Cryptographic Algorithm Implementation Requirements and Usage Guidance for Encapsulating Security Payload (ESP) and Authentication Header (AH)

The Encapsulating Security Payload (ESP) and the Authentication Header (AH) are the mechanisms for applying cryptographic protection to data being sent over an IPsec Security Association (SA) . This document provides guidance and recommendations so that ESP and AH can be used with a cryptographic algorithms that are up to date. The challenge of such document is to make sure that over the time IPsec implementations can use secure and up-to-date cryptographic algorithms while keeping IPsec interoperable. Cryptographic algorithms evolves over time... Sunsetting / introducing new algorithms ... This document is intended for IPsec developers as well as for IPsec users.

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 . Following , we define some additional key words: This term means the same as MUST. However, we expect that at some point in the future this algorithm will no longer be a MUST. This term means the same as SHOULD. However, it is likely that an algorithm marked as SHOULD+ will be promoted at some future time to be a MUST.

Name Status AEAD Comment NULLMUSTNo/Yes[RFC2410] AES-CBCMUST-No[1] [RFC3602] AES-GCM with a 16 octet ICVMUSTYes[1] [RFC4106] CHACHA20_POLY1305SHOULDYes [RFC7634] AES-CCM_16SHOULDYes[1][IoT] [RFC4309] AES-CCM_8SHOULDYes[1][IoT] [RFC4309] 3DESSHOULD NOTNo[RFC2451] DESMUST NOTNo[RFC2405] [1] - This requirement level is for 128-bit keys. 256-bit keys are at SHOULD. 192-bit keys can safely be ignored. [IoT] - This requirement is for interoperability with IoT. IPsec sessions may have very long life time, and carry multiple packets, so there is a need to move 256-bit keys in the long term. For that purpose requirement level is for 128 bit keys and 256 bit keys are at SHOULD (when applicable). In that sense 256 bit keys status has been raised from MAY in RFC7321 to SHOULD. NULL status remains to MUST to enable the use of ESP with only authentication. It status is expected to remain MUST by protocol requirements. ENCR_AES_CBC status remains to MUST in order to enable interoperability between implementation that followed RFC7321. AES-GCM status has been updated from SHOULD+ to MUST in order to favor the use of authenticated encryption and AEAD algorithms. The main motivation for adopting AES-GCM for ESP is performance as well as key longevity - compared to AES-CBC for example. This resulted in AES-GCM widely implemented for ESP. ENCR_CHACHA20_POLY1305 was not ready to be considered at the time of RFC7321. It has been recommended by the CRFG and others as an alternative to AES and AES-GCM. It is also being standardized for IPsec for the same reasons. At the time of writing, there were not enough ESP implementations of ENCR_CHACHA20_POLY1305 to be able to introduce it at the SHOULD+ level. ENCR_AES_CCM_8-16 status have been raised from MAY to SHOULD. This document considers it SHOULD be implemented in order to be able to interact with Internet of Things devices. As this case is not a general use case for VPNs, its status is expected to remain to SHOULD. ENCR_3DES status has been downgraded from MAY in RFC7321 to SHOULD NOT. However, ENCR_CHACHA20_POLY1305 seems a more modern approach alternative to AES than 3DES and so it expected to be favored over 3DES. ENCR_DES status remains to MUST NOT.

Encryption without authentication MUST NOT be used. As a result, authentication algorithm recommendations in this section are targeting two types of communications: Firstly authenticated only communications without encryption -- such communications can be ESP with NULL encryption or AH communications. Secondly, communications that are encrypted with non AEAD encryption algorithms mentioned above. In this case, they MUST be combined with an authentication algorithm. Name Status Comment AES-128-GMACMUST[RFC4553] AES-256-GMACSHOULD[RFC4553] HMAC_SHA2_256_128SHOULD[RFC4868] HMAC_SHA2_512_256SHOULD[RFC4868] HMAC_SHA1_96MUST-[RFC2404] AES_XCBC_96SHOULD[IoT] [RFC3566] NULLMUST NOTonly acceptable for authenticated encryption [RFC4303] [IoT] - This requirement is for interoperability with IoT HMAC_SHA2_256_128 was not mentioned in RFC7321, as no SHA2 based authentication was mentioned. HMAC_SHA2_256_128 SHOULD be implemented in order to replace HMAC_SHA1_96. HMAC_SHA2_512_256 SHOULD be implemented as a future replacement of HMAC_SHA2_256_128 or when stronger security is required. This value has been preferred to HMAC_SHA2_384, as the overhead of HMAC_SHA2_512 is negligible. HMAC_SHA1_96 status has been downgraded from MUST in RFC7321 to MUST- as there is an industry-wide trend to deprecate its usage. AES-XCBC is only recommended in the scope of IoT, as Internet of Things deployments tend to prefer AES based HMAC functions in order to avoid implementing SHA2. For the wide VPN deployment, as it has not been widely adopted, it has been downgraded from SHOULD in RFC4307 to MAY. AES-128-GMAC status is being upgraded from SHOULD+ to MUST in order to guarantee interoperability between implementation that are following RFC7321 and this document. In fact HMAC-SHA1-96 was used to be the algorithm that provided interoperability and it has been downgraded. AES-256-GMAC was not mentioned in RFC7321. Its status has been set to SHOULD as as a future replacement of AES-128-GMAC or when stronger security is required. NULL has been downgraded from MAY in RFC7321 to MUST NOT. The only reason NULL is acceptable is when authenticated encryption algorithms are selected from . In any other case, NULL MUST NOT be selected. As ESP and AH provides both authentication, one may be tempted to combine these protocol to provide authentication. As mentionned by RFC7321, it is NOT RECOMMENDED to use ESP with NULL authentication - with non authenticated encryption - in conjunction with AH; some configurations of this combination of services have been shown to be insecure . In addition, ESP NULL authentication cannot be combined with ESP NULL encryption.

I think that would be interesting to document and summarize the changes.

None.