some light reading
<auto97841-revL73yDgGBWk0Htik3J/[email protected]> Sun, 6 Jul 2003 17:33:36 -0700
| Newsgroups | gmane.comp.security.invisiblenet.iip.devel |
|---|---|
| Message-ID | <[email protected]> |
as promised, here's whats going on in my head for actual network topology / routing / etc. The latency this should pull, assuming 0ms message handling time (aka impossible, but tunable as per router implementations), came to ~ 600ms for a message to be sent over the internet anonymously with 5 million routers in the network. (2LK+Clog(N)/K: L = avg # hops / tunnel, K = avg ms / hop, C = % of routing table cached per router, N = # routers, but thats all way preliminary) This draws upon many, many, many different systems, and I've gone through a few hundred pdfs/.ps files, but from what I've been reading, the closest anonymizing network design out there to what I'm describing is Morphmix - http://citeseer.nj.nec.com/565365.html The two differences are: Morphmix is a gateway system (anonymized originator, not destination, while we provide sender, reciever, and server anonymity) Morphmix felt that distributed hash tables aren't worth the effort to implement (I obviously disagree [see log(N) above], and with caching DHTs, the overhead is nearly none with significant benefits [see the % factor of C off log(N)]) Anyway, everything in red is "um, wtf", but I thought it'd be good if I got this out even in draft form so people would be able to look it over *BEFORE* the tuesday meeting ;) Questions/problems/thoughs/frisbees, reply here or /msg jrand0m on iip -jrandom
draftNetSpec_july_7.zip
(application/zip, 36 KB) - not displayed