Cobalt Prototype Demo

• WARNING: Do not deploy this code to production---it is not secure! This implementation is intended only as a prototype. In particular the PyCrypto library has not been approved by the Google ISE team.

• Prerequisites

  • Currently this demo is only supported on Ubuntu Linux.

• One-time setup. You must install R and some other packages:

  • cd third_party/rappor
  • ./setup.sh
  • You will be asked to enter your sudo password.
  • The setup script may take a few minutes to install everything.
  • You may see a few error messages but if the script keeps going everything is probably ok.
  • cd ../..

• More one-time setup. You must install the Python cryptography package.

  • sudo apt-get install build-essential libssl-dev libffi-dev python-dev
  • sudo pip install cryptography

• ./cobalt.py build

  • This generates a fastrand python module that wraps a fast C random number generator and a fast_em module that is invoked from RAPPOR R code to perform a fast expectation maximization iteration.

• ./cobalt.py test

  • Runs all tests.

• ./cobalt.py run

  • This generates synthetic data, runs the straight-counting pipeline, runs the Cobalt prototype pipeline, and generates a visualization of the results. You can also run individual steps manually as follows:

  • ./cobalt.py clean

    • This deletes the |out| directory.

  • ./cobalt.py gen

    • This creates an out directory and generates synthetic data in the file input_data.csv in the out directory. This data is the input to both the straight counting pipeline and the Cobalt prototype pipeline.

    • This script also runs the straight counting pipeline that emits several output files into the out directory:

      • popular_help_queries.csv
      • usage_and_rating_by_city.csv
      • usage_by_hour.csv
      • usage_by_module.csv

  • ./cobalt.py randomize

    • This reads input_data.csv and runs all randomizers on that data. This constitutes the first stage of the Cobalt prototype pipeline. A randomizer emits its data to a csv file in the r_to_s subdirectory below out.

  • ./cobalt.py shuffle

    • This runs all shufflers. A shuffler reads randomizer output from the r_to_s directory and writes data in the s_to_a directory.

  • ./cobalt.py analyze

    • This runs all analyzers. An analyzer reads shuffler output from the s_to_a directory and writes final output into the out directory.

  • ./cobalt.py visualize

    • This reads the files output by the straight counting pipeline and the Cobalt prototype pipeline and generates data.js using the Google visualization API.

• Load ./visualization/visualization.html in your browser

  • This displays some charts based on the data in data.js