Composite ML-KEM for use in X.509 Public Key Infrastructure and CMS

2.1. Conventions and Terminology

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. These words may also appear in this document in lower case as plain English words, absent their normative meanings.¶

This document is consistent with all terminology from [I-D.ietf-pquip-pqt-hybrid-terminology]. In addition, the following terms are used in this document:¶

ALGORITHM: The usage of the term "algorithm" within this document generally refers to any function which has a registered Object Identifier (OID) for use within an ASN.1 AlgorithmIdentifier. This loosely, but not precisely, aligns with the definitions of "cryptographic algorithm" and "cryptographic scheme" given in [I-D.ietf-pquip-pqt-hybrid-terminology].¶

COMBINER: A combiner specifies how multiple shared secrets are combined into a single shared secret.¶

DER: Distinguished Encoding Rules as defined in [X.690].¶

KEM: A key encapsulation mechanism as defined in Section 3.¶

PKI: Public Key Infrastructure, as defined in [RFC5280].¶

SHARED SECRET KEY: A value established between two communicating parties for use as cryptographic key material suitable for direct use by symmetric cryptographic algorithms. This document is concerned with shared secrets established via public key cryptographic operations.¶

2.2. Composite Design Philosophy

[I-D.ietf-pquip-pqt-hybrid-terminology] defines composites as:¶

• Composite Cryptographic Element: A cryptographic element that incorporates multiple component cryptographic elements of the same type in a multi-algorithm scheme.¶

Composite keys, as defined here, follow this definition and should be regarded as a single key that performs a single cryptographic operation such as key generation, signing, verifying, encapsulating, or decapsulating -- using its internal sequence of component keys as if they form a single key. This generally means that the complexity of combining algorithms can and should be handled by the cryptographic library or cryptographic module, and the single composite public key, private key, ciphertext and signature can be carried in existing fields in protocols such as PKCS#10 [RFC2986], CMP [RFC4210], X.509 [RFC5280], CMS [RFC5652], and the Trust Anchor Format [RFC5914]. In this way, composites achieve "protocol backwards-compatibility" in that they will drop cleanly into any protocol that accepts an analogous single-algorithm cryptographic scheme without requiring any modification of the protocol to handle multiple algorithms.¶

3.1. Promotion of RSA-OAEP into a KEM

The RSA Optimal Asymmetric Encryption Padding (OAEP), as defined in section 7.1 of [RFC8017] is a public key encryption algorithm used to transport key material from a sender to a receiver. It is promoted into a KEM by having the sender generate a random 256 bit secret and encrypt it.¶

Note that, at least at the time of writing, the algorithm RSAOAEPKEM is not defined as a standalone algorithm within PKIX standards and it does not have an assigned algorithm OID, so it cannot be used directly with CMS KEMRecipientInfo [RFC9629]; it is merely a building block for the composite algorithm.¶

RSAOAEPKEM.Encap(pkR):
  shared_secret = SecureRandom(ss_len)
  enc = RSAES-OAEP-ENCRYPT(pkR, shared_secret)

  return shared_secret, enc

Note that the OAEP label L is left to its default value, which is the empty string as per [RFC8017]. The shared secret output by the overall Composite ML-KEM already binds a composite domain separator, so there is no need to also utilize the component domain separators.¶

The value of ss_len as well as the RSA-OAEP parameters used within this specification can be found in Section 7.4.¶

Decap(sk, ct) -> ss is accomplished in the analogous way.¶

RSAOAEPKEM.Decap(skR, enc):
  shared_secret = RSAES-OAEP-DECRYPT(skR, enc)

  return shared_secret

3.2. Promotion of ECDH into a KEM

An elliptic curve Diffie-Hellman key agreement is promoted into a KEM Encap(pk) -> (ss, ct) using a simplified version of the DHKEM definition from [RFC9180]; simplified to remove the context-binding labels since the shared secret output by the overall Composite ML-KEM already binds a composite domain separator, so there is no need to also utilize labels within DHKEM.¶

Note that, at least at the time of writing, the algorithm DHKEM is not defined as a standalone algorithm within PKIX standards and it does not have an assigned algorithm OID, so it cannot be used directly with CMS KEMRecipientInfo [RFC9629]; it is merely a building block for the composite algorithm.¶

DHKEM.Encap(pkR):
  skE, pkE = GenerateKeyPair()
  shared_secret = DH(skE, pkR)
  enc = SerializePublicKey(pkE)

  return shared_secret, enc

Decap(sk, ct) -> ss is accomplished in the analogous way.¶

DHKEM.Decap(skR, enc):
  pkE = DeserializePublicKey(enc)
  shared_secret = DH(skR, pkE)

  return shared_secret

This construction applies for all variants of elliptic curve Diffie-Hellman used in this specification: ECDH, X25519, and X448.¶

The simplifications from the DHKEM definition in [RFC9180] is that since the ciphertext and receiver's public key are included explicitly in the Composite ML-KEM combiner, there is no need to construct the kem_context object, and since a domain separator is included explicitly in the Composite ML-KEM combiner there is no need to perform the labeled steps of ExtractAndExpand().¶

4.1. Key Generation

To generate a new keypair for Composite schemes, the KeyGen() -> (pk, sk) function is used. The KeyGen() function calls the two key generation functions of the component algorithms for the Composite keypair in no particular order. Multi-process or multi-threaded applications might choose to execute the key generation functions in parallel for better key generation performance.¶

The following process is used to generate composite keypair values:¶

KeyGen() -> (pk, sk)

Explicit Inputs:
     None

Implicit Input:
  ML-KEM     A placeholder for the specific ML-KEM algorithm and
             parameter set to use, for example, could be "ML-KEM-65".

  Trad       A placeholder for the specific traditional algorithm and
             parameter set to use, for example "RSA-OAEP"
             or "X25519".

Output:
  (pk, sk)  The composite keypair.

Key Generation Process:

  1. Generate component keys

    (mlkemPK, mlkemSK) = ML-KEM.KeyGen()
    (tradPK, tradSK)   = Trad.KeyGen()

  2. Check for component key gen failure
    if NOT (mlkemPK, mlkemSK) or NOT (tradPK, tradSK):
      output "Key generation error"

  3. Output the composite public and private keys

    pk = (mlkemPK, tradPK)
    sk = (mlkemSK, tradSK)
    return (pk, sk)

In order to ensure fresh keys, the key generation functions MUST be executed for both component algorithms. Compliant parties MUST NOT use or import component keys that are used in other contexts, combinations, or by themselves as keys for standalone algorithm use.¶

Note that in step 2 above, both component key generation processes are invoked, and no indication is given about which one failed. This SHOULD be done in a timing-invariant way to prevent side-channel attackers from learning which component algorithm failed.¶

4.2. Encapsulation

The Encap(pk) of a Composite ML-KEM algorithm is designed to behave exactly the same as ML-KEM.Encaps(ek) defined in Algorithm 20 in Section 7.2 of [FIPS.203]. Specifically, Composite-ML-KEM.Encap(pk) produces a 256-bit shared secret key that can be used directly with any symmetric-key cryptographic algorithm. In this way, Composite ML-KEM can be used as a direct drop-in replacement anywhere that ML-KEM is used.¶

Composite-ML-KEM.Encap(pk) -> (ss, ct)

Explicit Input:

  pk          Composite public key consisting of encryption public keys
              for each component.

