[Swan-dev] defaults for ike= and esp= need updating?

[Paul Wouters]

On Thu, 8 Dec 2016, Andrew Cagney wrote:

> Given something like ike=sha1 or ike=aes or ...  pluto uses the following table to fill in the ENCR-PRF;MODP blanks:

It should basically act as a filter on the "default set".

>    DH:OAKLEY_GROUP_MODP2048, OAKLEY_GROUP_MODP1536, OAKLEY_GROUP_MODP1024,


Can this behaviour be tuned for IKEv1 or IKEv2? I would like IKEv2 to
not have 1536/1024. And I would like IKEv1 to not have 1024.

>    ENCRYPT: OAKLEY_AES_CBC, OAKLEY_3DES_CBC,

This is fine for IKEv1. For IKEv2 we would want AES_GCM and not 3DES.

>    PRF: OAKLEY_SHA1, OAKLEY_MD5,  (it's actually a HMAC based on)

Actually, we sort of only do PRF == INTEG, so for ike=sha1 it should
pick sha1 as prf. For ike=aes it should pick the strongest default,
so a sha2 flavour? (sha2_256 is fine for IKE, we just want to avoid
it for ESP due to the linux bug still existing and badly fixed in
2.6.33)

> Similarly, for esp:
> 
>     encrypt=AES

SHA2_512 or SHA1

> and esp/ah:
> 
>     INTEG="MD5", "SHA1" (its actually a truncated HMAC based on ...)
> 
> do these need updating?

Yes.

> (my preference is to drop the magic defaults but I suspect that is too radical)

Yes, that falls in the category "pony" and "world peace" :)

For guidance please see:

https://tools.ietf.org/html/draft-ietf-ipsecme-rfc7321bis

https://tools.ietf.org/html/draft-ietf-ipsecme-rfc4307bis

Paul