Adaptive DDoS Attack Detection Method Based on Multiple-Kernel Learning

Table 1

Comparison results of four algorithms for scaling attack flow and normal flow.

		The value of the random multiplier
		0.6–0.7	0.7–0.8	0.8–0.9	0.9–1.1	1.0–1.5	1.5–2.0	2.0–3.0	3.0–4.0	4.0–5.0
ADADM method	DR (%)	78.57	78.57	78.57	78.57	78.21	78.21	78.21	78.21	78.57
	FR (%)	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01
	ER (%)	12.22	12.22	12.22	12.22	12.42	12.42	12.42	12.42	12.22
SimpleMKL method	DR (%)	76.43	76.43	76.43	76.43	76.43	76.43	76.43	76.43	76.43
	FR (%)	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01
	ER (%)	13.44	13.44	13.44	13.44	13.44	13.44	13.44	13.44	13.44
SVM method	DR (%)	77.50	77.50	77.86	77.86	76.79	77.50	77.86	76.79	77.86
	FR (%)	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01	0.01
	ER (%)	12.83	12.83	12.63	12.63	13.24	12.83	12.63	13.24	12.63
Nezhad et al.’s [16] method	DR (%)	98.21	97.85	98.21	97.85	98.21	98.21	98.21	97.85	98.21
	FR (%)	74.29	74.76	74.29	74.76	75.71	74.29	74.29	74.29	72.38
	ER (%)	32.92	33.33	32.92	33.33	33.54	32.92	32.92	33.13	32.11