Implicit inputs:

  ML-KEM   A placeholder for the specific ML-KEM algorithm and
           parameter set to use, for example, could be "ML-KEM-768".

  Trad     A placeholder for the specific ML-KEM algorithm and
           parameter set to use, for example "RSA-OAEP"
           or "X25519".

  KDF      The KDF specified for the given Composite ML-KEM algorithm.
           See algorithm specifications below.

  Domain   Domain separator value for binding the ciphertext to the
           Composite OID. See section on Domain Separators below.

Output:

  ss      The shared secret key, a 256-bit key suitable for use with
          symmetric cryptographic algorithms.

  ct      The ciphertext, a byte string.

Encap Process:

  1. Separate the public keys.

      (mlkemPK, tradPK) = pk

  2.  Perform the respective component Encap operations according to
      their algorithm specifications.

      (mlkemCT, mlkemSS) = MLKEM.Encaps(mlkemPK)
      (tradCT, tradSS) = TradKEM.Encap(tradPK)

  3. If either ML-KEM.Encaps() or TradKEM.Encap() return an error,
     then this process must return an error.

      if NOT (mlkemCT, mlkemSS) or NOT (tradCT, tradSS):
        output "Encapsulation error"

  4. Encode the ciphertext

     ct = mlkemCT || tradCT

  5. Combine the KEM secrets and additional context to yield the composite shared secret

     ss = KDF(mlkemSS || tradSS || tradCT || tradPK || Domain)

  6. Output composite shared secret key and ciphertext

     return (ss, ct)

The specific values for KDF are defined per Composite ML-KEM algorithm in Table 2 and the specific values for Domain are defined per Composite ML-KEM algorithm in Section 7.¶

4.3. Decapsulation

The Decap(sk, ct) -> ss of a Composite ML-KEM algorithm is designed to behave exactly the same as ML-KEM.Decaps(dk, c) defined in Algorithm 21 in Section 7.3 of [FIPS.203]. Specifically, Composite-ML-KEM.Decap(sk, ct) produces a 256-bit shared secret key that can be used directly with any symmetric-key cryptographic algorithm. In this way, Composite ML-KEM can be used as a direct drop-in replacement anywhere that ML-KEM is used.¶

Composite-ML-KEM.Decap(sk, ct) -> ss

Explicit Input:

  sk    Composite private key consisting of decryption private keys for
        each component.

  ct      The ciphertext, a byte string.

Implicit inputs:

  ML-KEM   A placeholder for the specific ML-KEM algorithm and
           parameter set to use, for example, could be "ML-KEM-768".

  Trad     A placeholder for the specific traditional algorithm and
           parameter set to use, for example "RSA-OAEP"
           or "X25519".

  KDF      The KDF specified for the given Composite ML-KEM algorithm.
           See algorithm specifications below.

  Domain   Domain separator value for binding the ciphertext to the
           Composite OID. See section on Domain Separators below.

Output:

  ss      The shared secret key, a 256-bit key suitable for use with
          symmetric cryptographic algorithms.

Decap Process:

  1. Separate the private keys

      (mlkemSK, tradSK) = sk

  2. Parse the ciphertext

      (mlkemCT, tradCT) = ct

  3.  Perform the respective component Encap operations according to
      their algorithm specifications.

      mlkemSS = MLKEM.Decaps(mlkemSK, mlkemCT)
      tradSS  = TradKEM.Decap(tradSK, tradCT)

  4. If either ML-KEM.Decaps() or TradKEM.Decap() return an error,
     then this process must return an error.

      if NOT mlkemSS or NOT tradSS:
        output "Encapsulation error"

  5. Combine the KEM secrets and additional context to yield the composite shared secret

      ss = KDF(mlkemSS || tradSS || tradCT || tradPK || Domain)

  6. Output composite shared secret key

     return ss

It is possible to use component private keys stored in separate software or hardware keystores. Variations in the process to accommodate particular private key storage mechanisms are considered to be conformant to this document so long as it produces the same output and error handling as the process sketched above.¶

In order to properly achieve its security properties, the KEM combiner requires that all inputs are fixed-length. Since each Composite ML-KEM algorithm fully specifies its component algorithms, including key sizes, all inputs should be fixed-length in non-error scenarios, however some implementations may need to perform additional checking to handle certain error conditions. In particular, the KEM combiner step should not be performed if either of the component decapsulations returned an error condition indicating malformed inputs. For timing-invariance reasons, it is RECOMMENDED to perform both decapsulation operations and check for errors afterwards to to prevent an attacker from using a timing channel to tell which component failed decapsulation. Also, RSA-based composites MUST ensure that the modulus size (ie the size of tradCT and tradPK) matches that specified for the given Composite ML-KEM algorithm in Table 2; depending on the cryptographic library used, this check may be done by the library or may require an explicit check as part of the CompositeKEM.Decap() routine.¶

4.4. SerializePublicKey and DeserializePublicKey

The KEM public key serialization routine simply concatenates the fixed-length public keys of the constituent KEMs, as defined below.¶

Composite-ML-KEM.SerializePublicKey(pk) -> bytes

Explicit Input:

  pk    Composite ML-KEM public key

Implicit inputs:

  ML-KEM   A placeholder for the specific ML-KEM algorithm and
           parameter set to use, for example, could be "ML-KEM-768".

  Trad     A placeholder for the specific traditional algorithm and
           parameter set to use, for example "RSA-OAEP"
           or "X25519".

Output:

  bytes   The encoded public key

Serialization Process:

  1. Separate the public keys

     (mlkemPK, tradPK) = pk

  2. Serialize each of the constituent public keys

     mlkemEncodedPK = ML-KEM.SerializePublicKey(mlkemPK)
     tradEncodedPK = Trad.SerializePublicKey(tradPK)

  3. Combine and output the encoded public key

     bytes = mlkemEncodedPK || tradEncodedPK
     output bytes

Deserialization reverses this process, raising an error in the event that the input is malformed.¶

Composite-ML-KEM.DeserializePublicKey(bytes) -> pk

Explicit Input:

  bytes   An encoded public key

Implicit inputs:

  ML-KEM   A placeholder for the specific ML-KEM algorithm and
           parameter set to use, for example, could be "ML-KEM-768".

  Trad     A placeholder for the specific traditional algorithm and
           parameter set to use, for example "RSA-OAEP"
           or "X25519".

