Table 4. Overview of the used hyperparameters per method and possibly per diameter $\mathrm{d}$.

The “b” indicates baseline methods that aim at optimizing general performance without fairness considerations. For more details on these hyperparameters, see subsection “Fair leakage detection instations”.

Method	$c\in [0,\mathrm{\infty })$	$\lambda \in [0,1]$	$\mu$ ( $\mathrm{d}=$5, 10, 15)	$b$	T
TFPR-db (b)	–	–	–	100	0.8
TFPR+COV-db-log	$✓$	–	0.10 0.20 0.20	100	0.8
TFPR-ndb (b)	–	–	–	–	–
TFPR+COV-ndb-log	$✓$	–	0.20 0.25 0.25	–	–
TFPR+COV-ndb-max	$✓$	–	100	–	–
ACC-db (b)	–	–	–	100	0.8
ACC+COV-db-log	$✓$	–	0.15 0.05 0.05	100	0.8
ACC-ndb (b)	–	–	–	–	–
ACC+COV-ndb-log	$✓$	–	0.2 0.3 0.05	–	–
ACC+COV-ndb-max	$✓$	–	100	–	–
COV+ACC-ndb-log	–	$✓$	0.01 0.01 0.01	–	–
COV+ACC-ndb-max	–	$✓$	100	–	–
DI+ACC-ndb-log	–	$✓$	0.05 0.025 0.04	–	–
DI+ACC-ndb-max	–	$✓$	100	–	–