Re: Proposal: Remove MD5 / SHA1 support from veriexec

[Taylor R Campbell <campbell+netbsd-tech-security%mumble.net@localhost>]

> Date: Tue, 22 Aug 2017 01:50:30 +0100
> From: Sevan Janiyan <venture37%geeklan.co.uk@localhost>
> 
> It seem I forgot a couple of things, the reason why I'm proposing the
> change is because these ciphers are broken and are of no benefit in the
> context of this tool. I'm proposing the staged process as a migration
> path rather than ripping it flat out for NetBSD 8 though I have no
> objection to doing that if folks are happy with that.
> 
> On 08/22/17 00:12, Sevan Janiyan wrote:
> > Part 2:
> > http://www.netbsd.org/~sevan/patch-veriexec-nomd5-sha1.txt
> 
> This diff has been updated to cover all references to
> VERIFIED_EXEC_FP_MD5 & SHA1.

Sounds good to me.  Maybe also start to get rid of RIPEMD160 since
even if it's not specifically broken like MD5 and SHA1 are, the cost
of a generic collision attack on a 160-bit hash function is nearing
the realm of possibility.

(What's the threat model for which a collision attack is relevant?
Cook up a colliding pair of binaries and persuade someone to apply a
patch that causes one of those binaries to come out.)