Output:

  pk     The composite ML-KEM public key

Deserialization Process:

  1. Validate the length of the the input byte string

     if bytes is not the correct length:
      output "Deserialization error"

  2. Parse each constituent encoded public key

     (mlkemEncodedPK, tradEncodedPK) = bytes

  3. Deserialize the constituent public keys

     mlkemPK = ML-KEM.DeserializePublicKey(mlkemEncodedPK)
     tradPK = Trad.DeserializePublicKey(tradEncodedPK)

  4. If either ML-KEM.DeserializePublicKey() or
     Trad.DeserializePublicKey() return an error,
     then this process must return an error.

      if NOT mlkemPK or NOT tradPK:
        output "Deserialization error"

  5. Output the composite ML-KEM public key

     output (mlkemPK, tradPK)

4.5. SerializePrivateKey and DeserializePrivateKey

The same serialization and deserialization process as described in Section 4.4 should be used to serialize and deserialize the private keys. The only difference is that pk is the private key, and the output is the concatenation of the mlkem and traditional private keys for serialization, or the mlkem and traditional private keys for deserialization.¶

4.6. SerializeCiphertextValue and DeSerializeCiphertextValue

The serialization routine for the CompositeCiphertextValue simply concatenates the fixed-length ML-KEM cipherText value with the cipherText value from the traditional algorithm, as defined below:¶

Composite-ML-DSA.SerializeCiphertextValue(CompositeCiphertextValue) -> bytes

Explicit Input:

  CompositeCiphertextValue    The Composite CipherText Value obtained from Composite-ML-KEM.Encap(pk)

Implicit inputs:

  ML-KEM   A placeholder for the specific ML-KEM algorithm and
           parameter set to use, for example, could be "ML-KEM-768".

  Trad     A placeholder for the specific traditional algorithm and
           parameter set to use, for example "RSAOAEP" or "ECDH".

Output:

  bytes   The encoded CompositeCiphertextValue

Serialization Process:

  1. Separate the cipher texts

     (mldkemct, tradkemct) = CompositeCiphertextValue

  2. Serialize each of the constituent cipher texts

     mlkemEncodedCt = ML-KEM.SerializeCiphertext(mlkemct)
     tradkemEncodedCT = Trad.SerializeCiphertext(tradkemct)

  3. Combine and output the encoded composite ciphertext

     bytes = mlkemEncodedCt || tradkemEncodedCT
     output bytes

Deserialization reverses this process, raising an error in the event that the input is malformed.¶

Composite-ML-KEM.DeserializeCiphertextValue(bytes) -> CompositeCiphertextValue

Explicit Input:

  bytes   An encoded CompositeCiphertextValue

Implicit inputs:

  ML-KEM   A placeholder for the specific ML-KEM algorithm and
           parameter set to use, for example, could be "ML-KEM-768".

  Trad     A placeholder for the specific traditional algorithm and
           parameter set to use, for example "RSAOAEP" or "ECDH".

Output:

  CompositeCiphertextValue  The CompositeCiphertextValue

Deserialization Process:

  1. Validate the length of the the input byte string

     if bytes is not the correct length:
      output "Deserialization error"

  2. Parse each constituent encoded cipher text.
       The length of the mlkemEncodedCt is known based on the size of
       the ML-KEM component cipher text length specified by the Object ID

     (mlkemEncodedCt, tradkemEncodedCt) = bytes

  3. Deserialize the constituent cipher text values

     mlkemCt = ML-KEM.DeserializeCiphertext(mlkemEncodedCt)
     tradkemCt = Trad.DeserializeCiphertext(tradkemEncodedCt)

  4. If either ML-KEM.DeserializeCiphertext() or
     Trad.DeserializeCiphertext() return an error,
     then this process must return an error.

      if NOT mlkemCt or NOT tradkemCt:
        output "Deserialization error"

  5. Output the CompositeCiphertextValue

     output (mlkemCt, tradkemCt)

4.7. ML-KEM public key, private key and cipher text sizes for serialization and deserialization

As noted above in the public key, private key and CompositeCiphertextValue serialization and deserialization methods, ML-KEM uses fixed-length values for all of these components. This means the length encoding of the first component is known and does NOT need to be encoded into the serialization and deserialization process which simplifies the encoding. If future composite combinations make use of algorithms where the first component uses variable length keys or cipher texts, then that specification will need to ensure the length is encoded in a fixed-length prefix so the components can be correctly deserialized.¶

The following table shows the fixed length values in bytes for the public, private and cipher text sizes for ML-KEM which can be used to deserialzie the components.¶

5.1. CompositeKEMPublicKey

The wire encoding of a Composite ML-KEM public key is:¶

CompositeKEMPublicKey ::= BIT STRING

Since RSA and ECDH component public keys are themselves in a DER encoding, the following show the internal structure of the various public key types used in this specification:¶

When a CompositeKEMPublicKey is used with an RSA public key, the BIT STRING itself is generated by the concatenation of a raw ML-KEM key according to [I-D.ietf-lamps-kyber-certificates] and an RSAPublicKey (which is a DER encoded RSAPublicKey).¶

When a CompositeKEMPublicKey is used with an EC public key, the BIT STRING itself is generated by the concatenation of a raw ML-KEM key according to [I-D.ietf-lamps-kyber-certificates] and an ECDHPublicKey (which is a DER encoded ECPoint).¶

When a CompositeKEMPublicKey is used with an Edwards public key, the BIT STRING itself is generated by the concatenation of a raw ML-KEM key according to [I-D.ietf-lamps-kyber-certificates] and a raw Edwards public key according to [RFC8410].¶

Some applications may need to reconstruct the SubjectPublicKeyInfo objects corresponding to each component public key. Table 2 in Section 7 provides the necessary mapping between composite and their component algorithms for doing this reconstruction.¶

When the CompositeKEMPublicKey must be provided in octet string or bit string format, the data structure is encoded as specified in Section 5.3.¶

In order to maintain security properties of the composite, applications that use composite keys MUST always perform fresh key generations of both component keys and MUST NOT reuse existing key material. See Section 11.3 for a discussion.¶

The following ASN.1 Information Object Class is defined to allow for compact definitions of each composite algorithm, leading to a smaller overall ASN.1 module.¶

  pk-CompositeKEM {OBJECT IDENTIFIER:id, PublicKeyType}
    PUBLIC-KEY ::= {
      IDENTIFIER id
      KEY PublicKeyType
      PARAMS ARE absent
      CERT-KEY-USAGE { keyEncipherment }
    }

As an example, the public key type pk-MLKEM768-ECDH-P384 can be defined compactly as:¶

pk-MLKEM768-ECDH-P384 PUBLIC-KEY ::=
  pk-CompositeKEM {
    id-MLKEM768-ECDH-P384,
    EcCompositeKemPublicKey }

The full set of key types defined by this specification can be found in the ASN.1 Module in Section 9.¶

