[freenet-dev] On Sybil Attacks

[Matthew Toseland]

On Thu, Mar 22, 2007 at 11:36:17AM +0000, Thomas Bruderer wrote:
> Hello devl-list
> 
> Since nearly two weeks I am working on my Master Thesis in Theoretical 
> Computation. The topic is On sybil attacks on p2p systems, a bit general but I 
> will focus on a simple abstraction of freenet.
> 
> nodes have a unique id
> nodes can insert
> nodes can request
> nodes are in a small world
> (maybe nodes can swap ids : enforcing swaps would be intersting theoretical 
> topic too)

Indeed, and one that would be of real interest in 0.8.

Locations are swapped on 0.7, fairly frequently.
> 
> I think this are enough primitves to describe Freenet theoretically, if I am 
> mising an important part, please enlight me,.
> 
> new primitives might be introduced to achieve the goal.
> 
> Maybe I'll do a small prove about 0.5 that sybil attacks are easy, but the focus
> should be on the darknet, and maybe on the "Semi-Opennet"

Would be nice. :)
> 
> I read the discussions about sybil-attacks in the board, and alread read some 
> interesting Paper (I read douceurs paper first) I also read Ppaers about 
> SybilGuard and Establishing Distinct Identities.
> 
> Its not as hopeless as it looks in the paper from douceur, and in the darknet 
> model we have the premises about trusted links which are in fact a simple form 
> of a CA. (we all know the Web Of Trust for X509 certificates, which tries to 
> simulate a centralised Authority)

Right, in darknet Sybil isn't an issue unless people are stupid, no?
> 
> However also in Darknet its possible to do Sybilattacks, one trusted link to a 
> "Sybil" is enough... however this leads to Network parts not very good connected
>  to the rest. This leads to some ideas to prevent this. 

I don't understand. Sybil = creating lots of bogus identities. The most
you can do on darknet is get a small number of connections to a small
number of nodes and then pretend to be a much larger number of nodes -
which isn't a big deal, unless we are talking about premix routing, in
which case obviously we will need a counter.
> 
> I also will try to provide some ideas about the semi-opennet if possible, since 
> you know I am advocating it. However for me the theoretical aspects and the 
> practical usablity might differ :)
> 
> Ideas can be Peer Introduction, Topological characteristica, geometrical 
> characterisca of the network.
> 
> If there is alreay work done on this matter for freenet, I would be very glad to
> see it. The thesis will be written till September, and I'll publicate it (if 
> possible) and inform you about the results. However: the thisis will be 
> theoratical, I already have some Ideas which won't be that practical, however I 
> hope I can give a piece to this projects puzzles.
> 
> I hope I won't run in a dead end, and the final answer shouldn't be: "its 
> entierly impossible."

Cool.
> 
> I am very open to thoughts and Ideas.
> 
> Greets
> Thomas Bruderer
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
Url : http://emu.freenetproject.org/pipermail/devl/attachments/20070322/1b9a8306/attachment.pgp