]> UK National Cyber Security Centre

Ben.S3@ncsc.gov.uk

UK National Cyber Security Centre

Adam.R@ncsc.gov.uk

UK National Cyber Security Centre

Jonathan.C@ncsc.gov.uk

Security IPSECME ipsec sha-3 ikev2 kmac This document specifies the use of KMAC128 and KMAC256 within the Internet Key Exchange Version 2 (IKEv2), Encapsulating Security Payload (ESP), and Authentication Header (AH) protocols. These algorithms can be used as integrity protection algorithms for ESP, AH and IKEv2, and as Pseudo-Random Functions (PRFs) for IKEv2. Requirements for supporting signature algorithms in IKEv2 that use SHA3-256, SHA3-384, SHA3-512, SHAKE128 and SHAKE256 are also specified.

Changes in draft-ietf-ipsecme-sha3-01 Doubled the preferred key size of KMAC128 and KMAC256 from 128/256 to 256/512 to align with . This doesn't add any additional security, but does mean that the preferred key size matches the PRF output length. In places where IKEv2 would normally call prf+, KMAC is now used directly to produce as much key material as needed. This differs from the approach described in prior versions of this document, where prf+ was still used, but only the first iteration was ever needed. This aligns the document with . Added editorial notes about the use of customization strings for KMAC. This deviates from , but comments at IETF 123 suggested that further discussion is needed before committing to not using customization strings. Expanded editorial note around INTEG transforms, as the utility of these transforms was questioned at IETF 123.

Introduction specifies both the SHA3-256, SHA3-384 and SHA3-512 cryptographic hash functions, and the SHAKE eXtendable-output functions (XOFs). specifies KMAC128 and KMAC256, which use variants of SHAKE128 and SHAKE256 respectively to create a message authentication code (MAC). Like the output of SHAKE, the MAC output of KMAC can be of any length required by the application. This document specifies how to use KMAC128 and KMAC256 with IKEv2 and IPsec for integrity protection, and as a PRF for IKEv2. Use of KMAC as a PRF and in prf+ is aligned with the framework for use of XOFs in IKEv2 specified in . It also allocates values used for announcing support for SHA3-256, SHA3-384, SHA3-512, SHAKE128, and SHAKE256 when generating and validating signatures in IKEv2. EDNOTE: Support for SHA3-224 is not included, though draft-ietf-lamps-cms-sha3-hash includes support for SHA3-224 with ECDSA.

Conventions and Definitions 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 when, and only when, they appear in all capitals, as shown here. Additionally, this document uses several terms to collectively refer to sets of algorithms. The term "SHA-3 cryptographic hash functions" is used to collectively refer to SHA3-256, SHA3-384 and SHA3-512. The term "KMAC" is used to collectively refer to KMAC128 and KMAC256. The term "SHA-3" (without any other qualifiers) is used to collectively refer to the cryptographic algorithms defined in and . The term "SHA-2" (without any other qualifiers) is used to collectively refer to SHA-256, SHA-384, and SHA-512. The term "SHAKE" is used to collectively refer to SHAKE128 and SHAKE256.

SHA-3 and Keccak SHA-3 is a collection of cryptographic algorithms that all utilise the Keccak sponge construction. describes the SHA-3 cryptographic hash functions, which produce a fixed-length digest for any length of input. These hash functions are intended to be used in the same manner and contexts as other traditional hash functions such as SHA-2. also describes the SHAKE XOFs. An XOF differs from a traditional hash function in that the length of the XOF's output can be chosen by the application that uses it. describes cSHAKE, a customisable version of SHAKE, and KMAC, which is a PRF and keyed hash function that utilises cSHAKE. Like SHAKE and cSHAKE, the length of KMAC's output is application-dependent. SHA-3 was specified to provide applications with an alternative to SHA-2, which is based on the Merkle-Damgård construction. Use of the Merkle-Damgård construction in SHA-2 means that length extension attacks are possible if SHA-2 isn't used correctly. At the time of writing, use of SHA-2 in IPsec is believed to be secure, and hence there is no security motivation to migrate away from SHA-2 to SHA-3 in this context. However, in the event that a significant attack on SHA-2 is discovered, SHA-3 will be an immediately viable alternative. Migration to use of post-quantum algorithms in IKEv2 may make use of SHA-3 more appealing for minimal implementations of IPsec, as , , , and all make use of SHA-3 internally. Since support for SHA-3 is required to implement these algorithms, some implementers may find it preferable to implement SHA-3, and only SHA-3, if interoperability with general-purpose IKEv2 and IPsec implementations is not required. SHA-2 is used with HMAC in IKEv2 and IPsec. Whilst SHA-3 can be used with HMAC, KMAC is more efficient as it directly uses the Keccak sponge function to produce a MAC, rather than treating Keccak as a traditional cryptographic hash function and then feeding that hash function into a separate MAC algorithm. Therefore, this document only specifies use of KMAC, and not of HMAC-SHA3.