5.2. CompositeKEMPrivateKey

When a Composite ML-KEM private key is to be exported from a cryptographic module, it uses an analogous definition to the public keys:¶

CompositeKEMPrivateKey ::= OCTET STRING

Each element of the CompositeKEMPrivateKey Sequence is an OCTET STRING according to the encoding of the underlying algorithm specification and will decode into the respective private key structures in an analogous way to the public key structures defined in Section 5.1. This document does not provide helper classes for private keys. The PrivateKey for each component algorithm MUST be in the same order as defined in Section 5.1.¶

Use cases that require an interoperable encoding for composite private keys will often need to place a CompositeKEMPrivateKey inside a OneAsymmetricKey structure defined in [RFC5958], such as when private keys are carried in PKCS #12 [RFC7292], CMP [RFC4210] or CRMF [RFC4211]. The definition of OneAsymmetricKey is copied here for convenience:¶

 OneAsymmetricKey ::= SEQUENCE {
       version                   Version,
       privateKeyAlgorithm       PrivateKeyAlgorithmIdentifier,
       privateKey                PrivateKey,
       attributes            [0] Attributes OPTIONAL,
       ...,
       [[2: publicKey        [1] PublicKey OPTIONAL ]],
       ...
     }

  ...
  PrivateKey ::= OCTET STRING
                        -- Content varies based on type of key.  The
                        -- algorithm identifier dictates the format of
                        -- the key.

When a CompositeKEMPrivateKey is conveyed inside a OneAsymmetricKey structure (version 1 of which is also known as PrivateKeyInfo) [RFC5958], the privateKeyAlgorithm field SHALL be set to the corresponding composite algorithm identifier defined according to Section 7 and its parameters field MUST be absent. The privateKey field SHALL contain the CompositeKEMPrivateKey, and the publicKey field remains OPTIONAL. If the publicKey field is present, it MUST be a CompositeKEMPublicKey.¶

Some applications may need to reconstruct the OneAsymmetricKey objects corresponding to each component private key. Section 7 provides the necessary mapping between composite and their component algorithms for doing this reconstruction.¶

Component keys of a CompositeKEMPrivateKey MUST NOT be used in any other type of key or as a standalone key. For more details on the security considerations around key reuse, see section Section 11.3.¶

5.3. Encoding Rules

Many protocol specifications will require that the composite public key and composite private key data structures be represented by an octet string or bit string.¶

When an octet string is required, the DER encoding of the composite data structure SHALL be used directly.¶

When a bit string is required, the octets of the DER encoded composite data structure SHALL be used as the bits of the bit string, with the most significant bit of the first octet becoming the first bit, and so on, ending with the least significant bit of the last octet becoming the last bit of the bit string.¶

In the interests of simplicity and avoiding compatibility issues, implementations that parse these structures MAY accept both BER and DER.¶

5.4. Key Usage Bits

When any of the Composite ML-KEM AlgorithmIdentifier appears in the SubjectPublicKeyInfo field of an X.509 certificate [RFC5280], the key usage certificate extension MUST only contain¶

keyEncipherment

Composite ML-KEM keys MUST NOT be used in a "dual usage" mode because even if the traditional component key supports both signing and encryption, the post-quantum algorithms do not and therefore the overall composite algorithm does not.¶

6.1. kema-CompositeKEM

The ASN.1 algorithm object for a Composite ML-KEM is:¶

kema-CompositeKEM {
  OBJECT IDENTIFIER:id,
    PUBLIC-KEY:publicKeyType }
    KEM-ALGORITHM ::= {
         IDENTIFIER id
         VALUE CompositeCiphertextValue
         PARAMS ARE absent
         PUBLIC-KEYS { publicKeyType }
        }

6.2. CompositeCiphertextValue

The CompositeCipherTextValue is the DER encoding of a SEQUENCE of the ciphertexts from the underlying component algorithms. It is represented in ASN.1 as follows:¶

CompositeCiphertextValue ::= OCTET STRING

The order of the component ciphertexts is the same as the order defined in Section 5.1.¶

7.1. Composite-ML-KEM Algorithm Identifiers

For the use of HKDF [RFC5869]: a salt is not provided; ie the default salt (all zeroes of length HashLen) will be used. The output length of HKDF is the same as the block size of the underlying hash function; in particular, HKDF-SHA256/256 means HKDF-SHA256 with an output length L of 256 bits (32 octets).¶

Full specifications for the referenced algorithms can be found in Appendix B.¶

7.2. Domain Separators

The KEM combiner used in this document requires a domain separator Domain input. The following table shows the HEX-encoded domain separator for each Composite ML-KEM AlgorithmID; to use it, the value should be HEX-decoded and used in binary form. The domain separator is simply the DER encoding of the composite algorithm OID.¶

EDNOTE: these domain separators are based on the prototyping OIDs assigned on the Entrust arc. We will need to ask for IANA early allocation of these OIDs so that we can re-compute the domain separators over the final OIDs.¶

7.3. Rationale for choices

• Pair equivalent levels.¶

• NIST-P-384 is CNSA approved [CNSA2.0] for all classification levels.¶

• 521 bit curve not widely used.¶

A single invocation of SHA3 is known to behave as a dual-PRF, and thus is sufficient for use as a KDF, see Section 11.1, however SHA2 is not us must be wrapped in the HKDF construction.¶

The lower security levels (ie ML-KEM768) are provided with HKDF-SHA2 as the KDF in order to facilitate implementations that do not have easy access to SHA3 outside of the ML-KEM function. Higher security levels (ie ML-KEM1024) are paired with SHA3 for computational efficiency, and the Edwards Curve (X25519 and X448) combinations are paired with SHA3 for compatibility with other similar specifications.¶

While it may seem odd to use 256-bit hash functions at all security levels, this aligns with ML-KEM which produces a 256-bit shared secret key at all security levels. SHA-256 and SHA3-256 both have >= 256 bits of (2nd) pre-image resistance, which is the required property for a KDF to provide 128 bits of security, as allowed in Table 3 of [SP.800-57pt1r5].¶

7.4. RSA-OAEP Parameters

Use of RSA-OAEP [RFC8017] within id-MLKEM768-RSA2048, id-MLKEM768-RSA3072, and id-MLKEM768-RSA4096 requires additional specification.¶

First, a quick note on the choice of RSA-OAEP as the supported RSA encryption primitive. RSA-KEM [RFC5990] is more straightforward to work with, but it has fairly limited adoption and therefore is of limited backwards compatibility value. Also, while RSA-PKCS#1v1.5 [RFC8017] is still everywhere, but hard to make secure and no longer FIPS-approved as of the end of 2023 [SP800-131Ar2], so it is of limited forwards value. This leaves RSA-OAEP [RFC8017] as the remaining choice.¶

The RSA component keys MUST be generated at the 2048-bit and 3072-bit security levels respectively.¶

