[metrics-team] Relay diversity master thesis

[Iain Learmonth]

Hi,

On 18/12/17 08:53, Robin Descamps wrote:
> May I ask you advices/feedback about this master thesis plan? Since I would like this project to bring a real contribution to the Tor development, I want to make sure that all the steps I will perform are useful and/or worth it.

Have you contacted the Tor research safety board? If not, please do so.
I don't believe you will have any issues as you are only using test
networks and not collecting data from the public Tor network (other than
data that was already collected by Tor Metrics).

https://research.torproject.org/safetyboard.html

You may have noticed that a large number of relays are currently
concentrated into a few ASs. At the time of writing, over 15% of
consensus weight is in OVH alone so having research into what diversity
helps and what diversity doesn't is always useful.

https://atlas.torproject.org/#aggregate/as

I'm not entirely familiar with the path selection algorithms, but I
assume you have read about these and found merit in exploring them.

From reading the plan, I'm not sure if you are aware that flags are
assigned by the directory authorities and not chosen by the user. While
it's true that a relay can be configured to allow or disallow exit
traffic, and depending on its configuration may be awarded various
flags, its the directory authorities that decide.

You may be interested to investigate how diversity in the current
network could be improved (or harmed) by tweaking the parameters used by
the directory authorities to award flags. For an example (perhaps not a
great one) what happens if we limit the number of guards in any AS to be 25?

The consensus health tool shows how the directory authorities are voting
in the public Tor network.

http://consensus-health.torproject.org/

Finally, you may be interested to look at a proposal "Defending Against
Guard Discovery Attacks using Vanguards" that is looking to restrict the
second hop used in hidden service circuits to a smaller set.

https://gitweb.torproject.org/torspec.git/tree/proposals/247-hs-guard-discovery.txt

We cannot guarantee that anyone from any Tor team would have time
available to assist you during your project, but we do like it when
people would like to do research on Tor and we will assist you where we
are able. Please do not rely on our availability in your plan however.

Thanks,
Iain.

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 512 bytes
Desc: OpenPGP digital signature
URL: <http://lists.torproject.org/pipermail/metrics-team/attachments/20171218/903101bc/attachment.sig>