/PETS2017-paper

A user study testing and redesigning the Tor configuration launcher.

Primary LanguageBatchfile

Censorship environments:

  1. websites are blocked >> CONNECT == SUCCESS
  2. website and tor public relays are blocked >> CHOOSE ANY BRIDGE == SUCCESS
  3. website, tor public relays, and tor hard coded bridges are blocked >> MEEK OR OWN BRIDGE ADDRESS == SUCCESS

Participant tasks:

  • Go to cnn.com (non-blocked website) and write the headline.
  • Go to wikipedia.org (blocked website) and write down what the topic of the day is.

Hypotheses:

  1. people don't know what a bridge or proxy is. > ask participants to define these terms in the survey.
  2. people don't know what the difference between a bridge and a proxy is > ask if people can distinguish them.
  3. people do not know when bridges or proxies are necessary > ask them this in the survey.
  4. proxies cause confusion when configuring bridges. screen recording: see if people click on proxies (none of the enviornments require a proxy)
  5. people will favor familiar-sounding transports (i.e. meek-google versus meek-azure or scramblesuit) > see screen recordings, ask in survey for which transports were picked in which order and why they chose the final transport.
  6. (some dialogue is redundant. see opening window: "before you connect..." and bridge2 window: "you may use...") > we can try asking if people find these things useful, if they would prefer the interface without it, or if they read it.

Tor Survey Questions:

http://www.surveygizmo.com/s3/2085559/Tor-Usability-Survey/SG_TEST_RUN

Graphvis commands (for generating interface flow diagram):

  • generate png for digraph: dot -Tpng torconfig.dot > torconfig.png
  • generate pdf for digraph: dot -Tpdf torconfig.dot > torconfig.pdf Tor Circumvention UX Experiment

Scope:

  • Tor’s configuration dialogue only, with respect to censorship circumvention ! Not in scope: proxies, user trust models, rest of Tor’s UX, anonymity

Data collected:

  • configuration control flow: via video capture
  • participants information/feedback: via survey

Analyses:

  • show a markov chain version of the configuration tree, before and after changes
  • qualitative analysis on survey answers

Patches to Tor Browser:

https://github.com/nmalkin/tor-launcher (See branches instrumented and uxtest.)