As with the other Composite ML-KEM algorithms, when id-MLKEM768-RSA2048, id-MLKEM768-RSA3072, or id-MLKEM-RSA4096 is used in an AlgorithmIdentifier, the parameters MUST be absent. The RSA-OAEP SHALL be instantiated with the following hard-coded parameters which are the same for the 2048, 3072 and 4096 bit security levels.¶

where:¶

• id-sha256 is defined in [RFC8017].¶

• mgf1SHA256Identifier is defined in [RFC4055], which refers to the MFG1 function defined in [RFC8017] appendix B.2.1.¶

• pSpecifiedEmpty is defined in [RFC8017] to indicate that the empty string is used for the label.¶

Note: The mask length, according to [RFC8017], is k - hLen - 1, where k is the size of the RSA modulus. Since the choice of hash function and the RSA key size is fixed for each composite algorithm, implementations could choose to pre-compute and hard-code the mask length.¶

8.1. Underlying Components

A compliant implementation MUST support the following algorithm combinations for the KEMRecipientInfo kdf and wrap fields when the corresponding Composite ML-KEM algorithm is listed in the KEMRecipientInfo kem field. The KDFs listed below align with the KDF used internally within the KEM combiner. An implementation MAY also support other key-derivation functions and other key-encryption algorithms within CMS KEMRecipientInfo and SHOULD use algorithms of equivalent strength or greater.¶

Full specifications for the referenced algorithms can be found either further down in this section, or in Appendix B.¶

Note that here we differ slightly from the internal KDF used within the KEM combiner in Section 7 because [RFC9629] requires that the KDF listed in the KEMRecipientInfo kdf field must have an interface which accepts KDF(IKM, L, info), so here we need to use KMAC and cannot directly use SHA3. Since we require 256-bits of (2nd) pre-image resistance, we use KMAC256 for the Composite ML-KEM algorithms with internally use SHA3-256, as aligned with Table 3 of [SP.800-57pt1r5].¶

HKDF(salt, IKM, info, L)
  = HKDF-Expand(HKDF-Extract(salt, IKM), info, L)

8.2. RecipientInfo Conventions

When Composite ML-KEM is employed for a recipient, the RecipientInfo alternative for that recipient MUST be OtherRecipientInfo using the KEMRecipientInfo structure as defined in [RFC9629].¶

The fields of the KEMRecipientInfo MUST have the following values:¶

• version is the syntax version number; it MUST be 0.¶

• rid identifies the recipient's certificate or public key.¶

• kem identifies the KEM algorithm; it MUST contain one of the Composite ML-KEM identifiers listed in Section 7.¶

• kemct is the ciphertext produced for this recipient.¶

• kdf identifies the key-derivation algorithm. Note that the Key Derivation Function (KDF) used for CMS RecipientInfo process MAY be different than the KDF used within the Composite ML-KEM algorithm or one of its components.¶

• kekLength is the size of the key-encryption key in octets.¶

• ukm is an optional random input to the key-derivation function. ML-KEM doesn't place any requirements on the ukm contents.¶

• wrap identifies a key-encryption algorithm used to encrypt the content-encryption key.¶

8.3. Certificate Conventions

The conventions specified in this section augment RFC 5280 [RFC5280].¶

The willingness to accept a Composite ML-KEM Algorithm MAY be signaled by the use of the SMIMECapabilities Attribute as specified in Section 2.5.2. of [RFC8551] or the SMIMECapabilities certificate extension as specified in [RFC4262].¶

The intended application for the public key MAY be indicated in the key usage certificate extension as specified in Section 4.2.1.3 of [RFC5280]. If the keyUsage extension is present in a certificate that conveys a Composite ML-KEM public key, then the key usage extension MUST contain only the following value:¶

keyEncipherment

The digitalSignature and dataEncipherment values MUST NOT be present. That is, a public key intended to be employed only with a Composite ML-KEM algorithm MUST NOT also be employed for data encryption or for digital signatures. This requirement does not carry any particular security consideration; only the convention that KEM keys be identified with the keyEncipherment key usage.¶

8.4. SMIMECapabilities Attribute Conventions

Section 2.5.2 of [RFC8551] defines the SMIMECapabilities attribute to announce a partial list of algorithms that an S/MIME implementation can support. When constructing a CMS signed-data content type [RFC5652], a compliant implementation MAY include the SMIMECapabilities attribute that announces support for the RSA-OAEP Algorithm.¶

The SMIMECapability SEQUENCE representing a Composite ML-KEM Algorithm MUST include the appropriate object identifier as per Table 2 in the capabilityID field.¶

10.1. Object Identifier Allocations

EDNOTE to IANA: OIDs will need to be replaced in both the ASN.1 module and in Table 2.¶

11.1. KEM Combiner

EDNOTE: the exact text to put here depends on the outcome of the CFRG KEM Combiners and X-Wing discussion. If CFRG doesn't move fast enough for us, then we may need to leverage this security consideration directly on top of the X-Wing paper [X-Wing].¶

The following KEM combiner construction is as follows is used by both Composite-ML-KEM.Encap() and Composite-ML-KEM.Decap() and is split out here for easier analysis.¶

  KDF(mlkemSS || tradSS || tradCT || tradPK || Domain)

where:¶

• KDF(message) represents a key derivation function suitable to the chosen KEMs according to {tab-kem-combiners}. All KDFs produce a 256-bit shared secret key to match ML-KEM.¶

• mlkemSS is the shared secret key from the ML-KEM component.¶

• tradSS is the shared secret from the traditional component (elliptic curve or RSA).¶

• tradCT is the ciphertext from the traditional component (elliptic curve or RSA).¶

• tradPK is the public key of the traditional component (elliptic curve or RSA).¶

• Domain is the DER encoded value of the object identifier of the Composite ML-KEM algorithm as listed in Section 7.2.¶

• || represents concatenation.¶

Each registered Composite ML-KEM algorithm specifies the choice of KDF andDomain to be used in Section 7 and Section 7.2 below. Given that each Composite ML-KEM algorithm fully specifies the component algorithms, including for example the size of the RSA modulus, all inputs to the KEM combiner are fixed-size and thus do not require length-prefixing. The CompositeKEM.Decap() specified in Section 4.3 adds further error handling to protect the KEM combiner from malicious inputs.¶

The primary security property of the KEM combiner is that it preserves IND-CCA2 of the overall Composite ML-KEM so long as at least one component is IND-CCA2 X-Wing [GHP18]. Additionally, we also need to consider the case where one of the component algorithms is completely broken; that the private key is known to an attacker, or worse that the public key, private key, and ciphertext are manipulated by the attacker. In this case, we rely on the construction of the KEM combiner to ensure that the value of the other shared secret cannot be leaked or the combined shared secret predicted via manipulation of the broken algorithm. The following sections continue this discussion.¶