KMAC API The basic API for KMAC is defined below. The symbols used in this API description conform to those used for prf+ in , which clash with the API described in . uses S to describe the input string to prf+, whereas uses that symbol for the optional customization string. KMAC implementations used in IKEv2 and IPsec do not need to conform to these APIs exactly; they are merely used in this document for illustrative purposes. For the purposes of this document, the API for KMAC is defined as: KMAC(K, S, L, C) -> Z where: K is the key. It is a bit string of length between zero and (2^2040)-1 bits, inclusively. S is the input string. It is a bit string of any length, including zero. L is an integer representing the requested output length in bits. This parameter is typically fixed in the context of IPsec, except when extracting key material using prf+ in IKEv2, where it depends on the length of key material needed by the negotiated cipher suite. C is an optional customization string. It is a bit string of length between zero and (2^2040)-1 bits, inclusively. Z is the output string of KMAC, which is a message authentication code. It is a bit string of length L.

Constraints on KMAC inputs and outputs Per , the length of the key input K input to KMAC MUST be less than 2^2040 bits. In the context of IKEv2 and IPsec, there is no situation where a key that long would be expected. Initiator and Responder nonces Ni and Nr are used as inputs to IKEv2 PRF calls, although the length of these nonces combined cannot exceed 4096 bits. Pre-shared keys used for authentication in IKEv2 are used as keys with PRFs negotiated by IKEv2, and have no upper bound on their length. Therefore, KMAC implementations used with IKEv2 MUST at minimum accept values of K up to and including 4096 bits in length. Implementations MAY restrict the size of pre-shared key inputs such that they do not exceed 4096 bits. There is no algorithm-defined minimum size for the key input to KMAC, but and list the recommended key sizes to be used within the context of IKEv2 and IPsec, aligned to the security strength of each algorithm. Using a key smaller than the security strength of the chosen KMAC algorithm undermines the security properties of that algorithm. Where IKEv2 is used to create security associations, the size of most PRF keys is automatically managed at the protocol level, and there is no risk of selecting an undersized key in these cases. However, the size of keys used for PRFs in IKEv2 cannot always be controlled. As an example, in the case of pre-shared keys used for authentication or protection against a quantum computer as in , those secrets are used as the key input to a PRF negotiated by IKEv2. Those pre-shared keys could be arbitrarily chosen by a user or derived from a password rather than securely generated, even though strongly discourages this practice. Keys chosen in this manner may be shorter than any recommended key size. IKEv2 implementations following the recommendation laid out in can impose constraints on suitable pre-shared keys. Additionally, Ni and Nr are variable length and are used as the key for KMAC. states that each of these nonces MUST be at least 128 bits in size, and MUST be at least half the preferred key size for the negotiated PRF. If an IKEv2 peer sends an undersized nonce, the message containing that nonce can be rejected in the same way as any malformed IKEv2 message would be. Conformant KMAC implementations SHOULD reject keys that do not meet the security strength of the corresponding algorithm. The input string S can be a variety of lengths in practice, but in the context of IKE and IPsec the length will always be a multiple of eight, as IKE and IPsec only operate on entire octets. Similarly, KMAC's output length parameter L will always be a multiple of eight. Since the length of output required from KMAC is always known in advance, KMAC with arbitrary-length output as described in Section 4.3.1 of is never used, and thus L is never set to 0. KMAC's customization string C is fixed to a specific value depending on the context in which KMAC is used. Future specifications may define additional customization strings, but the set of valid strings used by KMAC in IKEv2 and IPsec will always be fixed-length context-dependent strings specified in IETF RFCs rather than dynamically created, e.g. via random data. EDNOTE: This is a deviation from . If working group consensus is that customization strings shouldn't be used, this section will specify that customization string input is null.

