Re: Deprecating RSA ssh host keys in 16

Shawn Webb <[email protected]> Tue, 24 Sep 2024 19:20:11 +0000
Newsgroups	gmane.os.freebsd.architechture
Message-ID	<7ujil5wxcwnmoobmjsmtdvfubmb3eiqcsblut3lwt7ussdxwxq@6qskqqcvfkcu>
On Tue, Sep 24, 2024 at 07:16:04PM UTC, Shawn Webb wrote:
> On Tue, Sep 24, 2024 at 06:41:00PM UTC, Colin Percival wrote:
> > Hi all,
> > 
> > Last week I turned off RSA host key generation for SSH in EC2 instances,
> > because (a) modern SSH clients support ecdsa and ed25519 keys, and (b)
> > generating RSA host keys was taking over 10% of the boot time when EC2
> > instances booted for the first time.
> > 
> > I don't think we should turn off RSA host key generation in general in
> > 15.x since for non-VM/cloud images the first boot time is less relevant
> > (if you're installing from an ISO image, the installer will take far
> > longer than the host key generation) but I think it would make sense to
> > deprecate RSA host keys in 15 and then turn them off by default in 16.
> > 
> > I'm not sure if there's any good way to announce the deprecation beyond
> > putting it into the release notes; we could print a warning in 15 when
> > RSA host keys are generated, but that will always fire regardless of
> > whether they're being *used* and I don't think there's any practical way
> > to warn specifically when RSA host keys are *used*.  So unless I'm
> > missing something, the deprecation would just take the form of a few lines
> > in the release notes.
> > 
> > Thoughts?
> 
> With commit e3f33c64ec168a48038309af0c237eda86d10c74[1], introduced on
> 14 Nov 2024, HardenedBSD has disabled the generation of RSA host keys
> by default.

Whoops, time travel hasn't been invented yet. (Or so we think? ;-P)
That would be 14 Nov 2023.

> 
> We haven't seen any reports of any breakage. While the change might be
> considered a POLA violation, it seems pretty harmless on today's
> 15-CURRENT systems.
> 
> We have a number of 15-CURRENT users, though we don't have any hard
> data, and likely pales in comparison to the FreeBSD side--enough so
> that the sample is too small to be a significant or reliable data
> point.
> 
> I have this commit taged as MFC-able, though I haven't MFC'd just yet.
> It completely spaced my mind and I'll likely MFC shortly after sending
> this email.
> 
> [1]: https://git.hardenedbsd.org/hardenedbsd/HardenedBSD/-/commit/e3f33c64ec168a48038309af0c237eda86d10c74
> 
> Thanks,
> 
> -- 
> Shawn Webb
> Cofounder / Security Engineer
> HardenedBSD
> 
> Tor-ified Signal: +1 303-901-1600 / shawn_webb_opsec.50
> https://git.hardenedbsd.org/hardenedbsd/pubkeys/-/raw/master/Shawn_Webb/03A4CBEBB82EA5A67D9F3853FF2E67A277F8E1FA.pub.asc



-- 
Shawn Webb
Cofounder / Security Engineer
HardenedBSD

Tor-ified Signal: +1 303-901-1600 / shawn_webb_opsec.50
https://git.hardenedbsd.org/hardenedbsd/pubkeys/-/raw/master/Shawn_Webb/03A4CBEBB82EA5A67D9F3853FF2E67A277F8E1FA.pub.asc
signature.asc (application/pgp-signature, 833 B)