This repository provides framework for testing ml-models for vulnerability to membership inference attack
Based on spring-epfl/mia repository.
Supported models:
- keras neural networks
- tensorflow neural networks
- sklearn models
There are 4 different libs for 4 different types of target models:
- estimators_bin_nn - for neural networks of binary classification
- estimators_bin_skl - for sklearn binary classificators
- estimators_multi_nn - for multiclass neural networks
- estimators_multi_skl - for multiclass sklearn-models
To use this framework you just need to pull membership_inference/MembershipInferenceTest and import MembershipInferenceTest
You can see some examples of using the tool in membership_inference/experiments