Padding Since the length of the input string S for KMAC varies, and KMAC operates on fixed-size input blocks, padding is required to use KMAC in IKEv2 and IPsec. The padding scheme for KMAC is specified in . A KMAC implementation conformant to that document is sufficient; no additional padding is required to use these algorithms in IKEv2 or IPsec.

KMAC Key Padding When KMAC is used as the PRF for an IKE SA, the size of the key input K is variable. If the size of a KMAC key is greater than the preferred key size as shown in , the key is used in its entirety without any kind of shortening or truncation. As described in , keys are always padded up to a multiple of the rate of the underlying Keccak sponge function; that is, 168 bytes and 136 bytes for KMAC-128 and KMAC-256 respectively. Any KMAC implementation conformant with is suitable for use in IKEv2 and IPsec; no protocol-specific additional padding of keys is required.

Parameters and security strengths for KMAC describes the general properties of KMAC, with the HMAC-SHA2 algorithms also listed for comparison purposes. The maximum security strengths listed are taken from . Note that these are maximum security strengths. Using keys that are shorter than the maximum security strength will constrain the maximum security strength of the chosen algorithm to be no higher than the length of that key. Keys that contain insufficient entropy to meet the maximum security strength constrain the maximum security of the chosen algorithm to be no higher than the bits of entropy represented in the key. Algorithm Name Output Length (bits) Maximum Security Strength (bits) HMAC-SHA-256 256 >=256 HMAC-SHA-384 384 >=256 HMAC-SHA-512 512 >=256 KMAC128 Variable 128 KMAC256 Variable >=256 presents the parameters of KMAC when it is used as a PRF in IKEv2, with the HMAC-SHA2 algorithms also listed for comparison purposes. Algorithm Name PRF variant Preferred Key Size (bits) Output Length (bits) HMAC-SHA-256 PRF_HMAC_SHA2_256 256 256 HMAC-SHA-384 PRF_HMAC_SHA2_384 384 384 HMAC-SHA-512 PRF_HMAC_SHA2_512 512 512 KMAC128 PRF_KMAC_128 256 256, or length of output required for prf+ KMAC256 PRF_KMAC_256 512 512, or length of output required for prf+ The security strength of these algorithms is equal to the maximum security strength indicated for that algorithm in , unless the entropy of the supplied key is insufficient to meet that strength. Note that the preferred key size for both KMAC variants is double the security strength of each algorithm. This is to align with the requirement in that, when an XOF is used as a PRF, the preferred key size must match the output length of the PRF. When key material is extracted from IKEv2's prf+ Key Derivation Function (KDF) for use with KMAC as a PRF in IKEv2, the length of keys extracted MUST conform to the preferred key sizes listed in . EDNOTE: The KMAC output lengths have been aligned with HMAC, but if we're not depending on collision resistance, could they be reduced to 128/256 bits respectively? This would also mean that the key size could be reduced back down to 128/256 bits; they only need to be 256/512 bits to maintain conformance with . presents the parameters of KMAC when it is used for data origin authentication and integrity protection in IKEv2 and IPsec, with the HMAC-SHA2 algorithms also listed for comparison purposes. Algorithm Name Integrity variant Key Size (bits) Output Length (bits) HMAC-SHA-256 AUTH_HMAC_SHA2_256_128 256 128 HMAC-SHA-384 AUTH_HMAC_SHA2_384_192 384 192 HMAC-SHA-512 AUTH_HMAC_SHA2_512_256 512 256 KMAC128 AUTH_KMAC_128 128 128 KMAC256 AUTH_KMAC_256 256 256 When used for authentication and integrity protection, KMAC message authentication codes are produced using a smaller value for the "requested output length" parameter L. In this case, the security strength of each given algorithm is constrained by its output length. When key material is extracted from IKEv2's prf+ KDF for use with KMAC for data origin authentication and integrity protection in IKEv2 or IPsec, the length of keys extracted MUST conform to the key sizes listed in .

