On Wed, Feb 06, 2002 at 03:21:01PM +0100, Nico Galoppo wrote: > > Since you'd have to do this at each hop it would probably hurt > > performance a lot. It's hard to quantify how much "extra anonymity" > > you'd derive. > > Actually, I was rather thinking of a prerouting step at each request, > instead of at each hop. Performance loss would be less than performing > the prerouting at each hop. When operating in 'basic freenet mode' (ie. > after the prerouting step) you would be sure that the preceding node is > just forwarding a request. Anonimity for data storers would then solely > rely on 'plausible deniability', while still providing extra anonimity > for inserters and retrievers through the prerouting step.
I started thinking about that just after sending the last message :) I suppose it can work, although failed links are still difficult. Somewhere code would have to be added to hold replies for Alice until she reopens the link. You could actually do this without touching Freenet at all: create an AnonNet+F network where in addition to commanding a node to set up an onion-step to another node, you can command it to make an ordinary Freenet request and return the result. You could even try just tunneling a stream through to a Freenet node's TCP/IP listening port but of course you'd be in trouble if the link failed. > > There's also the problem of links failing. They'd have to be > > re-established from the originating end with some kind of session key, > > which would force nodes to hold on to reply messages until Alice reopens > > the link. Not fun at all, particularly with the current architecture. > > This could be a real problem. We've been thinking about that. Also, > there's the possible threat of connection setup. It could be considered > 'a lightning' that flashes up in the dark. That reveals the routing that > you're trying to make 'secret', not something we want. Yeah, if you don't onion-route at each hop, then setting up the AnonNet link is a giveaway that a node is initiating a request, so you end up shifting the "not knowing whether the node is originating the request" to "not knowing whether the node is originating the onion route". In light of that the extra anonymity provided can be characterized easily. We are nullifying the information contained in the HTL value; no more, no less. > > As a side note, it might be worth exploring the addition of traffic > > shaping (or un-shaping ;-) features to Freenet like those found in > > pipenet/AnonNet, regardless of whether the onion-routing aspect > > can be made useful. > > I'll have a look into that. Unfortunately, I'll have to focus on the > anonimity aspects, due to the limitations of my thesis subject ;-) Anything related to defying traffic analysis is definitely related to anonymity. -tc _______________________________________________ freenet-tech mailing list [EMAIL PROTECTED] http://lists.freenetproject.org/mailman/listinfo/tech
