A Deep Intelligent Attack Detection Framework for Fog-Based IoT Systems

Table 2

Performance metrics of considered DLMs.

Dataset name	ML/DL model	Accr (%)	Precision (%)	Recall (%)	F1_S (%)
DDOS SDN	MLP	99.46	99.25	99.4	99.32
	LSTM	99.7	99.6	99.64	99.62
	Bi-LSTM	99.63	99.78	99.29	99.53
	GRU	99.66	99.55	99.6	99.57
	HEM	97.87	99.17	97.28	98.22
	CNN + LSTM	96.43	95.71	95.27	95.49
NSLKDD	MLP	99.05	99.04	99.12	99.08
	LSTM	99.12	99.22	99.08	99.15
	Bi-LSTM	98.95	98.85	99.13	98.99
	GRU	98.43	98.41	98.56	98.48
	HEM	98.28	99.62	96.8	98.19
	CNN + LSTM	98.19	97.79	98.72	98.25
UNSW-NB15	MLP	93.41	95.74	93.51	94.61
	LSTM	94.11	95.87	94.47	95.16
	Bi-LSTM	94.04	95.86	94.43	95.14
	GRU	93.57	95.75	93.76	94.74
	HEM	94.05	92.08	92.37	92.22
	CNN + LSTM	92.85	94.37	94.05	94.21
IoTID20	MLP	99.84	99.7	97.84	98.76
	LSTM	99.88	99.77	98.4	99.08
	Bi-LSTM	99.86	99.57	98.27	98.92
	GRU	99.84	99.61	98.01	98.8
	HEM	99.84	99.83	99.99	99.92
	CNN + LSTM	99.76	99.7	96.58	98.11