Well, a more obvious flaw which exists and is exploitable right now, and
is something of a "newbie crypto mistake" is the fact that we are still
using ephemeral diffie-hellman (with an outer encryption layer so you
need to know both refs). We really should fix that... Nextgens has
decided not to, should I? It's probably only a few days' work.

On Fri, Mar 23, 2007 at 10:48:18AM -0400, Colin Davis wrote:
> I worry that given that the Freenet Project has always been a very 
> public project, and relies heavily on donations to continue to develop, 
> headlines like "Freenet Authors add tracking code" aren't going to 
> endear you to people.
> 
> And while I certainly understand the idea, and I understand that the 
> security risks aren't actually that major compared to the existing 
> infrastructure and a determined hacker, I don't think it'd play well 
> with the Slashdot crowd. They aren't exactly world-renowned for 
> listening to nuanced arguments before making judgments ;)
> 
> Just my thoughts,
> Colin
> 
> 
> 
> Matthew Toseland wrote:
> > How far can we go in abusing the production network in order to make it
> > work? The testnet is never likely to be large enough to be a useful
> > model. What I propose is that on swap requests, which already include
> > the location of the node, and the locations of its peers, we also add a
> > unique ID (say the first 8 bytes of some hash of the identity) for the
> > node and each peer.
> > 
> > This would make it easier to map the network. It is already possible to
> > map the network but it is a lot of work and a lot of uncertainty,
> > because we don't know about every swap so we have to try to do partial
> > matches.
> > 
> > This may make some attacks easier. Having said that, with the current
> > swap requests, you can probably identify the topology close to you with
> > some confidence. The main benefit here is in identifying the topology
> > further away more reliably. Which isn't that interesting for attackers
> > unless they've been e.g. watching #freenet-refs and can match an IP
> > address to each node on the network. Even then, there are much easier
> > attacks, and correlation attacks on nodes 4 hops away may not have
> > enough information.
> > 
> > The benefit is we could test all our pet theories about the shape of the
> > network being completely broken due to #freenet-refs . We could gather
> > real world information about node uptimes, location swapping, location
> > clustering. It would of course be spoofable, but only to the extent that
> > location swapping is already spoofable. It would double the size of the
> > swap request packets, but these are fairly small.
> > 
> > What do you think?
> > 
> > 
> > ------------------------------------------------------------------------
> > 
> > _______________________________________________
> > Devl mailing list
> > Devl at freenetproject.org
> > http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
> 
> _______________________________________________
> Devl mailing list
> Devl at freenetproject.org
> http://emu.freenetproject.org/cgi-bin/mailman/listinfo/devl
> 
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL: 
<https://emu.freenetproject.org/pipermail/devl/attachments/20070323/2684c859/attachment.pgp>

Reply via email to