KMAC as a PRF in IKEv2 IKEv2 Security Associations (SAs) make use of a PRF for authentication purposes, and as a part of the prf+ Key Derivation Function (KDF). KMAC can act as the PRF for an IKE SA, but it is treated slightly differently to existing PRFs as it is capable of producing different output lengths depending on the context in which it is used. With KMAC, the key K is either a fixed-length key (such as SK_d) that is the preferred key size for the variant of KMAC being used, or the length of K is dependent on other factors. When the PRF is used with nonce inputs as the key K (e.g. when generating SKEYSEED), or when the PRF is used with a pre-shared key as the input key K, the length of K depends on implementation-specific details, user configuration options, etc. When KMAC is used as the PRF for an IKE SA, its "requested output length" parameter L and "customization string" parameter C are populated differently depending on whether KMAC is being used as a part of the prf+ construction in the IKEv2 KDF or not. This process is described in more detail in and .

KMAC as a PRF When used in IKEv2 as a PRF outside the prf+ construction, KMAC's output length L is 128 for KMAC128, and 256 for KMAC256. That is, the output length is the same size as the security strength and preferred key size of the given KMAC algorithm. Note that the situation is different when KMAC is used within the prf+ construction, as described in . When KMAC is used as a PRF outside the prf+ construction, the customization string C is set to the ASCII character string "ikev2 prf", without null termination. EDNOTE: This is a deviation from . If working group consensus is that customization strings shouldn't be used, this section will specify that customization string input is null.

KMAC in prf+ defines an iterative function called "prf+" that can be used with a PRF to produce a pseudorandom stream of arbitrary length. Typically prf+ is used to derive SA keys from SKEYSEED or SK_d. However, as described in , prf+ isn't necessary when using an XOF, as a single call to the XOF can produce as many pseudorandom bits are as needed. As such, prf+ MUST NOT be used when KMAC128 or KMAC256 have been negotiated. Instead, KMAC is called once, and the L input parameter is set to the length of the keying material required. When KMAC is used where prf+ would normally be used, the customization string C is set to the ASCII character string "ikev2 kdf", without null termination. EDNOTE: specifies that additional inputs like KMAC's customization string mustn't be used. This version of the draft still includes use of the customization string, as feedback at IETF 123 indicated some support for this approach. Use of the customization string would provide domain separation for uses of the same key across standard PRF calls and prf+. For instance, given that KMAC replaces prf+ entirely: SKEYSEED = prf(SK_d (old), g^ir (new) | Ni | Nr) and KEYMAT = prf+(SK_d, g^ir (new) | Ni | Nr) would produce the same value with the same inputs, whereas the same wouldn't be true of HMAC. This isn't obviously a security problem, but use of customization strings would remove any risk entirely.

KMAC for integrity protection in ESP, AH and IKEv2 IKE and IPsec SAs can make use of an integrity protection algorithm to provide data origin authentication and integrity protection services. KMAC can be used to provide these services. As described in , Authenticated Encryption with Associated Data (AEAD) ciphers are the fastest and most modern approach to providing these services in conjunction with confidentiality protection. KMAC MUST NOT be negotiated in IKEv2 in conjunction with an AEAD cipher. KMAC MAY be used as an integrity protection algorithm with: ESP in conjunction with a non-AEAD cipher ESP and null encryption (ENCR_NULL) IKEv2 in conjunction with a non-AEAD cipher AH EDNOTE BEGINS In general RFC 8247 and RFC 8221 recommend an AEAD algorithm, not using split ENCR and INTEG transforms. However, there are two cases where an INTEG transform is useful. The first is "authenticated-only communications without encryption, such as ESP with NULL encryption or AH communications" as described in RFC 8221. IKE always requires encryption, so this case only applies to Child SAs. Negotiating the ENCR_NULL transform with an INTEG transform is the preferred means to achieve this. Use of AH is an alternative, but not encouraged. The second is where a non-AEAD cipher is used for encryption. Most modern IPsec implementations (i.e. those that can be modified to support KMAC) shouldn't be doing this. However, RFC 8221 cites IoT devices that can't hold state as a motivation for continuing to use AES_ENCR_CBC, as risk of reuse of nonces for GCM, CCM or ChaCha20Poly1305 is high. So there is a niche for INTEG transforms, it's just very limited. They're simple to specify; is it worth it to specify them here, or is having a "new" INTEG transform likely to confuse users? We could recommend that KMAC not be used with AH, and recommend that an AEAD cipher is used where feasible. Is this the right place to make that recommendation, or should that be in an update to RFC 8247/RFC 8221? EDNOTE ENDS When using KMAC, the L input parameter is always set to the same value as the key size and security strength of the chosen KMAC algorithm. That is, the output length of KMAC128 is always set to 128 bits, and the output length of KMAC256 is always set to 256 bits. When used with ESP or AH, the "customization string" parameter C is set to the ASCII character string "ipsec integ", without null termination. When used with IKEv2 for integrity protection, the "customization string" parameter C is set to the ASCII character string "ikev2 integ", without null termination. EDNOTE: specifies that customization strings must be null for use of XOFs with prf/prf+, although it doesn't say anything about ESP/AH. See for further discussion as to why we're proposing using customization strings more generally. They're perhaps less important here as ESP/AH keys are only used to encrypt ESP/AH packets, whereas keys in IKE tend to be used for a wider range of purposes. If working group consensus is that customization strings shouldn't be used, this section will specify that customization string input is null.

