[144623] in cryptography@c2.net mail archive
Re: New Technology to Make Digital Data Disappear, on Purpose
daemon@ATHENA.MIT.EDU (Jerry Leichter)
Thu Jul 23 10:22:32 2009
Cc: dan@geer.org,
"Ali\, Saqib" <docbook.xml@gmail.com>,
Cryptography <cryptography@metzdowd.com>
From: Jerry Leichter <leichter@lrw.com>
To: Perry E. Metzger <perry@piermont.com>
In-Reply-To: <873a8pwhyf.fsf@snark.cb.piermont.com>
Date: Wed, 22 Jul 2009 20:30:45 -0400
On Jul 21, 2009, at 10:48 PM, Perry E. Metzger wrote:
> dan@geer.org writes:
>>> The pieces of the key, small numbers, tend to =93erode=94 over
>>> time as
>>> they gradually fall out of use. To make keys erode, or timeout,
>>> Vanish
>>> takes advantage of the structure of a peer-to-peer file system. Such
>>> networks are based on millions of personal computers whose Internet
>>> addresses change as they come and go from the network.
>>
>> One would imagine that as IPv6 rolls out, the need
>> for DHCP goes to zero excepting for mobile devices
>> attaching to public (not carrier) nets. Yes?
>
> Off topic, but actually DHCP is still needed. A machine needs to
> configure a lot more than just its address and router in common cases
> (it wants things like DNS servers, NTP servers, etc.) and in large
> deployments, it is often far easier to let machines autoconfigure
> these
> things during boot using DHCP even on comparatively hard wired
> networks.
>
> And with that, lets return to crypto...
The proposal makes use of an incidental property of existing DHT
implementations: Because many nodes are running on machines with
dynamic IP addresses, rehashes - which cause the table to change and
this leads to the loss of bits. It's not actually clear from the
paper how much of the bit loss is actually due to IP address changes
and how much to other phenomena. In any case, if this idea catches on
and there isn't enough "noise" in the network naturally to give an
adequate bit drop rate, it would be reasonable to add an explicit bit-
dropping mechanism to some new release. You'd need one to add IPv6
support anyway!
-- Jerry
---------------------------------------------------------------------
The Cryptography Mailing List
Unsubscribe by sending "unsubscribe cryptography" to majordomo@metzdowd.com
