/rfc/rfc2119>.

[RFC8174]

Leiba, B., "Ambiguity of Uppercase vs Lowercase in RFC 2119 Key Words", BCP 14, RFC 8174, DOI 10.17487/RFC8174, May 2017, <https://www.rfc-editor.org/rfc/rfc8174>.

Acknowledgments

Thanks to the Composite ML‑DSA authors and LAMPS WG for the existing combiner design and analyses of pre‑hashing, non‑separability, and key‑reuse risks which this document builds upon.¶

Author's Address