Use of SHA-3 in the Digital Signature authentication method in IKEv2 SHAKE and the SHA-3 cryptographic hash functions can generate digests for use with signature algorithms. For instance, specifies algorithm identifiers for using RSASSA-PSS and ECDSA with SHAKE, and NIST has assigned OIDs for using RSA PKCS #1 v1.5 signatures with SHA-3 . specifies the "Digital Signature" (14) authentication method, that allows IKEv2 to support any signature algorithm without the need to specify an authentication method for every new combination of signature algorithm and hash function. The Digital Signature authentication method is the only way to utilise SHA-3 with signatures in IKEv2, so if a peer uses SHA-3 in this context, it MUST specify the Digital Signature authentication method in its corresponding AUTH payload. The Digital Signature authentication method specifies use of a SIGNATURE_HASH_ALGORITHMS notification by each IKEv2 peer to announce the hash functions it supports for use with signatures. This specification defines values in for announcing support for SHA-3 algorithms in the SIGNATURE_HASH_ALGORITHMS notification. When an IKEv2 implementation supports SHA-3 in this context, and local policy permits use of SHA-3 to generate or verify signatures, it MUST include the corresponding values in its SIGNATURE_HASH_ALGORITHMS notification.

Considerations for implementers and for future IPsec documents KMAC's API differs from most PRF and Integrity Algorithm transforms as described in . Care should be taken with the output length parameter L in particular, as its behavior can be counter-intuitive when compared to other integrity algorithms where a truncated output is used as an authenticator value. Unlike SHAKE, KMAC outputs of different lengths are not related. That is, the value of L is factored into the calculation of the output value, and thus all bits of the output value are affected. This means that implementations cannot simply discard a portion of a KMAC output as a substitute for requesting the correct value for L. For example, a KMAC256 implementation used as the PRF transform for IKEv2 cannot be used as an Integrity transform simply by discarding bits from that implementation's output; a different value of L must be supplied. shows an example for the case where the key, input and (in the case of KMAC) customization string are all the empty string. Transform Name Output Length (bits) Output (hex) PRF_HMAC_SHA2_256 256 b613679a0814d9ec772f95d778c35fc5... AUTH_HMAC_SHA2_256_128 128 b613679a0814d9ec772f95d778c35fc5 PRF_KMAC_128 256 5c135c615152fb4d9784dd1155f9b603... AUTH_KMAC_128 128 e6aff27fef95903eb939bc3745730d34 This is also true for prf+ when used with KMAC. Typically prf+ is used iteratively to obtain at least as much key material as is required, and the amount of key material obtained will be a multiple of the output size of the negotiated PRF. This process will often produce more key material than required, and the excess is simply discarded. When KMAC is used, the amount of key material required is supplied to the PRF, and exactly the right amount of key material will be returned. Requesting more key material than required and discarding any excess will produce different keys, and interoperability will not be possible. Future authors of IPsec documents should be careful to consider whether related outputs from a PRF are required for their specification, and if so, describe how to handle KMAC and similar PRFs.

