MIT Press

Table 2:

Accuracy of membership inference per classifier type, Perceptron (P), Decision Tree (DT), Naïve Bayes (NB), Nearest Neighbors (NN), and Multi-layer Perceptron (MLP). Alice column shows the accuracy of attack on Alice probes

𝒜_{i n_p r o b e}

and

𝒜_{o u t_p r o b e}

⁠. Bob columns show the accuracy on the classifiers’ train, validation, and test set. Note that, following the evaluation protocol explained in Section 4.3, only Carol the evaluator can observe the accuracy of the attacks on Alice model.

	Alice	Bob:train	Bob:valid	Bob:test
P	50.0	50.0	50.0	50.0
DT	50.4	51.4	51.2	51.1
NB	50.4	51.2	51.1	51.0
NN	49.9	61.6	50.5	50.0
MLP	50.2	50.8	50.8	50.8

	Alice	Bob:train	Bob:valid	Bob:test
P	50.0	50.0	50.0	50.0
DT	50.4	51.4	51.2	51.1
NB	50.4	51.2	51.1	51.0
NN	49.9	61.6	50.5	50.0
MLP	50.2	50.8	50.8	50.8

View Large

This Feature Is Available To Subscribers Only