weihongli / regression-model-fitting-under-differential-privacy-and-model-inversion-attack-source-code Goto Github PK

Model_Accurancy_Without_Noise ---> compute the accuracy of released logistic model. 
Model_Inversion_Without_Noise ---> compute the accuracy of model inversion attack without noise added. 
Same_epsilon_value ---> add same noise into all attributes and test the accuracy of model inversion attack. 
different_epsilon_value ---> add different noise into each attribute(more noise into sensitve attribute, less noise into non-senstive attribute), 
                             and test the accuracy of model inversion attack.  

Functional Mechnism % Copyright 2012 Jun Zhang
Model Inversion and all rest of parts % Copyright 2015 Cheng Si