Re: [Curdle] Time for SSH3?

[Peter Gutmann <[e-mail]>]

Matt Johnston <[e-mail]> writes:

>In discussion of Terrapin I think there's been confusion of server-sig-algs
>being security relevant. It's not a security mechanism, it's a compatibility
>mechanism. The server decides whether to accept rsa sha1 signatures
>(regardless of whether a client thinks it can send rsa sha1 signatures), and
>dropping a server-sig-algs extension packet doesn't change that.

Ah, good point.  So perhaps before everyone rushes out work on their code with
a pair of pliers (and optionally a blowtorch), it'd help first to get an idea
of what the practical impact of this attack is.  I mean, apart from egg on the
face what could an attacker actually do that's going to cause a real security
problem?  Unless you exploit the @openssh.com ping extension (yet another
OpenSSH add-on causing problems) you've got a very low probability of removing
a message (0.04, CBC-EtM page 10) which even if you manage it probably won't
have any security implications.

What happens if you leave it as is, or at least mark it as "for future
attention when the opportunity arises"?

Peter.