ml-privacy-csf18

Code for the linear- and tree-model experiments in the CSF 2018 paper "Privacy Risk in Machine Learning: Analyzing the Connection to Overfitting"

• Link to paper
• Link to arXiv version

How to run the code

Note: The Netflix Prize dataset is not included in this repository due to its size

The code was (lightly) tested and should be compatible with both Python 2.7 and Python 3.6 on Bash.

To reproduce the results in the paper, clone the repository and go to the code/ directory. Then, run ./run-errors.sh to compute the training and test errors. This is required before running any of the other experiments. The script should take about 1-2 hours to run.

Membership Inference Attack

Run ./run-membership.sh. The results will be found in results-sklearn/iwpc/membership/ and results-sklearn/eyedata/membership/.

Attribute Inference Attack

Run ./run-attribute.sh cyp2c9 and ./run-attribute.sh vkorc1. The results will be found in results-sklearn/iwpc/attribute/.

Reduction from Membership Inference to Attribute Inference

Run ./run-reduction.sh cyp2c9 and ./run-reduction.sh vkorc1. The results will be found in results-sklearn/iwpc/reduction/.