Date: Fri, 16 Dec 2016 16:54:52 +0100
From: "M. Tariq Elahi" <Tariq.Elahi@esat.kuleuven.be>
Subject: Re: Tor Research Safety Board: default bridge reachability

Yesterday, I wrote to David et al. with a short summary of our review.
The tl;dr was that we (Damon McCoy and I) don't see anything
particularly wrong with their experimental setup and their proposed
research. Here is a less terse version of the review.

Summary of request:

The goal of the experiments is to gain more extensive empirical data
about how censors block Tor bridges. The experiments will measure where
in the world bridges are blocked, how many are blocked, how long before
they are blocked and other similar statistics and data points. The
overarching goal is to use this new found information to aid in better
bridge distribution schemes and censorship resistance in general.

Experimental setup:

Mainly the setup is to try to access bridges from many points on the
Internet (through VPS hosting providers) and record if they were
successfully connected to. There doesn't seem to be anything
particularly risky here. For bridges that do get burned because of these
probes we suggested that perhaps replacing them might be fair.

Risks:

The risk to users or the testing nodes do not seem high. The expected to
be released data does not contain identifying information. The risk to
public bridges is that they get blocked. However, as noted in the
request as well, bridges that are public are not particularly well
defended. However, this state of affairs seems to be OK in situations
where the censor is not actively trying to find bridges to block.


Recommendation:

We believe that these experiments do not significantly increase the
threat level of the Tor network, its operators, or its clients.