Recurrent Neural Network Model Based on a New Regularization Technique for Real-Time Intrusion Detection in SDN Environments

Table 2

Comparison of detection performance.

Algorithms	Legitimate class			Anomaly class
	Pr (%)	TPR (%)	F-Score (%)	Pr (%)	TPR (%)	F-Score (%)
Our work (KDD Cup 1999)	94.3	99.5	93.01	95	96	92.21
Our work (NSL-KDD)	92.2	97.43	91	94.57	93.5	91.13
Our work (UNSW-NB15)	93.5	98.4	94.03	95.2	97.01	92.3
VanilaRNN	43	90	58	57	10	17
SVM	71	32	44	64	90	75
DNN	67	89	76	88	66	76
GRU-DNN	87	89	88	91	90	90