From: PUMD: a PU learning-based malicious domain detection framework
Method | Labeled sample | Unlabeled sample | TrainSet | TestSet | ||
---|---|---|---|---|---|---|
Mali DN | Benign DN | P | N | |||
PUMD | \(P = c *Mali\) | â | \(U = D - P\) | W(P) | \(\cup w(U)\) | U |
PU_biased | W(U) | |||||
PU_empirical | \(W(P) \cup W(U)\) | |||||
mix_supervised* | Â | \(N_{mix} = Sample_{size(P)}(D-P)\) | \(D - P - N_{mix}\) | P | \(N_{mix}\) | |
pure_supervised | Â | \(N_{pure} = Sample_{size(P)}(Benign)\) | \(D - P - N_{pure}\) | \(N_{pure}\) | ||
unsupervised | â | â | D | D |