On Thu, Apr 30, 2020 at 09:33:28PM +0100, Stuart Henderson wrote:
> On 2020/04/30 20:11, Tobias Heider wrote:
> > Hi,
> > 
> > I would like to modernize our crypto defaults a bit and add some of the
> > supported ECDH Diffie-Hellman groups to the default IKE crypto proposal.
> > There should be no downside to this, if they are not supported by the
> > other side one of the old MODP groups will be used.
> > 
> > The same for AEADs in the ESP proposal.  We have support for AES-GCM
> > and CHACHA20 for some time now but they never made it into the
> > defaults.
> > 
> > ok?
> 
> ok to add them.

On second thought i would actually only add the ECDH groups for now.
For AEADs we would probably need a bit more boilerplate because they would
have to be sent in a second proposal without the AUTH transforms and
that can wait until after the release.

> 
> I'm really tempted to suggest dropping the worst of the rest from default
> transforms, users can still re-add them if needed. Not sure if that's a now
> thing or a post unlock thing though.
> 
> I was going to experiment some more (in particular to see what Windows
> comes up with by default nowadays) but the only box I'm running iked on
> that isn't going to interrupt other VPN users, is also running bgpd and
> I just discovered the hard way that starting iked clears out existing
> tcpmd5 SAs so I'm not going to touch that right now ;)
> 

According to the strongswan website [1] the Windows defaults should be:

128-CBC, AES-192-CBC, AES-256-CBC, 3DES, SHA-1,SHA-256, SHA-384 and MODP-1024.

If this is true we could actually drop 3DES and HMAC-SHA1-96, which would be
great.  MODP-1024 as the only Diffie-Hellman group however is already the
weakest group we offer (and if not for windows I would gladly drop it as well).

[1] https://wiki.strongswan.org/projects/strongswan/wiki/WindowsClients

Reply via email to