On Saturday 22 December 2007 09:24, Volodya wrote: > Matthew Toseland wrote: > > [ snip long security argument ] > > > > PROPOSAL: > > Add a flag RandomRoute. This may be set when a request starts (up to the > > user). There is a 50% chance of its being unset. So on average it adds 2 hops > > to the journey - but there is a small chance of requests going much further > > than that. The advantage is that it greatly obscures the picture for a > > distant attacker, by starting off in a somewhat random part of the keyspace. > > NOTES: > > We could not overload HTL=10 because HTL is reset to 10 every time we get > > closer to the target: we *do not* want to go into random route mode just > > because we got a bit closer to the target! > > PROBLEMS: > > It reveals that the request is relatively early. This will make local > > correlation attacks even easier. So we should do it *after* we have premix > > routing, at which point that won't be a problem any more. > > Would it be possible to have (a very small) probability of setting the RandomRoute flag when it's unset? In > that case if the attacker intercepts the random routed key one has only an inductive rather than deductive > proof that originator is near.
Maybe. It would change the probability from 50% to something a bit less say 25% if on average there are two stages of random routing on a typical request. Although IMHO two stages on average would be too many. > > P.S. Yes i realise that simply adding random at each step is not a positive thing. -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20071222/897d88d4/attachment.pgp>
