Algorithm 1 Detector Training Algorithm.

Require:  Normal Data $D_n$, Generator $\mathcal{G}$, Target Model $\mathcal{F}$   1:$D_m\leftarrow \mathcal{G}\left(D_n\right)$   2:$D\leftarrow$ Merge$(D_m,D_n)$   3:$D_{out}\leftarrow \mathcal{F}\left(D\right)$   4:$D_{train}\leftarrow$ Top$(k,D_{out})$   5:$\mathcal{D}\leftarrow$ Train$(\mathcal{D},D_{train})$