VHDRA: A Vertical and Horizontal Intelligent Dataset Reduction Approach for Cyber-Physical Power Aware Intrusion Detection Systems

Table 4

A comparison between the VHDRA vertical reduction and the other existing approaches.

	Binary class		Triple class		Multiclass
	Size: 81,664 MB, 128 features		Size: 81,669 MB, 128 features		Size: 78,963 MB, 128 features
Approach	No. of features	Detection rate (%)	No. of features	Detection rate (%)	No. of features	Detection rate (%)
VHDRA (PSO+ EPA)	15	99.21	14	98.91	18	97.03
	(9.57MB)	EPA Accu. gain:0.83	(8.93MB)	EPA Accu. gain:0.90	(11.10MB)	EPA Accu. gain: 3.0
VHDRA (PSO)	17	98.38	19	98.01	22	94.03
	(10.85MB)		(12.12MB)		(13.57MB)
NNGE (CFS)	28	94.68	28	94.62	28	87.77
	(17.86MB)		(17.91MB)		(17.77MB)
IBL	28	97.20	17	97.38	28	92.10
	(17.97MB)		(10.87MB)		(17.62MB)
J48	28	93.69	28	93.69	28	84.45
	(18.18MB)		(17.96MB)		(18.92MB)
Random Forest	28	96.77	28	96.77	28	90.41
	(18.29MB)		(18.20MB)		(17.54MB)
JRip	28	91.20	28	91.22	129	73.94