On Sun, Jun 03, 2001 at 10:44:11PM -0500, Timm Murray wrote:
> >In fact I think that if you tie this into Freenet itself you are directly
> >undermining one of the project's goals: plausible deniability.  "But I
> >didn't know my node was holding metadata information on those copyrighted
> >mp3s."  "Whaddya mean, it's right here in plain text"
> 
> What if a given plaintext metadata key didn't move with a given document.
> In that case, even if you had a metadata key that said "Naked Eskimo Pics",
> it wouldn't nessarily mean you had any naked eskimo pics in your node store.

Doesn't matter. You are running a node in a network where you know there
is significant illegal material transferred. You have the capability to
remove obviously illegal metadata from your node--and you refuse! How
will a judge and jury view that?

> We could propagate metadata keys just like normal keys; the search request
> moves through the network, each node checking it's metadata keys, then the
> metadata key(s) that get a positive response get sent back for the next node
> in line to cache.

Well, the requests would not stop until the HTL is exhausted, but
otherwise they could be normal. (The results of the search could be
appended to the StoreData sent back to the requester.)

I'm confident that I could destroy this system by running a set of evil
nodes who respond to every search with a deluge of hits. Soon, virtually
all searches would be routed to my nodes.

By selectively replying to search requests, I can position my nodes in
any area of the search-keyspace. I can flood anyone who searches for
"democracy" with child porn, for example. Because of this vulnerability,
the routing tables can't be shared between searches and requests--or I
could manipulate search results to gain and retain control over areas of
the request-keyspace.

If I understand the proposal correctly, metadata is inserted like:

freenet_insert --key CHK at foo --description "the decline and fall of \
the roman empire, by edward gibbon"

and then when some of those words are searched for, the reply contains
CHK at foo and the description that was originally inserted.

If the descriptions are actually descriptive, my evil nodes will not be
able to imitate them very well. So the user could plausibly pick out the
bogus replies and Unrequest them. But I suspect that users often will
simply use the filename as the description, or some simple permutation
of it. And that can be faked.


-- 
"...it must be held that third-party electronic monitoring, subject
only to the self-restraint of law enforcement officials, has no place
in our society..." Mark Roberts | mjr at statesmean.com

_______________________________________________
Devl mailing list
Devl at freenetproject.org
http://lists.freenetproject.org/mailman/listinfo/devl

Reply via email to