Network Traffic Anomaly Detection Model Based on Feature Reduction and Bidirectional LSTM Neural Network Optimization

Table 11

Detection results of the six models.

Dataset	Model	Evaluation metrics
		Accuracy (%)	Precision (%)	Recall (%)	F-score
NSL-KDD	HCRNNIDS	84.46	82.15	92.14	86.86
	ADASYN-LightGBM	86.25	82.97	94.05	88.16
	LNNLS-KH	89.16	83.64	96.44	89.58
	STL-HDL	91.02	82.35	94.98	88.21
	E-GraphSAGE	90.20	83.92	96.03	89.57
	FR-APPSO-BiLSTM	91.76	85.37	98.50	91.46
UNSW-NB1	HCRNNIDS	88.28	85.92	92.43	89.06
	ADASYN-LightGBM	89.36	90.27	91.44	90.85
	LNNLS-KH	90.44	94.35	91.08	92.69
	STL-HDL	90.29	92.19	96.14	94.12
	E-GraphSAGE	91.42	93.46	94.43	93.94
	FR-APPSO-BiLSTM	92.08	97.88	98.32	98.10
CICIDS-2017	HCRNNIDS	85.90	94.34	90.44	92.35
	ADASYN-LightGBM	92.00	97.30	95.44	96.36
	LNNLS-KH	91.57	89.31	90.35	89.82
	STL-HDL	90.46	95.03	93.23	94.12
	E-GraphSAGE	93.26	92.35	94.51	93.41
	FR-APPSO-BiLSTM	95.44	98.58	98.40	98.49