BLATTA: Early Exploit Detection on Network Traffic with Recurrent Neural Networks

Table 7

Comparison to previous works using the UNSW-NB15 dataset as the testing set.

Method	Detection rate (%)		FPR (%)
	Exploits in UNSW-NB15	Worms in UNSW-NB15
Blatta	99.04	100	1.93
PAYL	87.12	26.49	0.05
OCPAD	10.53	4.11	0
Decanter	67.93	90.14	0.03
Autoencoder	47.51	81.12	0.99