Re: [tor-talk] MTor (multicast tor), is it going to be released?

[Paul Syverson <paul.syverson@xxxxxxxxxxxx>]

On Wed, Aug 23, 2017 at 03:40:22AM -0400, Roger Dingledine wrote:
> On Mon, Aug 21, 2017 at 11:49:22PM -0700, Yuri wrote:
> > Here is the white paper with MTor design: https://www.degruyter.com/downloadpdf/j/popets.2015.2016.issue-2/popets-2016-0003/popets-2016-0003.pdf
> > 
> > 
> > So, what is MTor's status?
> 
> It's the standard story -- it's a research paper written by a research
> group to show a concept, and then they moved on.
> 

Right. The primary author had graduated and moved on to other things 
c. a year and a half ago.


> I don't remember the design in detail, but I remember based on the talk
> at PETS thinking that they had really changed the threat model for Tor
> to something much weaker, in exchange for better scalability in some
> situations.

Main differences were (1) a multicast root is selected amongst
(adequate-weight) available middle relays that will mate all
connections to it for that multicast session, (2) that any relay that
is part of a multicast session has a session group identifier (GID),
(3) that if a circuit building request hits a guard or middle relay
already participating in the session, it connects that circuit to its
existing group circuit, and (4) optionally, the set of relays to
select a middle hop from is restricted for deduplication benefit.

Our analysis showed that for small to moderate sized groups on the
existing Tor network, absent a pretty restricted middle-hop set, there
was virtually no deduplication (hence a star, which still saved over
all pairs).  If curious, you can see our security analysis of MTor
against targeting adversaries (also analysis of a group chatting via a
private IRC channel and of people connecting to the same onion
service) here: https://arxiv.org/abs/1706.10292

aloha,
Paul
-- 
tor-talk mailing list - tor-talk@xxxxxxxxxxxxxxxxxxxx
To unsubscribe or change other settings go to
https://lists.torproject.org/cgi-bin/mailman/listinfo/tor-talk