11.2. Decapsulation failure

Provided all inputs are well-formed, the key establishment procedure of ML-KEM will never explicitly fail. Specifically, the ML-KEM.Encaps and ML-KEM.Decaps algorithms from [FIPS.203] will always output a value with the same data type as a shared secret key, and will never output an error or failure symbol. However, it is possible (though extremely unlikely) that the process will fail in the sense that ML-KEM.Encaps and ML-KEM.Decaps will produce different outputs, even though both of them are behaving honestly and no adversarial interference is present. In this case, the sender and recipient clearly did not succeed in producing a shared secret key. This event is called a decapsulation failure. Estimates for the decapsulation failure probability (or rate) for each of the ML-KEM parameter sets are provided in Table 1 of [FIPS.203] and reproduced here in Table 6.¶

In the case of ML-KEM decapsulation failure, Composite ML-KEM MUST preserve the same behaviour and return a well-formed output.¶

11.3. Key Reuse

When using single-algorithm cryptography, the best practice is to always generate fresh keying material for each purpose, for example when renewing a certificate, or obtaining both a TLS and S/MIME certificate for the same device, however in practice key reuse in such scenarios is not always catastrophic to security and therefore often tolerated. With composite keys we have a much stricter security requirement. However this reasoning does not hold in the PQ / Traditional hybrid setting.¶

Within the broader context of PQ / Traditional hybrids, we need to consider new attack surfaces that arise due to the hybrid constructions and did not exist in single-algorithm contexts. One of these is key reuse where the component keys within a hybrid are also used by themselves within a single-algorithm context. For example, it might be tempting for an operator to take already-deployed RSA keys and add an ML-KEM key to them to form a hybrid. Within a hybrid signature context this leads to a class of attacks referred to as "stripping attacks" where one component signature can be extracted and presented as a single-algorithm signature. Hybrid KEMs using a concatenation-style KEM combiner, as is done in this document, do not have the analogous attack surface because even if an attacker is able to extract and decrypt one of the component ciphertexts, this will yield a different shared secret than the overall shared secret derived from the composite, so any subsequent symmetric cryptographic operations will fail. However there is still a risk of key reuse which relates to certificate revocation, as well as general key reuse security issues.¶

Upon receiving a new certificate enrollment request, many certification authorities will check if the requested public key has been previously revoked due to key compromise. Often a CA will perform this check by using the public key hash. Therefore, even if both components of a composite have been previously revoked, the CA may only check the hash of the combined composite key and not find the revocations. Therefore, it is RECOMMENDED to avoid key reuse and always generate fresh component keys for a new composite. It is also RECOMMENDED that CAs performing revocation checks on a composite key should also check both component keys independently.¶

11.4. Policy for Deprecated and Acceptable Algorithms

Traditionally, a public key or certificate contains a single cryptographic algorithm. If and when an algorithm becomes deprecated (for example, RSA-512, or SHA1), the path to deprecating it and removing it from operational environments is, at least is principle, straightforward.¶

In the composite model this is less obvious since implementers may decide that certain cryptographic algorithms have complementary security properties and are acceptable in combination even though one or both algorithms are deprecated for individual use. As such, a single composite public key or certificate may contain a mixture of deprecated and non-deprecated algorithms.¶

Since composite algorithms are registered independently of their component algorithms, their deprecation can be handled independently from that of their component algorithms. For example a cryptographic policy might continue to allow id-MLKEM512-ECDH-P256 even after ECDH-P256 is deprecated.¶

The Composite ML-KEM design specified in this document, and especially that of the KEM combiner specified in this document, and discussed in Section 11.1, means that the overall Composite ML-KEM algorithm should be considered to have the security strength of the strongest of its component algorithms; ie as long as one component algorithm remains strong, then the overall composite algorithm remains strong.¶

12.1. Normative References

[FIPS.180-4]

National Institute of Standards and Technology (NIST), "FIPS Publication 180-4: Secure Hash Standard", August 2015, <https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.180-4.pdf>.

[FIPS.202]

National Institute of Standards and Technology (NIST), "SHA-3 Standard: Permutation-Based Hash and Extendable-Output Functions", August 2015, <https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.202.pdf>.

[FIPS.203]

National Institute of Standards and Technology (NIST), "Module-Lattice-based Key-Encapsulation Mechanism Standard", August 2024, <https://nvlpubs.nist.gov/nistpubs/FIPS/NIST.FIPS.203.pdf>.

[I-D.ietf-lamps-cms-sha3-hash]

Housley, R., "Use of the SHA3 One-way Hash Functions in the Cryptographic Message Syntax (CMS)", Work in Progress, Internet-Draft, draft-ietf-lamps-cms-sha3-hash-04, 16 May 2024, <https://datatracker.ietf.org/doc/html/draft-ietf-lamps-cms-sha3-hash-04>.

[I-D.ietf-lamps-kyber-certificates]

Turner, S., Kampanakis, P., Massimo, J., and B. Westerbaan, "Internet X.509 Public Key Infrastructure - Algorithm Identifiers for the Module-Lattice-Based Key-Encapsulation Mechanism (ML-KEM)", Work in Progress, Internet-Draft, draft-ietf-lamps-kyber-certificates-06, 4 November 2024, <https://datatracker.ietf.org/doc/html/draft-ietf-lamps-kyber-certificates-06>.

[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/info/rfc2119>.

[RFC4055]

Schaad, J., Kaliski, B., and R. Housley, "Additional Algorithms and Identifiers for RSA Cryptography for use in the Internet X.509 Public Key Infrastructure Certificate and Certificate Revocation List (CRL) Profile", RFC 4055, DOI 10.17487/RFC4055, June 2005, <https://www.rfc-editor.org/info/rfc4055>.

[RFC5280]

Cooper, D., Santesson, S., Farrell, S., Boeyen, S., Housley, R., and W. Polk, "Internet X.509 Public Key Infrastructure Certificate and Certificate Revocation List (CRL) Profile", RFC 5280, DOI 10.17487/RFC5280, May 2008, <https://www.rfc-editor.org/info/rfc5280>.

[RFC5480]

Turner, S., Brown, D., Yiu, K., Housley, R., and T. Polk, "Elliptic Curve Cryptography Subject Public Key Information", RFC 5480, DOI 10.17487/RFC5480, March 2009, <https://www.rfc-editor.org/info/rfc5480>.

[RFC5652]

Housley, R., "Cryptographic Message Syntax (CMS)", STD 70, RFC 5652, DOI 10.17487/RFC5652, September 2009, <https://www.rfc-editor.org/info/rfc5652>.

[RFC5869]

Krawczyk, H. and P. Eronen, "HMAC-based Extract-and-Expand Key Derivation Function (HKDF)", RFC 5869, DOI 10.17487/RFC5869, May 2010, <https://www.rfc-editor.org/info/rfc5869>.

[RFC5958]

Turner, S., "Asymmetric Key Packages", RFC 5958, DOI 10.17487/RFC5958, August 2010, <https://www.rfc-editor.org/info/rfc5958>.

