On Thu, Nov 07, 2002 at 07:47:41AM -0500, Scott Young wrote: > If you compare this to the reasons behind the probabalistic HTL-1 > forwarding, I think it would be better just to abandon HTL altogether, and > go for a completely probabalistic approach. Each request would be forwarded > with a certain probability of being sent further. This would make it > impossible for any analysis of the HTL value (solving the first few hops > problem), and would make it much more difficult to determine wether or not a > node has specific data (solving the HTL-1 problem). A malicious node would > have no say in how far a request goes. > That is certainly possible. A few calculations by oskar suggest that it would increase the standard deviation of the request time by a factor of six. > > Scott Young > > > ----- Original Message ----- > From: "Matthew Toseland" <[EMAIL PROTECTED]> > To: <[EMAIL PROTECTED]> > Cc: <[EMAIL PROTECTED]>; <[EMAIL PROTECTED]> > Sent: Wednesday, November 06, 2002 6:09 PM > Subject: [Tech] The first few hops problem > > > Possible solution to the problem that you can see that a request has > > been initiated on a given node: > > Requests can have either HTL, or HTL|P, where P is a number between 0 > > and 1 (this would be limited to a more realistic range by each node it > > passed through). If request only has HTL, it is processed normally. If > > request has HTL|P, there is a P chance that it is forwarded as is, and a > > 1-P chance that it is turned into an HTL only request. So depending on > > the value of P, which can be set at the client end, we have a variable, > > random number of hops before the main HTL starts. This should greatly > > reduce the vulnerability to nodes seeing that requests are at a fixed > > request HTL, without needing huge packets (mixmastered first few hops), > > and without greatly increasing the variance of the request time, unless > > the probability is set to a very high value. The bounds are a topic of > > interest, as is the possible information leak of the probability - we > > probably want a limited set of probabilities available to clients, > > rather than the whole range, to avoid leaking too much information that > > could uniquely identify a requestor. What do people think? > > -- > > Matthew Toseland > > [EMAIL PROTECTED] > > [EMAIL PROTECTED] > > Freenet/Coldstore open source hacker. > > Employed full time by Freenet Project Inc. from 11/9/02 to 11/11/02. > > http://freenetproject.org/ > > > > > _______________________________________________ > Tech mailing list > [EMAIL PROTECTED] > http://hawk.freenetproject.org/cgi-bin/mailman/listinfo/tech >
-- Matthew Toseland [EMAIL PROTECTED] [EMAIL PROTECTED] Freenet/Coldstore open source hacker. Employed full time by Freenet Project Inc. from 11/9/02 to 11/11/02. http://freenetproject.org/
msg00992/pgp00000.pgp
Description: PGP signature