Security Considerations SHA-3 and SHA-2 are both believed to be secure at time of writing. Views on the security of cryptographic algorithms evolve over time, so implementers should pay attention to IETF RFCs reporting on recommendations for the use of cryptographic algorithms in IKEv2 and IPsec, such as any documents that update and . Quantum computing has a significant impact on the security of all IETF security protocols, as a cryptographically-relevant quantum computer (CRQC) could use Shor's algorithm to break many traditional asymmetric cryptographic algorithms. A CRQC can theoretically also attack hash functions, including SHA-3 and SHA-2, using Grover's algorithm. However, the impact of Grover's algorithm is significantly less dramatic than the impact of Shor's Algorithm. The worst-case impact of Grover's algorithm would be a reduction in security strength by a factor of two. However, since Grover's algorithm is difficult to parallelise, the security reduction for SHA-3 and SHA-2 caused by Grover's algorithm is expected to be far less significant in practice. See for a discussion on the practical cost of using Grover's algorithm to attack hash functions. The security properties offered by KMAC depend on limiting access to the keys used with those algorithms. Since both algorithms depend on a symmetric key, the key must be known by at least two parties in order to be useful. Sharing the key beyond two parties may erode the security offered by these algorithms. In the case of IKEv2 and IPsec, this typically means that access to keys must be limited to the peers participating in the security association that uses those keys. IKEv2 can be used to enforce this for IPsec SAs and most keys used in IKE SAs, but pre-shared keys are a notable exception here. Providing more than two peers with access to a single pre-shared key may undermine the security offered by that pre-shared key, and hence the security offered by KMAC. When IKEv2 is used to create IPsec SAs, the keys for KMAC are all ultimately derived from an ephemeral shared secret produced using one or more negotiated key exchange algorithms, with the exception of static pre-shared keys used in IKEv2 for authentication and/or protection against quantum computers. If the negotiated key exchange algorithm or encryption algorithm offers fewer bits of security than the negotiated PRF, this effectively caps the bits of security offered by the PRF as well. Negotiating a key exchange algorithm or encryption algorithm that offers more bits of security than the negotiated PRF does not improve the security offered by that PRF. As such, it is important to ensure that IKEv2 peers configure algorithm policies such that every algorithm negotiated always meets an acceptable minimum security strength. Where static keys are used with KMAC, these MUST contain at least as much entropy as the security strength of the chosen algorithm, and SHOULD be generated using a random number generator suitable for use with cryptography.

IANA Considerations For negotiating use of KMAC as a PRF for IKEv2, IANA is requested to assign two Transform IDs in the "Transform Type 2 - Pseudorandom Function Transform IDs" registry: Number Name Status Reference TBD PRF_KMAC_128   [This draft] TBD PRF_KMAC_256   [This draft] For negotiating use of KMAC for integrity protection in IKEv2 and IPsec protocols, IANA is requested to assign two Transform IDs in the "Transform Type 3 - Integrity Algorithm Transform IDs" registry: Number Name Status Reference TBD AUTH_KMAC_128   [This draft] TBD AUTH_KMAC_256   [This draft] For indicating support for the SHA-3 cryptographic hash functions and SHAKE XOFs in conjunction with a signature algorithm, IANA is requested to assign five Transform IDs in the "IKEv2 Hash Algorithms" registry: Value Hash Algorithm Reference TBD SHA3_256 [This draft] TBD SHA3_384 [This draft] TBD SHA3_512 [This draft] TBD SHAKE_128 [This draft] TBD SHAKE_256 [This draft]

