[tor-reports] George's status report: September 2014

[George Kadianakis]

# Activities of September 2014

- Continued work on the guardfraction project: #9321. Please refer to
  last month's report for the status of the project:
  https://lists.torproject.org/pipermail/tor-reports/2014-September/000644.html).

  I sent a mail to [tor-dev] asking dirauths for opinion on the
  guardfraction script:
  https://lists.torproject.org/pipermail/tor-dev/2014-September/007489.html

  I got very helpful feedback, and with the help of weasel and
  Sebastian I refactored my script to use an sqlite database instead
  of the awkward summary files:
  https://lists.torproject.org/pipermail/tor-dev/2014-September/007526.html

  I published the new script and asked for reviews. Nick Hopper
  reviewed the script and found a few flaws that I should plug:
  https://trac.torproject.org/projects/tor/ticket/13125

  I also worked on the little-t-tor side of this project, and wrote
  unittests and documentation. I need to give it more testing with
  chutney, rebase and publish for review.

  During working on the little-t-tor side I found a few bugs:
    #13064 [Tor]: TestingDirAuthVoteGuard doesn't work with nicknames
    #13066 [Tor]: smartlist_choose_node_by_bandwidth(q) does not consider bad exit
    #13297 [Tor]: compute_weighted_bandwidths() broken for dirauths 
  
- I also finished the little-t-tor Tor2Web patch that allows Tor2web
  clients to hardcode a Rendezvous Point which could potentially speed
  up circuit establishment for Tor2web:
  https://trac.torproject.org/projects/tor/ticket/12844

- While working on the Tor2web patch, I found a curious HS bug: #13151

  Basically, HS clients sent the IP address of RPs in host order,
  which meant that HSes tried to extend to the reverse IP (4.3.2.1
  instead of 1.2.3.4). Still, because of the canonical connections
  feature of Tor, the HS circuit establishment would succeed after a
  few failed INTRODUCE2 cells.

  It's possible that if you are using the latest patched TBB you might
  experience better performance during HS circuit establishment
  might. This bug discovery was well timed with the "increase
  client-side HS performance" sponsor project :)

- Wrote a bit more about the guard discovery problem:
  https://lists.torproject.org/pipermail/tor-dev/2014-September/007472.html
  which received good feedback from Roger, showing avenues for further
  research.

  This problem is something that should be fixed along with the
  upcoming next gen hidden services project.

- Reviewed a few tickets: 

    #12980: Implement ed25519 primitives for proposals 220, 224, 228 - https://bugs.torproject.org/12980
    #13213: Tor should tell its pluggable transports when DisableNetwork gets set/unset - https://bugs.torproject.org/13213
    #6456: Merge parse_client_transport_line() and parse_server_transport_line() - https://bugs.torproject.org/6456

- Helped Juha with his post-GSoC ahmia blog post:
  https://blog.torproject.org/blog/ahmia-search-after-gsoc-development

- Helped Arturo with the bridge reachability project. Also, debugged
  an OONI meter in China that was dying because the OOM killer was
  killing obfsproxy.

# Activities for October 2014 

- Work on the little-t-tor side of the guardfraction stuff. And also
  finalize the python script. The goal is to have everything for
  review by mid-October.

- Think more about guard discovery attacks and HSes.

- Help with the bridget OONI test.