----- Original Message -----
From: "Brandon" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Wednesday, December 27, 2000 2:42 PM
Subject: Re: [Freenet-dev] Don't Talk To Strangers
>
> > As Scott has pointed out, capturing the gateway node would provide a
nice
> > neat list of nodes whose owners previously believed were safe as houses,
> > where as capturing a normal Freenet node would give you a somewhat
random
> > list of other nodes which could be anywhere in the world.
>
> Same thing if you're trying to shut down Freenet nodes. Also, let's be
> clear about the MediaEnforcer attack. They scan IPs. They do not cease
> node lists from detected nodes. Public key crypto defeats the
> MediaEnforcer attack (scanning IPs doesn't work).
>
> Let us then consider the Enforcer Node attack in which they run a node to
> "fish" for IP addresses. Public key crypto does nothing. Rejecting unknown
> connections defeats this attack.
>
> Let us then consider the Totalitarian Government attack. They run a node
> and they cease node lists from detected nodes. Everything we have
> including rejecting unknown connections fails against this attack.
>
> So of the the 3 attacks (Enforcer, Enforcer Node, and Totalitarian
> Government), Freenet 0.3.6 fails against all of them, Freenet+PKI fails
> against the last two, Freenet+PKI+rejecting unknown connections fails
> against one, and nothing we've come up with defeats the last one.
>
> So rejecting unknown connections is still a stronger defense and is
> therefore good, not bad, in terms of defeating attacks in this genre. It
> doesn't solve everything, but neither does anything else so far.
>
> > On the second point, there are serious questions over whether one of
> > Brandon's clusters would be of significantly more value to Freenet than
an
> > individual Freenet node, not to mention the obvious client/server
> > central-point-of-failure and host of
>
> I have already addresses this. I'll address it again and then you can
> directly bring up the flaws in my response in the future. A cluster does
> not have a central point of failure. Clients connected to a trusted server
> do have a central point of failure. A cluster is a group of peers, one of
> which volunteers to be a gateway. If the gateway is shutdown, any cluster
> node can volunteer to be a gateway. There can be multiple gateways.
When Mallory volunteers to be a gateway, the entire cluster is screwed,
since Mallory knows all nodes in the group. Even if Bob and Alice are
secondary (or even primary) gateways, Mallory knows their nodes, too, and
can later take them over
>
> The advantages of having a cluster instead of clients connected to a
> trusted server is that cluster nodes are still useful members of the
> Freenet community. If you have a 10 node cluster, then you have 10
> nodes. If you have 10 clients then you have 1 node. Clusters are still
> part of the network. If the gateways get shut down, they can still trade
> within the cluster until a new gateway is established.
The only time I could possibly trust other nodes for estabilishing a cluster
is when I control every computer in that cluster. Other then that, Freenet
*must* operate in a zero-trust environment.
_______________________________________________
Freenet-dev mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/mailman/listinfo/freenet-dev