This document describes version 2 of the Internet Key Exchange (IKE) protocol. IKE is a component of IPsec used for performing mutual authentication and establishing and maintaining Security Associations (SAs). This document obsoletes RFC 5996, and includes all of the errata for it. It advances IKEv2 to be an Internet Standard. ELVIS-PLUS This document specifies the use of variable-length output Pseudo- Random Functions (PRFs) in the Internet Key Exchange Protocol Version 2 (IKEv2). Current IKEv2 specification relies on traditional PRFs with fixed output length for key derivation and uses iterative application of a PRF (called "prf+") in cases when longer output is required. Appearance of PRFs that can output as much bits as requested allows to streamline the key derivation functions of IKEv2. This document updates RFCs 5723, 6617, 6631, 7296, 8784, 9370 for the cases when variable-length output Pseudo-Random Functions are used in IKEv2 and its extensions. In many standards track documents several words are used to signify the requirements in the specification. These words are often capitalized. This document defines these words as they should be interpreted in IETF documents. This document specifies an Internet Best Current Practices for the Internet Community, and requests discussion and suggestions for improvements. RFC 2119 specifies common key words that may be used in protocol specifications. This document aims to reduce the ambiguity by clarifying that only UPPERCASE usage of the key words have the defined special meanings. National Institute of Standards and Technology European Telecommunications Standards Institute This document replaces RFC 7321, "Cryptographic Algorithm Implementation Requirements and Usage Guidance for Encapsulating Security Payload (ESP) and Authentication Header (AH)". The goal of this document is to enable ESP and AH to benefit from cryptography that is up to date while making IPsec interoperable. The IPsec series of protocols makes use of various cryptographic algorithms in order to provide security services. The Internet Key Exchange (IKE) protocol is used to negotiate the IPsec Security Association (IPsec SA) parameters, such as which algorithms should be used. To ensure interoperability between different implementations, it is necessary to specify a set of algorithm implementation requirements and usage guidance to ensure that there is at least one algorithm that all implementations support. This document updates RFC 7296 and obsoletes RFC 4307 in defining the current algorithm implementation requirements and usage guidance for IKEv2, and does minor cleaning up of the IKEv2 IANA registry. This document does not update the algorithms used for packet encryption using IPsec Encapsulating Security Payload (ESP). The possibility of quantum computers poses a serious challenge to cryptographic algorithms deployed widely today. The Internet Key Exchange Protocol Version 2 (IKEv2) is one example of a cryptosystem that could be broken; someone storing VPN communications today could decrypt them at a later time when a quantum computer is available. It is anticipated that IKEv2 will be extended to support quantum-secure key exchange algorithms; however, that is not likely to happen in the near term. To address this problem before then, this document describes an extension of IKEv2 to allow it to be resistant to a quantum computer by using preshared keys. Digital signatures are used to sign messages, X.509 certificates, and Certificate Revocation Lists (CRLs). This document updates the "Algorithms and Identifiers for the Internet X.509 Public Key Infrastructure Certificate and Certificate Revocation List (CRL) Profile" (RFC 3279) and describes the conventions for using the SHAKE function family in Internet X.509 certificates and revocation lists as one-way hash functions with the RSA Probabilistic signature and Elliptic Curve Digital Signature Algorithm (ECDSA) signature algorithms. The conventions for the associated subject public keys are also described. The Internet Key Exchange Version 2 (IKEv2) protocol has limited support for the Elliptic Curve Digital Signature Algorithm (ECDSA). The current version only includes support for three Elliptic Curve groups, and there is a fixed hash algorithm tied to each group. This document generalizes IKEv2 signature support to allow any signature method supported by PKIX and also adds signature hash algorithm negotiation. This is a generic mechanism and is not limited to ECDSA; it can also be used with other signature algorithms.

Test Vectors The following test cases include inputs and outputs for scenarios where KMAC is used in IKEv2 and IPsec. A key, input, customization string, and output are always supplied. These correspond to the K, S, C, and Z parameters described in . Note that in each context, the customization string is fixed. All inputs and outputs are encoded in hexadecimal. KMAC customization strings also have an ASCII character string representation. Data supplied to KMAC does not include quotation marks or null terminators. In some cases a description is supplied, which describes the case being tested in more detail. These descriptions are test vector metadata, and are not ever supplied to the relevant algorithm.

PRF Test Vectors These test cases correspond to use of KMAC as the PRF transform for an IKE SA.

KMAC128 PRF Test Vectors

KMAC256 PRF Test Vectors

KDF Test Vectors These test cases correspond to use of KMAC with IKEv2's prf+ function.

KMAC128 KDF Test Vectors

KMAC256 KDF Test Vectors

KMAC IKEv2 Integrity Protection Test Vectors These test cases correspond to use of KMAC as the integrity protection transform for an IKE SA. Note that, since different customization strings are used for integrity protection in IKEv2 and IPsec, different outputs are produced, so two sets of test vectors are supplied.

KMAC128 IKEv2 Integrity Protection Test Vectors

KMAC256 IKEv2 Integrity Protection Test Vectors

KMAC IPsec Integrity Protection Test Vectors These test cases correspond to use of KMAC as the integrity protection transform for an IPsec SA. Note that, since different customization strings are used for integrity protection in IKEv2 and IPsec, different outputs are produced, so two sets of test vectors are supplied.

KMAC128 IPsec Integrity Protection Test Vectors

KMAC256 IPsec Integrity Protection Test Vectors

Acknowledgments TODO