[RFC6234]

Eastlake 3rd, D. and T. Hansen, "US Secure Hash Algorithms (SHA and SHA-based HMAC and HKDF)", RFC 6234, DOI 10.17487/RFC6234, May 2011, <https://www.rfc-editor.org/info/rfc6234>.

[RFC8017]

Moriarty, K., Ed., Kaliski, B., Jonsson, J., and A. Rusch, "PKCS #1: RSA Cryptography Specifications Version 2.2", RFC 8017, DOI 10.17487/RFC8017, November 2016, <https://www.rfc-editor.org/info/rfc8017>.

[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/info/rfc8174>.

[RFC8410]

Josefsson, S. and J. Schaad, "Algorithm Identifiers for Ed25519, Ed448, X25519, and X448 for Use in the Internet X.509 Public Key Infrastructure", RFC 8410, DOI 10.17487/RFC8410, August 2018, <https://www.rfc-editor.org/info/rfc8410>.

[RFC8411]

Schaad, J. and R. Andrews, "IANA Registration for the Cryptographic Algorithm Object Identifier Range", RFC 8411, DOI 10.17487/RFC8411, August 2018, <https://www.rfc-editor.org/info/rfc8411>.

[RFC8619]

Housley, R., "Algorithm Identifiers for the HMAC-based Extract-and-Expand Key Derivation Function (HKDF)", RFC 8619, DOI 10.17487/RFC8619, June 2019, <https://www.rfc-editor.org/info/rfc8619>.

[RFC9629]

Housley, R., Gray, J., and T. Okubo, "Using Key Encapsulation Mechanism (KEM) Algorithms in the Cryptographic Message Syntax (CMS)", RFC 9629, DOI 10.17487/RFC9629, August 2024, <https://www.rfc-editor.org/info/rfc9629>.

[SP.800-185]

National Institute of Standards and Technology (NIST), "SHA-3 Derived Functions: cSHAKE, KMAC, TupleHash, and ParallelHash", December 2016, <https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-185.pdf>.

[SP.800-56Ar3]

National Institute of Standards and Technology (NIST), "Recommendation for Pair-Wise Key-Establishment Schemes Using Discrete Logarithm Cryptography", April 2018, <https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-56Ar3.pdf>.

[SP.800-56Cr2]

National Institute of Standards and Technology (NIST), "Recommendation for Key-Derivation Methods in Key-Establishment Schemes", August 2020, <https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-56Cr2.pdf>.

[SP.800-57pt1r5]

National Institute of Standards and Technology (NIST), "Recommendation for Key Management: Part 1 – General", May 2020, <https://nvlpubs.nist.gov/nistpubs/SpecialPublications/NIST.SP.800-57pt1r5.pdf>.

[X.690]

ITU-T, "Information technology - ASN.1 encoding Rules: Specification of Basic Encoding Rules (BER), Canonical Encoding Rules (CER) and Distinguished Encoding Rules (DER)", ISO/IEC 8825-1:2015, November 2015.

12.2. Informative References

[ANSSI2024]

French Cybersecurity Agency (ANSSI), Federal Office for Information Security (BSI), Netherlands National Communications Security Agency (NLNCSA), and Swedish National Communications Security Authority, Swedish Armed Forces, "Position Paper on Quantum Key Distribution", n.d., <https://cyber.gouv.fr/sites/default/files/document/Quantum_Key_Distribution_Position_Paper.pdf>.

[Aviram22]

Yogev, E., "Practical (Post-Quantum) Key Combiners from One-Wayness and Applications to TLS", n.d., <https://eprint.iacr.org/2022/065>.

[BSI2021]

Federal Office for Information Security (BSI), "Quantum-safe cryptography - fundamentals, current developments and recommendations", October 2021, <https://www.bsi.bund.de/SharedDocs/Downloads/EN/BSI/Publications/Brochure/quantum-safe-cryptography.pdf>.

[CNSA2.0]

"Commercial National Security Algorithm Suite 2.0", n.d., <https://media.defense.gov/2022/Sep/07/2003071834/-1/-1/0/CSA_CNSA_2.0_ALGORITHMS_.PDF>.

[FIPS-140-3-IG]

National Institute of Standards and Technology (NIST), "Implementation Guidance for FIPS 140-3 and the Cryptographic Module Validation Program", July 2024, <https://csrc.nist.gov/csrc/media/Projects/cryptographic-module-validation-program/documents/fips%20140-3/FIPS%20140-3%20IG.pdf>.

[GHP18]

Poettering, B., "KEM Combiners", 2018, <https://eprint.iacr.org/2018/024>.

[I-D.ietf-lamps-cms-kyber]

Prat, J., Ounsworth, M., and D. Van Geest, "Use of ML-KEM in the Cryptographic Message Syntax (CMS)", Work in Progress, Internet-Draft, draft-ietf-lamps-cms-kyber-05, 15 October 2024, <https://datatracker.ietf.org/doc/html/draft-ietf-lamps-cms-kyber-05>.

[I-D.ietf-pquip-pqt-hybrid-terminology]

D, F., P, M., and B. Hale, "Terminology for Post-Quantum Traditional Hybrid Schemes", Work in Progress, Internet-Draft, draft-ietf-pquip-pqt-hybrid-terminology-04, 10 September 2024, <https://datatracker.ietf.org/doc/html/draft-ietf-pquip-pqt-hybrid-terminology-04>.

[I-D.ietf-tls-hybrid-design]

Stebila, D., Fluhrer, S., and S. Gueron, "Hybrid key exchange in TLS 1.3", Work in Progress, Internet-Draft, draft-ietf-tls-hybrid-design-04, 11 January 2022, <https://datatracker.ietf.org/doc/html/draft-ietf-tls-hybrid-design-04>.

[RFC2986]

Nystrom, M. and B. Kaliski, "PKCS #10: Certification Request Syntax Specification Version 1.7", RFC 2986, DOI 10.17487/RFC2986, November 2000, <https://www.rfc-editor.org/info/rfc2986>.

[RFC4210]

Adams, C., Farrell, S., Kause, T., and T. Mononen, "Internet X.509 Public Key Infrastructure Certificate Management Protocol (CMP)", RFC 4210, DOI 10.17487/RFC4210, September 2005, <https://www.rfc-editor.org/info/rfc4210>.

[RFC4211]

Schaad, J., "Internet X.509 Public Key Infrastructure Certificate Request Message Format (CRMF)", RFC 4211, DOI 10.17487/RFC4211, September 2005, <https://www.rfc-editor.org/info/rfc4211>.

[RFC4262]

Santesson, S., "X.509 Certificate Extension for Secure/Multipurpose Internet Mail Extensions (S/MIME) Capabilities", RFC 4262, DOI 10.17487/RFC4262, December 2005, <https://www.rfc-editor.org/info/rfc4262>.

