Re: [DNSOP] New Version Notification for draft-muks-dns-message-checksums-00.txt

[Paul Vixie <[e-mail]>]

Robert Edmonds wrote:
> ...
>
> I am also curious why a cryptographic hash function (SHA-1) is needed
> for this.  Is a fast non-cryptographic checksum not suitable (e.g.,
> CRC-32C, which can be computed in hardware on x86 CPUs)?

in currently theorized attacks, the udp checksum is fooled by altering
two parts of a fragment:

first, alter the part you want to use to inject poison into a cache.

second, alter something else to fix up the checksum based on the first
alteration.

if CRC-32C is immune to that attack, i havn't heard, but i'd believe.

> Also, there is a long deployment tail for new EDNS options.  If it's
> urgent to deploy a countermeasure against off-path fragment spoofing,
> why not something like Unbound's "referral path hardening", or
> advertising a smaller EDNS buffer size which is much less likely to
> result in fragmentation?  (E.g., I believe OpenDNS advertises a ~1.4
> Kbyte EDNS buffer size.)  Those countermeasures can be deployed
> unilaterally by the resolver, and on a shorter time scale than a new
> EDNS option.
>

those things should also be done in the short term.

but it's the internet. it'll outlive us all. we ought to have a long
term plan as well.

-- 
Paul Vixie