Table 7. MIA result summary of trained attack on breast cancer dataset for moderate noise level.

Moderate noise ( $\epsilon =1,\delta ={10}^{-4}$) Attack type = trained attack
Slice feature/	Attacker	AUC	Precision	Recall	F1 score
Slice value	advantage
Entire dataset	0.07	0.49	0.50	0.49	0.49
Class 0	0.38	0.42	0.53	0.46	0.49
Class 1	0.24	0.54	0.46	0.51	0.49
Correctly classified	0.21	0.47	0.53	0.49	0.51
True
Correctly classified	0.48	0.45	0.59	0.45	0.51
False