[RFC5083]

Housley, R., "Cryptographic Message Syntax (CMS) Authenticated-Enveloped-Data Content Type", RFC 5083, DOI 10.17487/RFC5083, November 2007, <https://www.rfc-editor.org/info/rfc5083>.

[RFC5639]

Lochter, M. and J. Merkle, "Elliptic Curve Cryptography (ECC) Brainpool Standard Curves and Curve Generation", RFC 5639, DOI 10.17487/RFC5639, March 2010, <https://www.rfc-editor.org/info/rfc5639>.

[RFC5914]

Housley, R., Ashmore, S., and C. Wallace, "Trust Anchor Format", RFC 5914, DOI 10.17487/RFC5914, June 2010, <https://www.rfc-editor.org/info/rfc5914>.

[RFC5990]

Randall, J., Kaliski, B., Brainard, J., and S. Turner, "Use of the RSA-KEM Key Transport Algorithm in the Cryptographic Message Syntax (CMS)", RFC 5990, DOI 10.17487/RFC5990, September 2010, <https://www.rfc-editor.org/info/rfc5990>.

[RFC6090]

McGrew, D., Igoe, K., and M. Salter, "Fundamental Elliptic Curve Cryptography Algorithms", RFC 6090, DOI 10.17487/RFC6090, February 2011, <https://www.rfc-editor.org/info/rfc6090>.

[RFC7292]

Moriarty, K., Ed., Nystrom, M., Parkinson, S., Rusch, A., and M. Scott, "PKCS #12: Personal Information Exchange Syntax v1.1", RFC 7292, DOI 10.17487/RFC7292, July 2014, <https://www.rfc-editor.org/info/rfc7292>.

[RFC7296]

Kaufman, C., Hoffman, P., Nir, Y., Eronen, P., and T. Kivinen, "Internet Key Exchange Protocol Version 2 (IKEv2)", STD 79, RFC 7296, DOI 10.17487/RFC7296, October 2014, <https://www.rfc-editor.org/info/rfc7296>.

[RFC8446]

Rescorla, E., "The Transport Layer Security (TLS) Protocol Version 1.3", RFC 8446, DOI 10.17487/RFC8446, August 2018, <https://www.rfc-editor.org/info/rfc8446>.

[RFC8551]

Schaad, J., Ramsdell, B., and S. Turner, "Secure/Multipurpose Internet Mail Extensions (S/MIME) Version 4.0 Message Specification", RFC 8551, DOI 10.17487/RFC8551, April 2019, <https://www.rfc-editor.org/info/rfc8551>.

[RFC9180]

Barnes, R., Bhargavan, K., Lipp, B., and C. Wood, "Hybrid Public Key Encryption", RFC 9180, DOI 10.17487/RFC9180, February 2022, <https://www.rfc-editor.org/info/rfc9180>.

[SP800-131Ar2]

Barker, E. and A. Roginksy, "Transitioning the Use of Cryptographic Algorithms and Key Lengths", n.d., <https://nvlpubs.nist.gov/nistpubs/specialpublications/nist.sp.800-131ar2.pdf>.

[X-Wing]

Barbosa, M., Connolly, D., Duarte, J., Kaiser, A., Schwabe, P., Varner, K., and B. Westerbaan, "X-Wing The Hybrid KEM You’ve Been Looking For", 9 January 2024, <https://eprint.iacr.org/2024/039.pdf>.

D.1. FIPS Certification

TODO -- update this once NIST SP 800-227 is published.¶

EDNOTE At time of writing, it is unclear that the KEM combiner presented in this document would pass a FIPS certification. The SHA3 instantiations should pass under SP 800-56Cr2 Option 1, but the HKDF-SHA2 combinations are technically not allowed under SP 800-56Cr2 even though Option 2 allows HMAC-based constructions, but unfortunately only HKDF-Extract is FIPS-allowed, not HKDF-Expand. The authors have been in contact with NIST to ensure compatibility either by NIST making SP 800-227 more flexible, or by changing this specification to only use HKDF-Extract.¶

One of the primary design goals of this specification is for the overall composite algorithm to be able to be considered FIPS-approved even when one of the component algorithms is not. Implementers seeking FIPS certification of a Composite ML-KEM algorithm where only one of the component algorithms has been FIPS-validated or FIPS-approved should credit the FIPS-validated component algorithm with full security strength, the non-FIPS-validated component algorithm with zero security, and the overall composite should be considered full strength and thus FIPS-approved.¶

The authors wish to note that this gives composite algorithms great future utility both for future cryptographic migrations as well as bridging across jurisdictions; for example defining composite algorithms which combine FIPS cryptography with cryptography from a different national standards body.¶

D.2. Backwards Compatibility

As noted in the introduction, the post-quantum cryptographic migration will face challenges in both ensuring cryptographic strength against adversaries of unknown capabilities, as well as providing ease of migration. The composite mechanisms defined in this document primarily address cryptographic strength, however this section contains notes on how backwards compatibility may be obtained.¶

The term "ease of migration" is used here to mean that existing systems can be gracefully transitioned to the new technology without requiring large service disruptions or expensive upgrades. The term "backwards compatibility" is used here to mean something more specific; that existing systems as they are deployed today can inter-operate with the upgraded systems of the future.¶

These migration and interoperability concerns need to be thought about in the context of various types of protocols that make use of X.509 and PKIX with relation to key establishment and content encryption, from online negotiated protocols such as TLS 1.3 [RFC8446] and IKEv2 [RFC7296], to non-negotiated asynchronous protocols such as S/MIME signed email [RFC8551], as well as myriad other standardized and proprietary protocols and applications that leverage CMS [RFC5652] encrypted structures.¶

D.3. Decapsulation Requires the Public Key

ML-KEM always requires the public key in order to perform various steps of the Fujisaki-Okamoto decapsulation [FIPS.203], and for this reason the private key encoding specified in FIPS 203 includes the public key. Therefore it is not required to carry it in the OneAsymmetricKey.publicKey field, which remains optional, but is strictly speaking redundant since an ML-KEM public key can be parsed from an ML-KEM private key, and thus populating the OneAsymmetricKey.publicKey field would mean that two copies of the public key data are transmitted.¶

With regard to the traditional algorithms, RSA or Elliptic Curve, in order to achieve the public-key binding property the KEM combiner used to form the Composite ML-KEM, the combiner requires the traditional public key as input to the KDF that derives the output shared secret. Therefore it is required to carry the public key within the respective OneAsymmetricKey.publicKey as per the private key encoding given in Section 5.2. Implementers who choose to use a different private key encoding than the one specified in this document MUST consider how to provide the component public keys to the decapsulate routine. While some implementations might contain routines to computationally derive the public key from the private key, it is not guaranteed that all implementations will support this; for this reason the interoperable composite private key format given in this document in Section 5.2 requires the public key of the traditional component to be included.¶